(Md. Rabiul Islam MMC Design

Green Energy and Technology
Md. RabiulIslam
YouguangGuo
JianguoZhu
Power Converters
for Medium
Voltage
Networks
Green Energy and Technology
More information about this series at http://www.springer.com/series/8059
Md. Rabiul Islam Youguang Guo Jianguo Zhu
Power Converters for

Medium Voltage Networks
13
Md. Rabiul Islam
Youguang Guo
Jianguo Zhu
School of Electrical, Mechanical and
Mechatronic Systems
Faculty of Engineering and Information
Technology
University of Technology Sydney (UTS)
Ultimo
NSW
Australia
ISSN 1865-3529 ISSN 1865-3537 (electronic)

ISBN 978-3-662-44528-0 ISBN 978-3-662-44529-7 (eBook)
DOI 10.1007/978-3-662-44529-7
Library of Congress Control Number: 2014947125
Springer Heidelberg New York Dordrecht London
Springer-Verlag Berlin Heidelberg 2014

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts
in connection with reviews or scholarly analysis or material supplied specifically for the purpose of
being entered and executed on a computer system, for exclusive use by the purchaser of the work.
Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright
Law of the Publishers location, in its current version, and permission for use must always be obtained
from Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance
Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
Medium and large-scale renewable power plants require large areas of land, and
are usually installed in offshore or remote areas, far from cities, in order to cap-
ture large amount of energy. For power transmission and grid integration, the low
voltage, e.g., 690 V for a typical wind turbine generator, at which the electricity is
generated from renewable energy sources is transformed to the medium-voltage
(e.g., 1133-kV) commonly by using the step-up transformers. In many cases, the
power transformers operated at the frequency of 50/60Hz are heavy, large, and
inefficient. These technological drawbacks would have significant impacts on off-
shore and remote renewable power plants where the costs of installation and regu-
lar maintenance are extremely high.
With the fast development of power electronics, it is becoming a reality to
replace the combination of low voltage inverter and step-up-transformer by a
medium-voltage converter for direct grid connection to reduce the system volume
and weight, as well as the cost. Traditionally, two- and three-level converters are
commonly used for high power applications. When applied to medium-voltage
systems, very expensive semiconductor switching devices of high-voltage ratings
are required. To reduce the cost, multilevel converters made of cascaded switch-
ing devices of relatively low-voltage ratings are developed. In comparison with
the conventional two- and three-level converters, the multilevel converters present
lower switching losses, lower voltage stress on switching devices, and higher qual-
ity output power, and thus suit better the medium-voltage applications. Although
several multilevel converter topologies have been developed in the last few dec-
ades, most of them are not suitable for medium-voltage applications as their
number of auxiliary components scales quadratically with the number of levels.
Because of some special features (e.g., the number of components scales linearly
with the number of levels, and individual modules are identical and completely
modular in construction hence enabling high-level number attainability) the modu-
lar multilevel cascaded (MMC) converter topology can be considered as a possible
candidate for medium voltage applications. However, the MMC converter requires
multiple isolated DC sources that must be balanced, and as a result its application
is not straightforward, especially in wind power generation systems. Moreover,
v
vi Preface
the multilevel converter requires a number of switching and control PWM signals,
which cannot be generated by a digital signal processor (DSP) because the cur-
rently available DSP can only provide about six pairs of PWM channels. In this
instance, a field programmable gate array (FPGA) becomes the natural choice.
Most of the available design techniques require special software such as HDL
coder, System generator, PSIM and ModelSim, which increases the developmental
time and cost.
Chapter 1 of this book discusses various aspects, such as historical growth of
two dominating renewable, i.e., wind and solar, energy sources, the existing tech-
nologies, technical challenges, and possible solutions for large-scale power gen-
eration from these energy sources.
An extensive literature survey has been conducted in Chap.2 focusing on vari-
ous aspects of medium-voltage converter development for step-up-transformer-less
direct grid integration of photovoltaic (PV) power plants. The main objective is
to show how power electronic converter topologies, power electronic devices, and
control complexities have affected the development of medium-voltage converters,
and how to make an excellent choice of the suitable converter topologies for step-
up-transformer-less grid integration through medium-voltage converters, which is
really a critical problem and highly affects the converter performance and cost.
The main aim of Chap.3 is to find out a suitable converter topology, which can
interconnect the renewable generation units directly to the medium-voltage grid
with the commercially available matured semiconductor devices. Different multi-
level converter topologies, such as the neutral point clamped (NPC), flying capaci-
tor (FC), and MMC converters, have been considered and compared for the design
of an 11-kV converter system. The comparison is made in terms of the number of
semiconductors, semiconductor cost and commercial availability, total harmonic
distortions (THDs), filter size and control complexity of the converters. The per-
formance is analyzed and compared in the MATLAB/Simulink environment.
To couple a renewable energy source to the MMC converter, a high-frequency
magnetic-link with multiple secondary windings can be an excellent option and
is explored in Chap.4. The high-frequency magnetic-link is used to generate the
isolated balanced multiple DC supplies for all of the H-bridge inverter cells of
the MMC converter from a single low voltage power source. Compared with the
conventional power frequency transformers operated at 50 or 60Hz, the high-fre-
quency magnetic-links (in the range of a few kilohertz to megahertz) have much
smaller and lighter magnetic cores and windings, and thus much lower costs.
The capability of parallel processing of the FPGA affords the opportunity to
the switching controller to update all gate signals simultaneously. Various design
techniques and software environments are available for the modeling of switching
control schemes with the FPGA technology. Most of the techniques require spe-
cial software, which increases the development time and cost. In Chap.5, the most
common software such as the MATLAB/Simulink and Xilinx ISE-based alterna-
tive design technique is proposed, which may reduce the developmental time and
cost of the switching controller.
Preface vii
To verify the feasibility of the high-frequency MMC converter, a scaled down

1-kV laboratory prototype test platform with a 5-level MMC converter is devel-
oped. Furthermore, the prototype design and implementation, test platform, and
experimental results are analyzed and discussed in Chap. 6. The component selec-
tion, converter fabrication, and experimentation techniques are equally applicable
to any other converter applications.
In Chap. 7, an 11-kV system and a 33-kV system are designed and analyzed
taking into account the specified system performance, control complexity, and cost
and market availability of the power semiconductors. It is found that, the 19-level
and 43-level converters are the optimal choice for the 11-kV and 33-kV systems,
respectively. Besides the design and analysis of medium voltage converters, the
traditional low voltage converters with power frequency step-up-transformers are
also discussed.
Chapter 8 concludes the book. Future directions have been recommended for
further research and development. It is expected that the technology presented in
this book has a great potential to be implemented in future wind farms, PV power
plants, and smart micro-grid applications.
Contents
1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Most Dominating Renewable Power Sources . . . . . . . . . . . . . . . . . . 1
1.1.1 Historical Growth of Wind Power Generation Capacity . . . . 2
1.1.2 Technical Challenges and Possible Solutions . . . . . . . . . . . . 3
1.1.3 Historical Growth of Solar PV Power Generation Capacity. . . 7
1.1.4 Technical Challenges of Solar PV Power
Generation Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.1.5 Possible Solution to Technical Challenges for
Large-Scale Renewable Generation Systems. . . . . . . . . . . . . 10
1.2 Major Objectives of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Contribution of the Book. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Organization of the Book. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2 Power Converters for Small- to Large-Scale Photovoltaic

Power Plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Solar Photovoltaic Arrays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.1 Solar Photovoltaic Array Modeling. . . . . . . . . . . . . . . . . . . . 22
2.2.2 Solar Photovoltaic Array Characteristics. . . . . . . . . . . . . . . . 24
2.2.3 Solar Photovoltaic Array Maximum Power Point . . . . . . . . . 25
2.3 Inverters in Small-Scale Solar PV Systems. . . . . . . . . . . . . . . . . . . . 28
2.3.1 Two-Stage Solar PV Inverters for Small-Scale Systems. . . . 28
2.3.2 Multiple-Stages Solar PV Inverters for Small-Scale Systems. . . 31
2.3.3 Single-Stage Solar PV Inverter for Small-Scale Systems . . . 34
2.3.4 ABB and Siemens Solar PV Inverters
for Small-Scale Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Medium- and Large-Scale Solar PV Systems . . . . . . . . . . . . . . . . . . 38
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
ix
x Contents
3 Power Converter Topologies for Grid-Integrated

Medium-Voltage Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Two-Level Converters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.1 Single-Phase Converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.2 Three-Phase Converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2.3 Space Vector Pulse Width Modulation (SV-PWM). . . . . . . . 60
3.3 Multilevel Converters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.3.1 Neutral Point Clamped Converter . . . . . . . . . . . . . . . . . . . . . 78
3.3.2 Flying Capacitor Converter . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.3.3 Modular Multilevel Cascaded Converter. . . . . . . . . . . . . . . . 81
3.3.4 Switching Schemes of Multilevel Converters . . . . . . . . . . . . 83
3.4 Selection of Multilevel Converter Topology
for Medium-Voltage Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.4.1 Detailed Design and Analysis of 5-Level
and 11-Level NPC Converters. . . . . . . . . . . . . . . . . . . . . . . . 97
and 11-Level FC Converters. . . . . . . . . . . . . . . . . . . . . . . . . . 99
and 11-Level MMC Converters. . . . . . . . . . . . . . . . . . . . . . . 101
3.4.4 Performance Analysis and Topology Selection. . . . . . . . . . . 102
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4 Design and Characterization of High-Frequency Magnetic

Links Used in Power Electronic Converters. . . . . . . . . . . . . . . . . . . . . . 109
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.2 Design of High-Frequency Magnetic Links. . . . . . . . . . . . . . . . . . . . 113
4.2.1 Flowchart of High-Frequency Magnetic-Link Design. . . . . . 114
4.2.2 Core Material Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.2.3 Calculation of Number of Turns. . . . . . . . . . . . . . . . . . . . . . . 118
4.2.4 Winding Wire Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.2.5 Parasitic Calculation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.2.6 Design Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.2.7 Transformer Core Development. . . . . . . . . . . . . . . . . . . . . . . 131
4.3 Characterization of High-Frequency Magnetic Links. . . . . . . . . . . . 136
4.3.1 Windings DC Resistance Calculation . . . . . . . . . . . . . . . . . . 136
4.3.2 Voltage Transformation Ratio Calculation. . . . . . . . . . . . . . . 138
4.3.3 Measurement of Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.3.4 Measurement of BH Characteristics. . . . . . . . . . . . . . . . . . . 144
4.3.5 Measurement of DC-Link Voltages. . . . . . . . . . . . . . . . . . . . 149
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Contents xi
5 FPGA-Based Digital Switching Controller for Multilevel Converters. . . 153

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.1.1 ASIC Technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.1.2 DSP Technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.1.3 FPGA Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
5.2 FPGA-Based Switching Controller Design Techniques. . . . . . . . . . . 158
5.3 Modeling and Schematic Symbol Creation Using VHDL. . . . . . . . . 163
5.3.1 Carrier Signal Generation Unit. . . . . . . . . . . . . . . . . . . . . . . . 163
5.3.2 Reference Signal Generation Unit. . . . . . . . . . . . . . . . . . . . . 165
5.3.3 Comparator Unit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.4 Model of Complete Switching Controller . . . . . . . . . . . . . . . . . . . . . 173
5.5 Behavioral Simulation of Switching Controller. . . . . . . . . . . . . . . . . 174
5.6 Model Implementation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.7 Design Verification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.8 FPGA Programming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.9 Experimental Testing and Verifications . . . . . . . . . . . . . . . . . . . . . . . 182
5.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
6 Experimental Validation of 1-kV Modular Multilevel

Cascaded Converter with High-Frequency Magnetic Link . . . . . . . . . 189
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.2 Development of High-Frequency Magnetic-Link MMC Converter. . . 194
6.2.1 Development of a 3-Phase 5-Level MMC Converter. . . . . . . 194
6.2.2 Design of Driver Circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.2.3 Development of High-Frequency Rectifiers
with Fast Recovery Diodes. . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.2.4 Development of High-Frequency Inverter. . . . . . . . . . . . . . . 203
6.2.5 Development of High-Frequency Magnetic Link . . . . . . . . . 208
6.3 Experimental Testing and Performance Analysis. . . . . . . . . . . . . . . . 209
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7 Design and Analysis of 11- and 33-kV Modular

Multilevel Cascaded Converters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.1.1 Traditional Converters for Wind Power Generation Systems. . . 228
7.1.2 Traditional Converters for Solar PV Power Plants. . . . . . . . . 231
7.1.3 Power Transformers in Traditional Converter-Based
Renewable Generation Systems. . . . . . . . . . . . . . . . . . . . . . . 234
7.1.4 Medium-Voltage Converters for Direct Grid Integration. . . . 240
7.2 Design and Analysis of 11-kV Converter Systems . . . . . . . . . . . . . . 242
7.2.1 Design and Analysis of 9-Level 11-kV Converter. . . . . . . . . 243
7.2.2 Design and Analysis of 11-Level 11-kV Converter. . . . . . . . 246
xii Contents
7.2.3
Design and Analysis of 15-Level 11-kV Converter. . . . . . . . 248
7.2.4
7.2.5
7.2.6
Selection of Number of Levels for 11-kV
Converter Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
7.3 Design and Analysis of 33-kV Converter Systems . . . . . . . . . . . . . . 258
7.3.6 Selection of Optimal Number of Levels
for 33-kV Converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
8 Conclusions and Future Works. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
8.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
8.3 Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Figures
Fig. 1.1 A photograph of Anholt offshore wind power plant was

officially inaugurated in September 2013 [16]. . . . . . . . . . . . . . . . . 3
Fig. 1.2 An ideal model of renewable power source-based smart
power systems [17]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Fig. 1.3 Wind turbine with step-up transformer . . . . . . . . . . . . . . . . . . . . . . 5
Fig. 1.4 A photograph of installation process of offshore wind
power plant with the vessel Aeolus [24] . . . . . . . . . . . . . . . . . . . . . 6
Fig. 1.5 A photograph of offshore wind farm under maintenance [25]. . . . . 6
Fig. 1.6 A photograph of Beneixama 20 MW PV power plant [27]. . . . . . . 8
Fig. 1.7 A photograph of step-up transformer installation of
Tritec 5.2 MW solar PV power plant [30]. . . . . . . . . . . . . . . . . . . . 9
Fig. 2.1 World oil consumption [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Fig. 2.2 World coal consumption [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Fig. 2.3 World natural gas consumption [1] . . . . . . . . . . . . . . . . . . . . . . . . . 18
Fig. 2.4 World CO2 emission form fossil fuel burning [1] . . . . . . . . . . . . . . 19
Fig. 2.5 Average annual world energy growth rates by source [1]. . . . . . . . 19
Fig. 2.6 Average photovoltaic module cost [1]. . . . . . . . . . . . . . . . . . . . . . . 20
Fig. 2.7 A photograph of PV cell, which may produce 0.5 V DC. . . . . . . . . 21
Fig. 2.8 A photograph of 24-V PV module, where 72 cells are
interconnected . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Fig. 2.9 A photograph of PV array with six 12-V modules,
where 36 cells are interconnected in each module. . . . . . . . . . . . . . 22
Fig. 2.10 Equivalent circuit of a PV cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Fig. 2.11 Simplified equivalent circuit of a PV cell . . . . . . . . . . . . . . . . . . . . 24
Fig. 2.12 Monthly average global solar insolation at Dhaka City
of Bangladesh; recording period: 19881998 [11] . . . . . . . . . . . . . 25
Fig. 2.13 Daily average bright sunshine hour at Dhaka City
of Bangladesh; recording period: 19611980 [11] . . . . . . . . . . . . . 25
Fig. 2.14 Calculated IV curves at different irradiance ranging
from 600 to 1,000 W/m2; 25 C temperature was
considered for the calculation [12, 13]. . . . . . . . . . . . . . . . . . . . . . . 26
xiii
xiv Figures
Fig. 2.15 Calculated PV curves at different irradiance ranging

from 600 to 1,000 W/m2; 25 C temperature was
Fig. 2.16 Calculated IV curves at different temperature ranging
from 25 to 75 C; irradiance 1,000 W/m2 was
Fig. 2.17 Calculated PV curves at different temperature ranging
from 25 to 75 C; irradiance 1,000 W/m2 was
Fig. 2.18 Maximum power point (MPP) determination: The slope
is positive on the left side of the PV curve, negative
on the right side, and zero at the MPP [12, 13]. . . . . . . . . . . . . . . . 27
Fig. 2.19 Fundamental circuit of two-stage PV inverter [14]. . . . . . . . . . . . . 28
Fig. 2.20 Most commonly used DC to DC boost converter circuit. . . . . . . . . 28
Fig. 2.21 Equivalent circuit of a DC to DC boost converter
when transistor is turned on. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Fig. 2.22 Equivalent circuit of a DC to DC boost converter
when the transistor is turned off. . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Fig. 2.23 Voltage across inductor L for time period T. . . . . . . . . . . . . . . . . . . 29
Fig. 24 Waveform of inductor current in the continuous conduction
mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Fig. 2.25 Time-sharing chopper-based-modified circuit of two-stage
PV inverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Fig. 2.26 Parallel resonant DC-link-based-modified two-stage
PV inverter circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Fig. 2.27 Basic circuit of half-bridge topology-based PV inverter. . . . . . . . . 32
Fig. 2.28 Sunny Boy 5000TL model half-bridge topology-based
PV inverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Fig. 2.29 High-frequency transformer-link (HFT-L)-based
PV inverter power circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Fig. 2.30 High-frequency (HF) transformer-based PV inverter
power circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Fig. 2.31 High-frequency transformer (HFT)-based PowerLynx
Powerlink PV 4.5 kW inverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Fig. 2.32 PV-battery-powered DC bus system . . . . . . . . . . . . . . . . . . . . . . . . 35
Fig. 2.33 Full-bridge with MPPT-based circuit topology
of single-stage inverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Fig. 2.34 Circuit topology of single-stage inverter: NPC with
grid neutral connected to the middle point of DC link. . . . . . . . . . . 36
Fig. 2.35 NPC topology-based three-phase PV inverter. . . . . . . . . . . . . . . . . 36
Fig. 2.36 Single-stage power circuit with boost converter . . . . . . . . . . . . . . . 37
Fig. 2.37 Single-stage power circuit with universal converter . . . . . . . . . . . . 37
Fig. 2.38 ABB PVS300 string inverter design and grid connection. . . . . . . . 38
Fig. 2.39 Siemens SINVERT PVM string inverter design
and grid connection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Figures xv
Fig. 2.40 Centralized PV inverter topology. . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Fig. 2.41 String PV inverter topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Fig. 2.42 Battery-integrated boost converter. . . . . . . . . . . . . . . . . . . . . . . . . . 40
Fig. 2.43 Battery-integrated boost converter-based PV inverter system. . . . . 40
Fig. 2.44 ABB central inverter design and medium-voltage
(MV) grid connection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Fig. 2.45 Siemens central inverter design and medium-voltage
grid connection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Fig. 2.46 Quasi-Z source converter-based medium-voltage
PV inverter (one of 3 phases). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Fig. 2.47 Multiphase isolated converter-based medium-voltage
PV inverter (one of 3 phases). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Fig. 2.48 Common DC-link-based medium-voltage PV inverter
(one of 3 phases) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Fig. 2.49 Common magnetic-link-based 3-phase medium-voltage
PV inverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Fig. 3.1 Circuit diagram of half-bridge voltage source converter. . . . . . . . . 53
Fig. 3.2 Modulation scheme of half-bridge PWM converter. . . . . . . . . . . . . 54
Fig. 3.3 Output voltage waveform of half-bridge PWM converter. . . . . . . . 54
Fig. 3.4 Frequency spectrum of output voltage of half-bridge
PWM converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Fig. 3.5 Circuit diagram of full-bridge PWM converter. . . . . . . . . . . . . . . . 56
Fig. 3.6 Output voltage waveform of full-bridge PWM converter . . . . . . . . 57
Fig. 3.7 Frequency spectrum of output voltage of full-bridge
PWM converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Fig. 3.8 Output voltage of full-bridge PWM converter with
square wave approximation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Fig. 3.9 Circuit diagram of three-phase two-level converter. . . . . . . . . . . . . 59
Fig. 3.10 Modulation scheme of three-phase two-level converter. . . . . . . . . . 59
Fig. 3.11 Output line voltage of three-phase two-level converter. . . . . . . . . . 60
Fig. 3.12 Frequency spectrum of output line voltage of three-phase
two-level converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Fig. 3.13 Space vector diagram of three-phase two-level converter. . . . . . . . 61
Fig. 3.14 The dq plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Fig. 3.15 d-axis component calculation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Fig. 3.16 q-axis component calculation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Fig. 3.17 Algorithm to select the sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Fig. 3.18 Seven-state switching pattern at sector 1. . . . . . . . . . . . . . . . . . . . . 65
Fig. 3.19 Output line-to-line voltage with seven-state switching pattern. . . . 65
Fig. 3.20 Frequency spectrum of output voltage with seven-state
switching pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Fig. 3.21 Line current with seven-state switching pattern. . . . . . . . . . . . . . . . 66
Fig. 3.22 Three-state switching pattern (only one zero vector). . . . . . . . . . . . 67
Fig. 3.23 Output line-to-line voltage with three-state switching
pattern (only one zero vector) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
xvi Figures
Fig. 3.24 Frequency spectrum of output line-to-line voltage with

three-state switching pattern (only one zero vector) . . . . . . . . . . . . 68
Fig. 3.25 Line current with three-state switching pattern
(only one zero vector) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Fig. 3.26 Switching pattern of three-state switching (both zero
vectors alternatively) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Fig. 3.27 Output line-to-line voltage of three-phase two-level
converter with three-state switching (both zero
vectors alternatively) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Fig. 3.28 Harmonic spectrum of line-to-line voltage with three-state
switching (both zero vectors alternatively) . . . . . . . . . . . . . . . . . . . 70
Fig. 3.29 Line current of three-phase two-level converter with
three-state switching (both zero vectors alternatively). . . . . . . . . . . 70
Fig. 3.30 Four-state switching pattern. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Fig. 3.31 Output line-to-line voltage of three-phase two-level
converter with four-state switching pattern. . . . . . . . . . . . . . . . . . . 71
Fig. 3.32 Harmonic spectrum of line-to-line voltage with four-state
switching pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Fig. 3.33 Line current of three-phase two-level converter with
four-state switching pattern. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Fig. 3.34 Five-level power converter circuit topologies [2], a NPC
converter topology, b FC converter topology,
and c MMC converter topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Fig. 3.35 Line-to-line output voltage waveform of five-level converter. . . . . 75
Fig. 3.36 Staircase output phase voltage waveform of five-level converter . . 75
Fig. 3.37 THDs of 33 and 11 kV multilevel converters [3]. . . . . . . . . . . . . . . 76
Fig. 3.38 THDs of 33 and 11 kV multilevel converters [3]. . . . . . . . . . . . . . . 78
Fig. 3.39 Three-phase seven-level NPC converter circuit topology . . . . . . . . 79
Fig. 3.40 Output line-to-line voltage of seven-level NPC converter. . . . . . . . 81
Fig. 3.41 Three-phase seven-level FC converter circuit topology. . . . . . . . . . 82
Fig. 3.42 Three-phase seven-level modular multilevel
cascaded converter topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Fig. 3.43 Phase-shifted switching scheme for the three-phase
seven-level FC and MMC multilevel converters. . . . . . . . . . . . . . . 85
Fig. 3.44 MATLAB/Simulink model to generate phase-shifted
carrier signals for seven-level converter. . . . . . . . . . . . . . . . . . . . . . 85
Fig. 3.45 Three-generated phase-shifted carrier signals. . . . . . . . . . . . . . . . . 86
Fig. 3.46 Three-inverted phase-shifted carrier signals . . . . . . . . . . . . . . . . . . 86
Fig. 3.47 Three reference signals and six phase-shifted carrier
signals-based modulation scheme of three-phase
seven-level converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Fig. 3.48 Generation of gate pulse PS1 with phase-shifted carrier. . . . . . . . . 87
Fig. 3.49 Generation of gate pulse PS2 with phase-shifted carrier. . . . . . . . . 87
Fig. 3.50 MATLAB/Simulink model to generate gate pulses for
three-phase seven-level converter. . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Figures xvii
Fig. 3.51 Three reference signals and four level-shifted carrier signals. . . . . 90
Fig. 3.52 Level-shifted modulation technique for a three-phase
seven-level NPC converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Fig. 3.53 Generation of gate pulse PS7 with level-shifted carrier. . . . . . . . . . 92
Fig. 3.54 Generation of gate pulse PS2 with level-shifted carrier. . . . . . . . . . 92
Fig. 3.55 MATLAB/Simulink model to generate
sinusoidal reference signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Fig. 3.56 SPWM modulation scheme for three-phase five-level converter. . . 93
Fig. 3.57 MATLAB/Simulink model to generate third-harmonic
injected sinusoidal reference signals for three-phase
five-level converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Fig. 3.58 Third-harmonic injected pulse width modulation
schemes for three-phase five-level converter. . . . . . . . . . . . . . . . . . 94
Fig. 3.59 MATLAB/Simulink model to generate trapezoidal
reference signals for three-phase five-level converter. . . . . . . . . . . 94
Fig. 3.60 Trapezoidal pulse width modulation schemes for
three-phase five-level converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Fig. 3.61 MATLAB/Simulink model to generate sixty degree
sinusoidal reference signals for three-phase five-level converter . . . 95
Fig. 3.62 Sixty degree pulse width modulation schemes
for three-phase five-level converter. . . . . . . . . . . . . . . . . . . . . . . . . 95
Fig. 3.63 Calculated THD at different level numbers ranging
from seven-level to 19-level [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Fig. 3.64 Line voltage of five-level neutral point clamped converter. . . . . . . 98
Fig. 3.65 Frequency spectrum of five-level neutral point
clamped converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Fig. 3.66 Line voltage of 11-level neutral point clamped converter. . . . . . . . 99
Fig. 3.67 Frequency spectrum of 11-level neutral point clamped converter. . . 100
Fig. 3.68 Line voltage of five-level flying capacitor converter. . . . . . . . . . . . 100
Fig. 3.69 Frequency spectrum of five-level flying capacitor converter. . . . . . 100
Fig. 3.70 Line voltage of 11-level flying capacitor converter. . . . . . . . . . . . . 100
Fig. 3.71 Frequency spectrum of 11-level flying capacitor converter. . . . . . . 101
Fig. 3.72 Line voltage of five-level modular multilevel cascaded converter. . . 101
Fig. 3.73 Frequency spectrum of five-level modular multilevel
cascaded converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Fig. 3.74 Line voltage of 11-level modular multilevel cascaded converter. . . 102
Fig. 3.75 Frequency spectrum of 11-level modular multilevel
cascaded converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Fig. 3.76 Number of auxiliary devices in NPC and FC converters [32]. . . . . 104
Fig. 3.77 Overall comparisons of multilevel converters [32] . . . . . . . . . . . . . 104
Fig. 4.1 Power circuit of a 3-phase 13-level MMC converter. . . . . . . . . . . . 110
Fig. 4.2 Common magnetic-link-based medium-voltage
converter for wind power generation systems [8] . . . . . . . . . . . . . . 111
Fig. 4.3 Cascaded high-frequency magnetic-link-based
medium-voltage converter for wind power generation systems. . . . 112
xviii Figures
Fig. 4.4 Modular medium-voltage converter for wind

power generation systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Fig. 4.5 Flowchart of high-frequency magnetic-link design process [10]. . . 115
Fig. 4.6 Metglas amorphous alloy 2605S3A stripe of 20 m
thickness and 25 mm width. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Fig. 4.7 Compound VITROPERM stripe [16]. . . . . . . . . . . . . . . . . . . . . . . . 117
Fig. 4.8 Specific core loss in terms of flux density and frequency
of Metglas alloy 2605SA1 under square-wave excitation. . . . . . . . 118
Fig. 4.9 Specific core loss in terms of flux density and frequency
of Metglas alloy 2605S3A under square-wave excitation. . . . . . . . 119
Fig. 4.10 Voltage and flux in a square-wave transformer . . . . . . . . . . . . . . . . 119
Fig. 4.11 Skin depth versus excitation frequency of a copper conductor. . . . 121
Fig. 4.12 Power losses due to skin effect in a round conductor
for different penetration ratios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Fig. 4.13 Power losses due to proximity effects in a round conductor
for different penetration ratios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Fig. 4.14 Graphical representation of the AC/DC resistance ratio Kr,
at 10 kHz, showing the effect of conductor diameter
and number of layers on AC resistance. . . . . . . . . . . . . . . . . . . . . . 124
Fig. 4.15 A two-winding transformer with stray capacitances. . . . . . . . . . . . 126
Fig. 4.16 Equivalent circuit of a two-winding transformer with stray
capacitances; all parameters referred to the primary side [22] . . . . 126
Fig. 4.17 Equivalent circuit of a two-winding transformer with leakage
inductances; all parameters referred to the primary side [22]. . . . . 128
Fig. 4.18 Maxwell desktop with model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Fig. 4.19 Maxwell desktop windows to select material or add a new
material. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Fig. 4.20 Maxwell desktop windows to create excitation coil . . . . . . . . . . . . 131
Fig. 4.21 Maxwell desktop windows to select element size and numbers . . . 132
Fig. 4.22 Maxwell desktop with solve setup window. . . . . . . . . . . . . . . . . . . 132
Fig. 4.23 Magnetic field illustration for Metglas alloy 2605S3A-based
high-frequency magnetic link in the Ansoft Maxwell 3D [26] . . . . 133
Fig. 4.24 Magnetic field illustration for Metglas alloy 2605SA1-based
high-frequency magnetic link in the Ansoft Maxwell 3D [26] . . . . 133
Fig. 4.25 Transformer core dimensions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Fig. 4.26 Transformer core development setup (raping of Metglas sheet). . . 134
Fig. 4.27 A photograph of the transformer core with frame; just after
wrapping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Fig. 4.28 Photograph of the transformer core with Metglas alloy
2605SA1 ribbon of 30 m thickness and 25 mm of width. . . . . . . 135
Fig. 4.29 Photograph of high-frequency magnetic link
with Metglas alloy 2605SA1 core. . . . . . . . . . . . . . . . . . . . . . . . . . 136
Fig. 4.30 Photograph of high-frequency magnetic link
with Metglas alloy 2605S3A core. . . . . . . . . . . . . . . . . . . . . . . . . . 136
Fig. 4.31 Wheatstone bridge circuit diagram . . . . . . . . . . . . . . . . . . . . . . . . . 138
Figures xix
Fig. 4.32 DC resistance measurement technique. . . . . . . . . . . . . . . . . . . . . . . 138

Fig. 4.33 Photograph of DC resistance measurement experimental setup . . . 139
Fig. 4.34 Photograph of experimental setup to calculate
voltage transformation ratios at 50-Hz sinusoidal excitation. . . . . . 141
Fig. 4.35 A photograph of the experimental setup for
characterization of high-frequency magnetic links . . . . . . . . . . . . . 142
Fig. 4.36 Measured primary and secondary side voltage waveforms. . . . . . . 143
Fig. 4.37 Measured voltage and current waveforms. . . . . . . . . . . . . . . . . . . . 143
Fig. 4.38 Measured total losses of Metglas alloy 2605SA1 against
excitation frequencies at excitation current of 0.35 A; only
one coil is energized at a time while the others are open circuited . . . 144
Fig. 4.39 Measured total losses of Metglas alloy 2605SA1 against
excitation currents of 10 kHz; only one coil is energized
at a time while the others are open circuited. . . . . . . . . . . . . . . . . . 145
Fig. 4.40 Measured core losses of Metglas alloy 2605SA1 at different
frequency ranging from 6 to 12 kHz; experiments were
carried out at 40 C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Fig. 4.41 Measured BH loops of Metglas alloy 2605SA1 at 40 C;
maximum magnetic flux density at 1, 2, and 3 A is 0.65, 0.74,
and 0.80 T, respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Fig. 4.42 Measured BH loops of Metglas alloy 2605SA1 at
different temperature ranging from 40 to 100 C. . . . . . . . . . . . . . . 147
Fig. 4.43 Measured maximum magnetic flux density of Metglas alloy
2605SA1 at different temperature ranging from 40 to 120 C:
maximum magnetic flux density at 40, 60, 80, 100, and
100 C is 0.65, 0.65, 0.64, 0.63, and 0.62 T, respectively. . . . . . . . 147
Fig. 4.44 Thermal image of high-frequency magnetic link when the
probe was pointed at the surface of the core. . . . . . . . . . . . . . . . . . 147
Fig. 4.45 Thermal image of high-frequency magnetic link when
the probe was pointed at the center of the core. . . . . . . . . . . . . . . . 148
Fig. 4.46 Measured BH loops of Metglas alloy 2605S3A at
excitation current of 3 and 5 A. The link is excited
by 12-kHz square-wave primary voltages. . . . . . . . . . . . . . . . . . . . 148
Fig. 4.47 Measured BH loops of Metglas alloy 2605S3A at various
temperatures. The link is excited by 12-kHz square-wave
primary voltages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Fig. 4.48 Measured BH loops of Metglas alloy 2605S3A at different
frequencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Fig. 4.49 Measured BH loops of Metglas alloys 2605SA1 and 2605S3A;
the links are excited by 12-kHz square-wave primary voltages . . . . . 149
Fig. 4.50 DC-link voltage of an H-bridge inverter cell. . . . . . . . . . . . . . . . . . 150
Fig. 4.51 Measured phase voltage of the prototype 5-level MMC
converter before the LC line filter circuit. . . . . . . . . . . . . . . . . . . . . 150
Fig. 5.1 Photograph of an ASIC by AMI. . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Fig. 5.2 Photograph of a DSP by TI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
xx Figures
Fig. 5.3 Photograph of an FPGA by Xilinx. . . . . . . . . . . . . . . . . . . . . . . . . . 156

Fig. 5.4 Multiplications per second of texas instruments
DSP and Altera FPGA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Fig. 5.5 Power circuit of a 3-phase 5-level converter . . . . . . . . . . . . . . . . . . 157
Fig. 5.6 MATLAB/Simulink and HDL coder-based FPGA programming. . . 159
Fig. 5.7 MATLAB/Simulink- and system generator-based FPGA
programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Fig. 5.8 Interfacing with target system in simulation environment. . . . . . . . 160
Fig. 5.9 Switching scheme for a 5-level MMC converter. . . . . . . . . . . . . . . 160
Fig. 5.10 Switching controller architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Fig. 5.11 Simulink and Xilinx ISE 13.2 Design Suite software-based
design technique. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Fig. 5.12 MATLAB/Simulink model for two carrier
signals, where = t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Fig. 5.13 Two carrier signals; one for the top H-bridge and 90 shifted
carrier for the bottom H-bridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Fig. 5.14 Updown counter-based carrier signal generation, where H
is the highest count, and t the step size. . . . . . . . . . . . . . . . . . . . . 163
Fig. 5.15 Schematic symbols of updown counter: a without phase
shift and b with 90 phase shift. . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Fig. 5.16 Schematic symbols of inverter, 180 phase shifter . . . . . . . . . . . . . 164
Fig. 5.17 Reference signals generation model. . . . . . . . . . . . . . . . . . . . . . . . . 165
Fig. 5.18 Three-phase 50-Hz reference signals (in MATLAB/Simulink
environment) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Fig. 5.19 Three-phase reference signals (in ISim Simulator environment). . . 166
Fig. 5.20 Schematic symbol of reference signal block. . . . . . . . . . . . . . . . . . 167
Fig. 5.21 Simulink block diagram of comparator module. . . . . . . . . . . . . . . . 168
Fig. 5.22 Carrier and reference signals in Simulink environment:
thick waves represent sine reference signals and dotted
triangular lines |represent the inverted form of carrier
signals as presented by triangular
solid lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Fig. 5.23 Gate pulses for the top H-bridge cell in phase A: a left top,
b left bottom, c right top and d right bottom. . . . . . . . . . . . . . . . . . 169
Fig. 5.24 Gate pulses for the bottom H-bridge cell in phase A: a left top,
Fig. 5.25 Gate pulses for the top H-bridge cell in phase B: a left top,
Fig. 5.26 Gate pulses for the bottom H-bridge cell in phase B: a left top,
Fig. 5.27 Gate pulses for the top H-bridge cell in phase C: a left top,
Fig. 5.28 Gate pulses for the bottom H-bridge cell in phase C: a left top,
Fig. 5.29 Switching technique without deadband. . . . . . . . . . . . . . . . . . . . . . 171
Fig. 5.30 Switching technique with deadband. . . . . . . . . . . . . . . . . . . . . . . . . 171
Figures xxi
Fig. 5.31 Schematic symbol of the comparator and PWM pulse generator
unit, which also has an output terminal called DBA_L to access
the deadband wave against PWM waves at pwm_L_T
and pwm_L_B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Fig. 5.32 Basic block diagram of FPGA-based 5-level converter
control algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Fig. 5.33 Basic block diagram of FPGA-based 3-phase 5-level
converter control circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Fig. 5.34 Gate pulses and deadband signals for switches in phase A,
where the narrow pulses represent the deadband (adjustable
in software environment). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Fig. 5.35 Gate pulses and deadband signals for switches in phase B,
Fig. 5.36 Gate pulses and deadband signals for switches in phase C,
Fig. 5.37 Assigning I/O pins via I/O port properties. . . . . . . . . . . . . . . . . . . . 177
Fig. 5.38 Assigning I/O pins by dragging into package view. . . . . . . . . . . . . 177
Fig. 5.39 I/O ports viewer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Fig. 5.40 Design summary/report viewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Fig. 5.41 Design summary/pinout report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Fig. 5.42 RTL schematic viewer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Fig. 5.43 View technology schematic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Fig. 5.44 Photograph of FPGA board with PC connection through
USB cable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Fig. 5.45 iMPACT process view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Fig. 5.46 Photograph of the FPGA-based switching controller section . . . . . 183
Fig. 5.47 Measured gate pulses for the top H-bridge cell in phase A:
1 left top, 2 left bottom, 3 right top, and 4 right bottom . . . . . . . . . 183
Fig. 5.48 Measured gate pulses for the bottom H-bridge cell in phase A:
Fig. 5.49 Measured gate pulses for the top H-bridge cell in phase B: 1
left top, 2 left bottom, 3 right top, and 4 right bottom. . . . . . . . . . . 184
Fig. 5.50 Measured gate pulses for the bottom H-bridge cell in phase B:
1 left top, 2 left bottom, 3 right top, and 4 right bottom. . . . . . . . . 185
Fig. 5.51 Measured gate pulses for the top H-bridge cell in phase C:
Fig. 5.52 Measured gate pulses for the bottom H-bridge cell in phase C:
Fig. 5.53 Measured phase voltage of a prototype 5-level converter. . . . . . . . 186
Fig. 5.54 Simulated phase voltage of a prototype 5-level converter. . . . . . . . 186
Fig. 6.1 Multi-coil generator-based wind turbine power generation
systems [3, 7]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Fig. 6.2 Multiple generator-based wind turbine system: single-phase
layout [6, 7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
xxii Figures
Fig. 6.3 Wind turbine generator system for direct medium-voltage

grid integration [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Fig. 6.4 Detailed power conversion circuit of high-frequency
magnetic-link MMC converter for step-up-transformer-less
direct grid connection of wind turbine generators. . . . . . . . . . . . . . 193
Fig. 6.5 Power conversion circuit of high-frequency magnetic-link-based
3-phase 5-level MMC converter for grid connection
of wind turbine generators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Fig. 6.6 A Photograph of Semikron IGBT modules SK30GH123. . . . . . . . 195
Fig. 6.7 A photograph of 3-phase 5-level 1-kV MMC converter . . . . . . . . . 196
Fig. 6.8 A photograph of Spartan-3E starter kit . . . . . . . . . . . . . . . . . . . . . . 197
Fig. 6.9 A photograph of ribbon cable and header socket. . . . . . . . . . . . . . . 197
Fig. 6.10 Detailed circuit diagram of driver circuit to step-up
the gate signal to 15 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Fig. 6.11 A photograph of extra driver circuit to step-up
the gate signal to 15 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Fig. 6.12 Layout of the distribution board. . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Fig. 6.13 A photograph of the switching and control
signal distribution board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Fig. 6.14 High-frequency rectifier circuits with fast recovery diodes. . . . . . . 201
Fig. 6.15 A photograph of the fast recovery rectifier section . . . . . . . . . . . . . 202
Fig. 6.16 Capacitor voltage of the fast recovery rectifier [1]. . . . . . . . . . . . . . 202
Fig. 6.17 A photograph of high-frequency inverter; generating
high-frequency square wave primary excitation voltage
of the high-frequency magnetic link [23]. . . . . . . . . . . . . . . . . . . . . 204
Fig. 6.18 A photograph of Texas Instruments high-speed
PWM controller UC3825BN. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Fig. 6.19 Simplified circuit of the switching signal generator
using Texas Instruments high-speed PWM
controller UC3825BN [23] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Fig. 6.20 A photograph of control circuit using Texas Instruments
high-speed PWM controller UC3825BN. . . . . . . . . . . . . . . . . . . . . 205
Fig. 6.21 Basic internal circuit of the Texas Instruments high-speed
PWM controller UC3825BN (oscillator section only). . . . . . . . . . . 206
Fig. 6.22 Alternative two output topology . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Fig. 6.23 Measured switching signals; on-time calculation . . . . . . . . . . . . . . 207
Fig. 6.24 Measured switching signals; dead-time calculation. . . . . . . . . . . . . 208
Fig. 6.25 Photograph of the Metglas amorphous alloy 2605SA1-based
high-frequency magnetic link [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Fig. 6.26 A photograph of Litz wire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Fig. 6.27 A photograph of the experimental setup of high-frequency
magnetic-link-based MMC converter (1 kV, 1.73 kVA) . . . . . . . . . 210
Fig. 6.28 Measured gate pulses of the high-frequency inverter. . . . . . . . . . . . 210
Fig. 6.29 Measured primary and secondary voltage waveforms
of high-frequency magnetic link . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Figures xxiii
Fig. 6.30 A photograph of Tektronix digital phosphor

oscilloscope DPO 2024 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Fig. 6.31 A photograph of Tektronix current probes. . . . . . . . . . . . . . . . . . . . 212
Fig. 6.32 A photograph of Tektronix high-voltage differential probe P5200. . . 213
Fig. 6.33 A photograph of Tektronix power analyzer PA4000. . . . . . . . . . . . 213
Fig. 6.34 A photograph of Fluke infrared temperature probe 80T-IR. . . . . . . 214
Fig. 6.35 A photograph of Flir thermal imaging camera i7. . . . . . . . . . . . . . . 215
Fig. 6.36 Measured gate pulses of four IGBTs in the top
module in phase A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Fig. 6.37 Measured phase voltage of the prototype system
(phase voltage before LC filter circuit) . . . . . . . . . . . . . . . . . . . . . . 216
Fig. 6.38 Zoomed phase voltage of the prototype system (phase
voltage before LC filter circuit; with 1.525-kHz PWM frequency. . . 216
Fig. 6.39 A photograph of MTE reactor to design LC filter. . . . . . . . . . . . . . 217
Fig. 6.40 A photograph of RS AC capacitor. . . . . . . . . . . . . . . . . . . . . . . . . . 217
Fig. 6.41 A photograph of LC line filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Fig. 6.42 Measured line voltage with SPWM of the prototype
system before filter 218
Fig. 6.43 Measured line voltage with SPWM of the prototype system
after the filter when the filter capacitors are connected in Y. . . . . . 219
Fig. 6.44 Measured line voltage with SPWM of the prototype system
after the filter when the filter capacitors are connected in . . . . . . 219
Fig. 6.45 Measured line voltage (before filter circuit) of the prototype
1-kV modular multilevel cascaded converter with THPWM. . . . . . 220
Fig. 6.46 Zoomed measured line voltage (before filter circuit) of the
prototype 1-kV modular multilevel cascaded converter
with THPWM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Fig. 6.47 Frequency spectrums of line voltages (before the filter circuit). . . . 221
Fig. 6.48 Measured line voltage (after filter circuit) of the prototype
1-kV modular multilevel cascaded converter with THPWM. . . . . . 221
Fig. 6.49 Zoomed measured line voltage (after filter circuit) of the
prototype 1-kV modular multilevel cascaded converter
with THPWM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Fig. 6.50 Frequency spectrum of line voltages (after the filter circuit). . . . . . 222
Fig. 7.1 A photograph of Semikrons SEMIKUBE converter
module IGDD6-4-426-D3816-E1F12-BL-FA. . . . . . . . . . . . . . . . . 228
Fig. 7.2 Electrical circuit of Semikrons SEMIKUBE converter
module IGDD6-4-426-D3816-E1F12-BL-FA. . . . . . . . . . . . . . . . . 229
Fig. 7.3 Power conditioner circuit for PMSG-based wind turbine
generator systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Fig. 7.4 A photograph of MITSUBISHI ELECTRIC IGBT
module CM100MXA-24S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Fig. 7.5 A photograph of Semikron SEMISTAK RE converter
module SKS B2 140 GDD 69/12 U-A11 MA PB. . . . . . . . . . . . . . 230
Fig. 7.6 Electrical power circuit of Semikron SEMISTAK RE
converter module SKS B2 140 GDD 69/12 U-A11 MA PB. . . . . . 231
xxiv Figures
Fig. 7.7 A photograph of Semikrons SEMIKUBE converter

modules designed for 110-, 220-, 400-, and 900-kW
(from the left) solar PV power converters . . . . . . . . . . . . . . . . . . . . 232
Fig. 7.8 Electrical equivalent circuit of Semikrons
SEMIKUBE converter modules. . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Fig. 7.9 ABB central solar PV inverter PVS800. . . . . . . . . . . . . . . . . . . . . . 232
Fig. 7.10 Housing of ABB central solar PV inverter PVS800
with step-up transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Fig. 7.11 Siemens PV inverter SINVERT PVS; the left two cabinets
are the DC cabinets and the other two the AC cabinets. . . . . . . . . . 233
Fig. 7.12 Function units of Siemens PV inverter SINVERT PVS. . . . . . . . . . 234
Fig. 7.13 ABB cast coil dry-type transformer. . . . . . . . . . . . . . . . . . . . . . . . . 235
Fig. 7.14 Housing of ABB cast coil dry-type transformer . . . . . . . . . . . . . . . 236
Fig. 7.15 ABB liquid-filled transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Fig. 7.16 Housing of liquid-filled transformer in onshore wind farm. . . . . . . 238
Fig. 7.17 Pauwels SLIM transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Fig. 7.18 A Pauwels transformer being lifted through the door
of a wind turbine tower during installation . . . . . . . . . . . . . . . . . . . 239
Fig. 7.19 Pauwels Bio-SLIM transformer for offshore applications. . . . . . . . 239
Fig. 7.20 Number of IGBTs/control complexity versus
converter level numbers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Fig. 7.21 THD (%) and semiconductor cost versus converter
level numbers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Fig. 7.22 Three-phase 9-level MMC converter. . . . . . . . . . . . . . . . . . . . . . . . 243
Fig. 7.23 Line voltage of 9-level MMC converter. . . . . . . . . . . . . . . . . . . . . . 244
Fig. 7.24 Line voltage frequency spectrum of 9-level MMC converter . . . . . 244
Fig. 7.25 Three line voltages of 9-level MMC converter . . . . . . . . . . . . . . . . 244
Fig. 7.26 Photograph of Mitsubishi Electric 4.5-kV 400-A IGBT
CM400HB-90H. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Fig. 7.27 Collectoremitter saturation voltages of 600-A-rated
IGBTs [10]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Fig. 7.28 Line voltages of 11-level MMC converter. . . . . . . . . . . . . . . . . . . . 246
Fig. 7.29 Line voltage frequency spectrum of 11-level MMC converter . . . . 247
Fig. 7.31 Infineon 3.3-kV IGBT module FZ800R33KL2C. . . . . . . . . . . . . . . 248
Fig. 7.32 Line voltage of 15-level MMC converter. . . . . . . . . . . . . . . . . . . . . 248
Fig. 7.35 ABB 2.5-kV IGBT module 5SNA 1200E250100. . . . . . . . . . . . . . 250
Fig. 7.36 Three-phase 19-level MMC converter [1]. . . . . . . . . . . . . . . . . . . . 251
Fig. 7.37 Voltage waveform across an IGBT in 19-level MMC converter. . . 251
Fig. 7.38 Voltage waveform of an H-bridge inverter cell in 19-level
MMC converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
Fig. 7.39 Phase voltage of 19-level MMC converter. . . . . . . . . . . . . . . . . . . . 252
Figures xxv
Fig. 7.42 Three line voltages of three-phase 19-level MMC converter. . . . . . 253
Fig. 7.43 Current waveform in an IGBT of 19-level MMC converter . . . . . . 253
Fig. 7.44 Semikron 1.7 IGBT module SKM 600GA176D. . . . . . . . . . . . . . . 254
Fig. 7.47 Three line voltages of 21-level MMC converter . . . . . . . . . . . . . . . 256
Fig. 7.48 Graphical representation converter performance
parameters [9, 10]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Fig. 7.49 Total performance indexes of different 11-kV MMC converters . . . 257
Fig. 7.50 Phase voltage of 33-kV 15-level MMC converter. . . . . . . . . . . . . . 259
Fig. 7.51 Three line voltages of three-phase 33-kV 15-level MMC
converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
Fig. 7.52 Line voltage frequency spectrum of 33-kV 15-level
MMC converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
Fig. 7.53 Photograph of ABB 6.5-kV IGBT module 5SNA 0400J650100. . . 261
Fig. 7.54 Circuit diagram of a 33-kV 23-level MMC converter. . . . . . . . . . . 262
Fig. 7.55 Phase-shifted carrier-based switching scheme for 23-level
MMC converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Fig. 7.62 Circuit diagram of a 43-level MMC converter. . . . . . . . . . . . . . . . . 266
Fig. 7.63 Switching control scheme of 43-level MMC converter. . . . . . . . . . 267
Fig. 7.64 Voltage across an IGBT in 43-level 33-kV MMC converter. . . . . . 268
Fig. 7.65 Output voltage of an H-bridge inverter cell in 43-level
33-kV MMC converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Fig. 7.69 IGBT current of 43-level MMC converter. . . . . . . . . . . . . . . . . . . . 269
Fig. 7.73 Graphical representation converter parameters [10, 21] . . . . . . . . . 271
Fig. 7.74 Total indexes of different 33-kV converters. . . . . . . . . . . . . . . . . . . 273
Tables
Table 1.1 Land area covered by offshore wind farms. . . . . . . . . . . . . . . . . . 2

Table 1.2 Global offshore wind farms installed capacity. . . . . . . . . . . . . . . 4
Table 1.3 Global wind farms (onshore and offshore) installed capacity . . . 4
Table 1.4 Global solar PV installations [26]. . . . . . . . . . . . . . . . . . . . . . . . . 7
Table 2.1 Electrical characteristics of solar module SX-10. . . . . . . . . . . . . 22
Table 2.2 Capacity of PVS version inverters . . . . . . . . . . . . . . . . . . . . . . . . 41
Table 3.1 Switching scheme of half-bridge converter . . . . . . . . . . . . . . . . . 54
Table 3.2 Maximum rms output voltage of half-bridge converters
with modulation index at 0.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Table 3.3 Switching scheme of full-bridge converter. . . . . . . . . . . . . . . . . . 56
Table 3.4 Maximum rms output voltage of full-bridge converters
with modulation index at 0.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Table 3.5 Eight switching states of three-phase two-level converter. . . . . . 62
Table 3.6 Comparison of switching patterns . . . . . . . . . . . . . . . . . . . . . . . . 73
Table 3.7 Market prices of some available power IGBTs [8]. . . . . . . . . . . . 76
Table 3.8 Voltage handling capacity of available power IGBTs. . . . . . . . . . 77
Table 3.9 Voltage handling capacity of power converters . . . . . . . . . . . . . . 77
Table 3.10 Converter specification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Table 3.11 Components of different multilevel converters [16]. . . . . . . . . . . 97
Table 3.12 Cost and performance of different multilevel converters [16] . . . 103
Table 3.13 Comparison of different multilevel converter topologies [16]. . . 104
Table 4.1 Properties of Metglas alloys [15]. . . . . . . . . . . . . . . . . . . . . . . . . 116
Table 4.2 Properties of nanocrystalline magnetic materials. . . . . . . . . . . . . 117
Table 4.3 Coefficients of Steinmetz equation with sinusoidal
wave excitation of Metglas alloy 2605SA1 . . . . . . . . . . . . . . . . . 118
Table 4.4 Coefficients of Steinmetz equation with square-wave
high-frequency excitation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Table 4.5 Optimal parameters for 2.5-kVA magnetic links [26]. . . . . . . . . . 129
Table 4.6 DC resistance measurement using DMM. . . . . . . . . . . . . . . . . . . 138
Table 4.7 DC resistance measurement using Wheatstone bridge. . . . . . . . . 139
Table 4.8 Transformation ratio calculations at 50 Hz (AC). . . . . . . . . . . . 140
xxvii
xxviii Tables
Table 4.9 Transformation ratio calculations at 50 Hz (DF) . . . . . . . . . . . . . . 140

Table 4.10 Transformation ratio calculations at 1 kHz (AC) . . . . . . . . . . . . . . 141
Table 4.11 Transformation ratio calculations at 1 kHz (DF) . . . . . . . . . . . . . . 141
Table 4.12 Transformation ratio calculations at 10 kHz (AC) . . . . . . . . . . . . . 142
Table 4.13 Transformation ratio calculations at 10 kHz (DF) . . . . . . . . . . . . . 142
Table 4.14 Total loss calculations with respect to each secondary
coil with 0.35-A excitation current. . . . . . . . . . . . . . . . . . . . . . . . . . 144
Table 4.15 Total loss calculations with respect to each secondary
coil at 10-kHz Metglas alloy 2605SA1-based magnetic link) . . . . . 145
Table 5.1 A comparison between the DSP and the FPGA technology. . . . . . . 157
Table 5.2 FPGA developmental boards and kits. . . . . . . . . . . . . . . . . . . . . . . . 158
Table 5.3 I/O pin configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Table 6.1 Converter specifications in the experiment. . . . . . . . . . . . . . . . . . . . 195
Table 6.2 Specification of module rectifier. . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Table 6.3 Voltage transformation ratios against primary winding [27]. . . . . . . 211
Table 7.1 Transformer data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Table 7.2 DVUF with different level number; 11-kV system. . . . . . . . . . . . . . 242
Table 7.3 Commercially available 4.5-kV-rated IGBTs. . . . . . . . . . . . . . . . . . 245
Table 7.6 Commercially available 1.7-kV-rated IGBT modules. . . . . . . . . . . . 254
Table 7.7 Converter comparison for an 11-kV system [9, 10] . . . . . . . . . . . . . 256
Table 7.8 Performance indexes of 11-kV converters [9, 10]. . . . . . . . . . . . . . . 256
Table 7.9 DVUFs of 33-kV MMC converters of different levels. . . . . . . . . . . 258
Table 7.11 Converter comparison for a 33-kV system [10, 21] . . . . . . . . . . . . . 272
Table 7.12 Total indexes of different 33-kV MMC converters [10, 21]. . . . . . . 272
Symbols
A Cross-sectional area (m2)

V Air velocity (m/s)
Mass density (kg/m3)
P Power (W)
r Rotor speed
1 Synchronous speed
P p Number of poles of power winding
fp Frequency of power winding current (Hz)
Pc Number of poles of control winding
fc Frequency of the control winding current (Hz)
nr Rotational speed (RPM)
G Generator
C Capacitor
D Diode
S Switching device
L Inductor
Rscell Intrinsic series resistance of a solar cell ()
Rshcell Equivalent shunt resistance of a solar cell ()
IL Light-generated current (A)
Icell Cell output current (A)
IPV PV array output current (A)
ns Number of cells connected in series
n p Number of cells connected in parallel
VPV PV array output voltage (V)
Rs PV array series resistance ()
Rsh PV array shunt resistance ()
Io Cell saturation current (A)
q Electronic charge (1.6021019 coulomb for an
electron)
A i Ideality factor (1.92)
K Boltz-mans constant (1.381023 J/K)
xxix
xxx Symbols
T Temperature (C)
P Change of PV array output power (W)
V Change of PV array output voltage (V)
Vrms PV inverter output AC voltage (V)
R Reservation factor
VA Minimum PV array voltage (V)
Vdc DC-link voltage of converter (V)
Van, Vbn, Vcn Phase voltages of converter
n Neutral point
a, b, c Terminals of three-phase system
Vcom@100FIT Device commutation voltage for a device reliability
of 100 FIT (V)
m Converter number of levels
m f Frequency modulation index
fm Frequency of reference signals (Hz)
fca Frequency of carrier signals (Hz)
Am Peak to peak amplitude of reference signal
Ac Peak to peak amplitude of carrier signal with
phase-shifting
map Amplitude modulation index with phase-shifting
B n Number of H-bridge inverter cell on a particular
phase leg
ps Carrier phase-shifting (Degree)
mal Amplitude modulation index with level-shifting
Acl Peak-to-peak amplitude of carrier signal with
level-shifting
Vll(rms) Grid line to line voltage (V)
Vdc(min) Minimum voltage of each DC-link capacitor (V)
Vdc(nom) Nominal voltage of each DC-link capacitor (V)
Ip(rms) Converter phase current (A)
Sc Apparent output power of the converter (VA)
nc Number of series-connected flying capacitor cells
Vc Capacitor voltage ripple (V)
Idc Converter DC output current (A)
Id Normalized index value
y Given value
ymin Minimum value
ymax Maximum value
Pcore Core loss (W)
f Frequency (Hz)
B Flux density (T)
k, y, z Coefficients of Steinmetz's equation
Vmax Maximum excitation voltage (V)
max Maximum flux (Wb)
Symbols xxxi
T Time period (s)

v(t) Time varying voltage (V)
Flux (Wb)
(t) Time varying flux (Wb)
N Number of turns
Bmax Maximum flux density (T)
Np Number of turns in the primary winding
Ns Number of turns in the secondary winding
Permeability (H/m)
Resistivity (m)
Rdc DC resistance ()
Pskin Skin effect loss (W/m)
d Diameter of the conductor (mm)
Skin depth (mm)
Pp_int Power loss due to internal proximity effect (W/m)
Conductivity (Siemens per meter)
U Number of layers
ns Total number of strands in a bundle
ls Total length of a single strand (m)
r b Radius of the bundle (mm)
He External magnetic field strength (A/mm)
Pp_ext Power loss due to external proximity effect (W/m)
Kr AC/DC resistance ratio
Rac AC resistance ()
ip, is Current of primary and secondary coils (A)
J Current density (A/mm2)
ap Cross-sectional area of a strand in primary winding
(mm2)
dp Diameter of a strand in primary winding (mm)
as Cross-sectional area of a strand in secondary
winding (mm2)
ds Diameter of a strand in secondary winding (mm)
Rp, Rs Resistances of the primary and secondary
windings ()
Cpo, Cso Self capacitances of primary and secondary wind-
ings (F)
Cpso Mutual capacitance between primary and secondary
windings (F)
Np, Ns Number of turns in primary and secondary windings
Rm, Lm Core loss resistance and magnetization inductance
Cp, Cs, Cps Equivalent capacitances referred to the primary side
(F)
Lex External inductance (H)
Ceqp, Ceqs Equivalent capacitances seen from primary and
secondary sides (F)
xxxii Symbols
Ldl Differential leakage inductance (H)

Ldlp, Ldls Differential leakage inductances of primary and
secondary windings
lp, ls Leakage fluxes of primary and secondary
windings (Wb)
Aw Area required by the primary and secondary
windings (mm2)
NA, NB, NC, ND, NE, NF Number of turns in secondary windings
vA, vB, vC, vD, vE, vF Secondary coils induced voltages (V)
TRA, TRB, TRC, TRD, TRE, TRF Voltage transformation ratios
i(t) Excitation current (A)
le Mean length of the core (m)
N1, N2 Number of turns of excitation and pick-up coils
Ae Cross-sectional area of the core (cm2)
VL Pick-up coil voltage (V)
k1 Clock divider value in carrier generator unit

pi
Angular resolution (Degree)
Ln Number of entries for a cycle
Aout Amplitude of the output signal
Phase angle (Degree)
k2 Clock divider value for reference generator unit
Doff Diode off time (s)
Vr(P-P) Peak-to-peak rms output voltage (V)
fs Frequency of the source voltage (Hz)
r Ripple factor
TJ, TC, TA Temperatures of junction, case, and ambient (C)
JA, JC, CS, SA Thermal resistances of total, transistor, insulator,
and heatsink (C/W)
PD Power dissipation (W)
RT, CT Timing resistor and capacitor (, F)
Dmax Desired maximum duty cycle
D% Percentage of duty cycle
Thigh, Thigh Switch on and off time (s)
fH Frequency of the medium frequency inverter (Hz)
Ploss_inv Power losses in the inverter section (W)
Pc_inv, Psw_inv Conduction and Switching losses (W)
Psw_inv_2, Psw_inv_M Switching loss of 2-level and multilevel
converters (W)
fc_2, fc_M Carrier frequencies of 2-level and multilevel con-
verters (Hz)
Pc_sw, Pc_D Conduction losses of IGBT and diode (W)
ma Amplitude modulation index
pf Power factor
Symbols xxxiii
Vt, Vf Voltage drop at zero current of IGBT and diode (V)

RCE, RAK Forward resistances of IGBT and Diode ()
Efficiency
Pout Output power (W)
Pin Input power (W)
% Percentage of efficiency
Pc_9 Conduction loss of 9-level converter (W)
VCE(sat) Collector-emitter saturation voltage (V)
Acronyms
A Ampere
ABB ASEA Brown Boveri
AC Alternating current
ALOs Arithmetic and logic operations
AMSC American Superconductor Corporation
ASIC Application-specific integrated circuit
AUD Australian dollar
BDFM Brushless doubly-fed machine
BSCCO Bismuth strontium calcium copper oxide
DC Direct current
DDS Direct digital synthesis
DFIG Doubly-fed induction generator
DMM Digital Multimeter
DSP Digital signal processor
DTC Direct torque control
DVUF Device voltage utilization factor
EMC Electromagnetic compatibility
EMI Electromagnetic interference
FC Flying capacitor
FPGA Field programmable gate array
g/cm3 Gram per cubic centimeter
GW Gigawatt
HDL Hardware description language
HFT High frequency transformer
HT Height
HTS High temperature superconductor
HVAC High voltage alternating current
HVDC High voltage direct current
Hz Hertz
I/O Input and (or) output
IBM International Business Machines
xxxv
xxxvi Acronyms
IC Integrated circuit
ID Inner diameter
IGBTs Insulated gate bipolar transistors
I-V Current versus voltage
k Kilo
kB Kilobyte
kg kilogram
kHz Kilohertz
km Kilometer
km2 Square kilometer
kV kilovolt
kVA Kilo volt ampere
kVAR Kilo volt ampere reactive
kW Kilo watt
kWh Kilowatt hour
LC Inductor-capacitor
LED Light emitting diode
LEs Logic elements
LUT Look-up table
LVRT Low voltage ride through
m2 Square meter
m3 Cubic meter
MHz Megahertz
MMC Modular multilevel cascaded
MOSFET Metal oxide semiconductor field effect transistor
MPP Maximum power point
MPPT Maximum power point tracker
MRI Magnetic resonance imaging
MVA Mega volt ampere
MW Megawatt
NCD Native circuit description
nF/kWp Nino farad per kilo watt peak
NGD Native generic database
NPC Neutral point clamped
NRE Nonrecurring engineering
OD Outer diameter
PM Permanent magnet
PMG Permanent magnet generator
PMSG Permanent magnet synchronous generator
PV Photovoltaic
P-V Power versus voltage
PWM Pulse width modulation
rms Root mean square
RTL Register transfer level
SCIG Squirrel-cage induction generator
Acronyms xxxvii
SDFM Single doubly-fed machine

SDPWM Sixty degree pulse width modulation
SliM Spoked lightweight machine
SPI Serial peripheral interface
SPWM Sine pulse width modulation
SVM Space vector modulation
THDs Total harmonic distortions
THPWM Third harmonic injected pulse width modulation
TRPWM Trapezoidal pulse width modulation
US United States
USD United States dollar
V Volt
VHDL Very high speed integrated hardware description language
VSC Voltage source converter
VSIs Voltage source inverters
W Watt
W/kg Watt per kilogram
WRSG Wound rotor synchronous generator
F/kWp Micro farad per kilowatt peak
Authors Biography
Md. Rabiul Islam (M14, IEEE) received the

B.Sc. and M.Sc. in Electrical and Electronic En-
gineering (EEE) from Rajshahi University of
Engineering and Technology (RUET), Rajsha-
hi, Bangladesh, in 2003 and 2009, respectively;
and the Ph.D. in Electrical and Electronics from
University of Technology Sydney (UTS),
Sydney, Australia, in 2014.
From 2005 to 2008, he lectured with the
Department of EEE, RUET, where he became
an Assistant Professor in June 2008. He is cur-
rently a Research Associate with the School of
Electrical, Mechanical and Mechatronic Sys-
tems, UTS. He has authored and coauthored more than 40 technical papers and 2
book chapters. His research interests are in the fields of power electronic converters,
renewable energy technologies, and smart grid.
Dr. Islam is a member of the Institution of Engineers, Bangladesh and the Aus-
tralian Institute of Energy, Australia. He received the University Gold Medal and
Joynal Memorial Award from RUET for his outstanding academic performance
while pursuing the B.Sc. engineering degree. He also received the Best Paper Award
at IEEE PECon-2012. He acts as a reviewer for several prestigious international
journals.
xxxix
xl Authors Biography
Youguang Guo (S02-M05-SM06, IEEE) re-

ceived the B.E. from Huazhong University of
Science and Technology (HUST), Wuhan, Chi-
na, in 1985, the M.E. from Zhejiang University,
Zhejiang, China, in 1988, and the Ph.D. from
University of Technology Sydney (UTS), Syd-
ney, Australia in 2004, all in Electrical Engi-
neering.
From 1988 to 1998, he lectured in the De-
partment of Electric Power Engineering, HUST.
From March 1998 to July 2008, he was a Vis-
iting Research Fellow, Ph.D. candidate, Post-
doctoral Fellow, and Research Fellow in the
Center for Electrical Machines and Power Electronics, Faculty of Engineering, UTS.
He is currently an Associate Professor at the School of Electrical, Mechanical and
Mechatronic Systems, UTS. His research fields include measurement and modeling
of magnetic properties of magnetic materials, numerical analysis of electromagnetic
field, electrical machine design and optimization, power electronic drives and control.
Jianguo Zhu (S93-M96-SM03, IEEE) re-

ceived the B.E. from the Jiangsu Institute of
Technology, Zhenjiang, China, in 1982, the
M.E. from Shanghai University of Technology,
Shanghai, China, in 1987, and the Ph.D. from
University of Technology Sydney (UTS), Syd-
ney, Australia, in 1995.
He is currently a Professor of Electrical
Engineering and the Head of the School of
Electrical, Mechanical and Mechatronic Systems,
UTS. His research interests include electromag-
netics, magnetic properties of materials, electrical
machines and drives, power electronics, renew-
able energy systems, and smart micro-grids.
Keywords
Wind farms PV power plants Power converter topologies Multilevel converters

Modulation techniquesMedium-voltage smart micro-gridsStep-up-transformers
Amorphous alloys NanocrystallineHigh-frequency magnetic-linksMultiple
isolated and balanced DC supplies Digital switching controller Field programmable
gate array (FPGA) Design and implementation
xli
Chapter 1
Introduction
AbstractThis chapter presents a succinct preface focusing on various aspects,

including the historical growth of two dominating renewable energy sources, i.e.,
wind and solar sources, the existing technologies for large-scale generation of
these renewable powers, technical challenges, and possible solutions. Reduction
of volume and weight of power conversion system would solve effectively some
technical problems of the existing technologies for large-scale renewable power
generation systems, which is the foremost intention of this book.
KeywordsEnergy and environment challengesRenewable energyOffshore

wind farms Photovoltaic (PV) power plants Remote area power supply Grid
connection of renewable power plants
1.1Most Dominating Renewable Power Sources
The energy and environment represent two major areas of current global crisis,
and it is more and more widely recognized that renewable energy sources, espe-
cially wind and solar energy, can offer effective solutions to these enormous chal-
lenges [1] as the wind and solar power development is experiencing dramatic
growth [26]. Over 318GW of wind power generation and over 130GW solar
photovoltaic (PV) power generation have been installed by 2013. Accordingly, the
cumulative growth of wind and solar PV installation has directly pushed the wind
turbine and solar PV converter technologies into a more competitive area [712].
In this propitious climate, it is therefore essential for scientists and engineering
researchers to find the most effective converter technologies for integration of the
wind and solar power generation systems into the main power grids.
Springer-Verlag Berlin Heidelberg 2014 1

M.R. Islam et al., Power Converters for Medium Voltage Networks,
Green Energy and Technology, DOI 10.1007/978-3-662-44529-7_1
2 1Introduction
1.1.1Historical Growth of Wind Power Generation Capacity
Wind speed varies continuously with time and height because of the changes in
the thermal conditions of air masses. The motion of air masses is not only a global
phenomenon but also a regional and local phenomenon. The annual peak hours are
normally around 2,5003,000h at good sites. A wind turbine generator converts
wind energy to electricity energy. If A is the cross-sectional area through which
the air of velocity V flows, and d is the air density, the theoretical power P avail-
able in a wind stream can be calculated from [13, 14]
1
P= d AV 3 . (1.1)
2
Usually, offshore winds tend to flow at higher speeds than onshore winds. This
allows the turbine to produce more electricity as the possible energy produced
from the wind is proportional to the cube of the wind speed. Also unlike onshore
wind, offshore breezes can be strong in the afternoon, matching the time when
load demands are at peak level. Moreover, wind farms cover large areas of land.
The land area covered by a 3.6-MW turbine can be almost 0.37km2, such that
54 turbines would cover about a land area of 20km2. Table1.1 summarizes the
land covered by some offshore wind farms [15]. For example, the Anholt offshore
wind power plant is a Danish offshore wind power plant with nameplate capacity
of 400MW officially inaugurated in September 2013, which covers a land area
of 88km2. It is the third largest offshore wind farm in the world and the largest in
Denmark. Siemens supplied, installed, and commissioned 111 wind turbines, each
with a capacity of 3.6MW and a rotor diameter of 120m. Figure1.1 shows a pho-
tograph of the Anholt offshore wind power plant. Since offshore wind farms can
save land rental expense which is equivalent to 1018% of the total operating and
maintenance costs of a wind farm, offshore-based wind farms have attracted great
attention in the last few years. Considering the wind speed and capacitance effect
of the transmission-line cables, offshore wind turbines are usually installed about
Table1.1Land area covered by offshore wind farms

Wind farms Power capacity Distance from Number of Area covered
(MW) shore (km) turbines (km2)
Anholt 400 21 111 88.00
Barrow 108 7 30 10.00
Gunfleet Sands 172 7 48 17.50
Horns Rev 160 15 80 20.00
Horns Rev 2 200 27 91 35.00
Lynn and Inner 194 9 54 20.00
Ormonde 150 10 30 8.70
Princess 120 23 60 14.00
Amalia
1.1 Most Dominating Renewable Power Sources 3
Fig.1.1A photograph of Anholt offshore wind power plant was officially inaugurated in
September 2013 [16]
727km from the shore. Figure1.2 shows an ideal model of smart power systems
with offshore wind farms.
According to the statistical data, the cumulative installed capacity of offshore
wind farms in 2008, 2010, and 2012 is 1.50, 3.08, and 5.41GW, respectively.
The global annually installed capacity of offshore wind farms is summarized in
Table 1.2. The capacity therefore almost doubled in every 2years [18], whereas
the cumulative installed capacity of total (onshore and offshore) wind power gen-
eration in 2008, 2010, and 2012 was substantially less than this at 120.26, 197.68,
and 282.43GW, respectively. The global annually installed capacity of farms is
summarized in Table1.3. It is expected that the global offshore installed capac-
ity will increase to approximately 20GW by 2015 and rise sharply to 104GW
by 2025. According to the Global Wind Energy Council and Green Peace
International estimations, it is possible to mitigate 20% of global electricity
demand with wind power [19].
1.1.2Technical Challenges and Possible Solutions
To integrate scattered wind turbine generators into a medium-voltage grid (e.g.,

1133kV), a power frequency (50 or 60Hz) transformer is commonly used to
4 1Introduction
Fig.1.2An ideal model of renewable power source-based smart power systems [17]
Table1.2Global offshore Years Annual addition (MW) Cumulative

wind farms installed capacity (MW)
2005 90 723
2007 318 1,134
2009 577 2,084
2011 1,037 4,120
2013 1,647 7,062
Table1.3Global wind Years Annual addition (MW) Cumulative

farms (onshore and offshore) (MW)
installed capacity
2005 11,471 59,091
2007 19,951 93,889
2009 38,351 158,975
2011 40,125 238,126
2013 34,923 318,117
Transformer
Generator Converter
380~690 V
Grid
Pitch
drive
Switching &
turbine control
Fig.1.3Wind turbine with step-up transformer
step up the voltage for long-distance transmission. Figure1.3 shows the traditional
wind turbine with step-up transformer. In an offshore wind turbine power genera-
tion system, this transformer is usually installed at a height of about 80m inside
the nacelle together with other equipment, such as the generator and power con-
verter. The tower provides support to the rotating parts and nacelle (the stationary
parts). The nacelle weight of a 5-MW turbine is about 300t, while the rotor repre-
sents only about 120t. Therefore, the tower diameter and strength depend mostly
on the weight of the nacelle and expected wind loads. The tower cost accounts for
26%, the largest component, of the total turbine component cost [20]. The tower
is normally held by a heavy foundation to ensure that it withstands the overturning
moment created from the turbine. The foundation size of a 2.3-MW wind turbine
is 314m2, and the approximate weight is 2,000t or more [21]. In an offshore area,
the cost of installation is extremely high. On average, approximately 20% of the
capital costs are associated with the installation [22, 23]. During the past decade,
Van Oord has been involved in many offshore wind farm projects in North West
Europe, providing construction services such as foundation and turbine installa-
tion, cable installation, and scour protection. Van Oord has an innovative and
advanced transport and installation vessel named Aeolus. The vessel is 139m
long and 38m wide with a draft of 5.7m. The vessel is equipped with a crane
capable of hoisting 900t at 30m radius and has accommodation for 74 persons.
Aeolus is used for the installation of foundations and turbines. Figure1.4 shows a
photograph of wind turbine installation process. Therefore, a reduction in mechan-
ical loading represents an enormous saving of tower construction and turbine
installation costs. Moreover, regular monitoring and maintenance are also critical
in offshore applications. For example, an offshore wind farm requires about 20%
higher operating and maintenance cost compared with an onshore farm. Figure1.5
shows photograph of offshore wind power plant under maintenance. As the power
rating and the distance of the wind park from the shore increase, the engineering
6 1Introduction
Fig.1.4A photograph of installation process of offshore wind power plant with the vessel
Aeolus [24]
Fig.1.5A photograph of offshore wind farm under maintenance [25]
challenges associated with the installation and maintenance of the power genera-
tion systems also grow. Recently, many researchers/scientists have begun to focus
on a number of areas including proposing new converters to eliminate the step-up
transformers, applying medium-frequency transformers to reduce the weight and
volume of step-up transformers, and superconducting wind generators to reduce
the weight and volume of wind turbine generators. These steps are aimed at reduc-
ing the weight and volume of the wind turbine generator system. The weight and
volume reduction of wind turbine power generation system still needs further
investigation because the wind power has become one of the main renewable
energy sources for future electricity supply within smart microgrids.
1.1.3Historical Growth of Solar PV Power Generation

Capacity
Solar energy is the second main renewable energy source for future electricity sup-
ply. Solar PV generates electricity in well over 100 countries and continues to be
the fastest growing renewable resource in the world. By the end of 2011, a total
of 69.68GW PV power capacity had been installed, sufficient to generate around
80billionkWh/year and enough to cover the annual power supply needs for
more than 20million households as reported by European Photovoltaic Industry
Association (EPIA). Solar PV installations substantially increased over the last
5years. The annual installation of new PV power capacity rose from 16.80GW
in 2010 to 29.66GW in 2011, and in 2009, it was only 7.43GW. The globally
installed PV power capacity is summarized in Table1.4 [26]. According to statisti-
cal data, the cumulative installed capacity of solar PV generation in 2007, 2009,
and 2011 was 9.44, 23.21, and 69.68GW, respectively. This means that the capac-
ity of solar PV generation effectively tripled every 2years.
Since 2007, medium- and large-scale PV power plants have attracted great
interest and PV power plants of more than 10MW in capacity have now become
a reality. More than 200 PV power plants have already been installed in the world,
each of them generating an output of more than 10MW. Of these plants, 34 are
located in Spain and 26 in Germany. The number of PV power plants will con-
tinue to rise. The literature indicates that more than 250 PV power plants will be
installed within the next few years [27]. Future PV power plants will have higher
power capacity. Indeed, some are to have a capacity in excess of 250MW. These
multi-megawatt PV power plants require large areas of land. Owing to this, they
Table1.4Global solar PV Years Annual addition (MW) Cumulative

installations [26] (MW)
2001 328 1,753
2003 578 2,798
2005 1,429 5,340
2007 2,528 9,443
2009 7,438 23,210
2011 29,665 69,684
2012 32,472 102,156
8 1Introduction
Fig.1.6A photograph of Beneixama 20MW PV power plant [27]
are usually installed in remote areas, far from cities. The 20MW PV power plant
in Beneixama, Spain, used about 200 SINVERT 100M inverters and installed
approximately 100,000 PV modules in a land area of 500,000m2 [27]. Therefore,
for power transmission, a medium-voltage grid is usually used. Figure1.6 shows
the photograph of Beneixama 20MW PV power plant.
1.1.4Technical Challenges of Solar PV Power Generation

Systems
Although different power electronic converters have been developed in the last few
decades using conventional topologies for solar PV systems, it is hard to connect
the traditional converters to the grids directly as the distortion in generated out-
put voltages is high and a single switch cannot stand at grid voltage level. With
the rapid growth of grid-connected PV generations, the total harmonic distortions
(THDs) generated from PV inverters is becoming a major concern [28]. In this
regard, conventional systems utilizing the power frequency step-up transformer,
filter, and booster not only increase the size, weight, and loss but also increase the
cost and complexity of the system installation and operation. Swiss solar company
Tritec built a PV power plant with 5.2MW of power for the commercial company
Migros. The installation feeds directly from the roofs of the Migros-Verteilbetriebs
Fig.1.7A photograph of step-up transformer installation of Tritec 5.2MW solar PV power

plant [30]
Neuendorf AG into the medium-voltage grid. Tritec installed SolarMax trans-

former compact stations from the Swiss manufacturer Sputnik Engineering on
three roof areas of the factory. To reduce transmission losses, Tritec feeds solar
power directly from the roof into a medium-voltage grid through step-up trans-
former. Figure1.7 shows photograph of step-up transformer installation of Tritec
5.2MW solar PV power plant.
These heavy and large-size power frequency step-up transformers significantly
increase the weight and volume of the renewable power generation systems.
Because of the heavy weight and large size of the power frequency transformer,
the wind turbine generator and PV inverter system can be expensive and com-
plex in terms of installation and maintenance especially in offshore and remote
area applications. For example, the weight and volume of a 0.69/33kV, 2.6 MVA
transformer are typically in the range of 68t and 59m3, respectively, and the
weight and volume of a 0.4/33kV, 1 MVA power frequency transformer are typi-
cally in the range of 35t and 45m3, respectively [29]. A liquid-filled 2-MVA
step-up transformer uses about 900kg of liquid as the coolant and insulator, which
requires regular monitoring and replacement. These levels are critical in offshore
and remote area applications where the costs of installation and regular mainte-
nance are extremely high.
10 1Introduction
1.1.5Possible Solution to Technical Challenges

for Large-Scale Renewable Generation Systems
To ensure that the goals of smart grids are met, the current industrial trend is to
move away from these heavy and large-size passive components to compact and
lightweight systems that use more and more semiconductor devices controlled by
advanced digital controls. In comparison with the conventional two-level convert-
ers, multilevel converters present lower switching losses, lower voltage stress on
switching devices, and better power quality. They also enable the development
of medium-voltage converters using matured and cheap power semiconductor
devices. The development of multilevel medium-voltage converters enables the
connection of renewable energy systems directly to the grid without using large,
heavy, and costly power transformers. Although several multilevel converter
topologies have been used in low-voltage applications, most of them are not suit-
able for medium-voltage applications. Because of some special features, such as
the number of components scaling linearly with the number of levels and high-
level number attainability by the use of identical modules, the modular multilevel-
cascaded (MMC) converter topology can be considered as a possible candidate for
medium-voltage applications [31] to connect renewable power generation units to
the medium-voltage grid directly. Since the component number and control com-
plexity increase linearly with the number of levels, the optimal selection of the
number of converter levels is important in order to achieve the best performance/
cost ratio for the medium-voltage converter systems. For example, the 19-level
and 43-level converters are found to be optimal for 11- and 33-kV systems,
respectively [32]. However, the MMC converter requires multiple isolated DC
sources that must be balanced, and therefore, its application is not straightforward,
especially in wind power generation systems [33, 34]. Moreover, the multilevel
converter requires a number of switching and control pulse width modulation
(PWM) signals, which cannot be generated by the available digital signal proces-
sor (DSP) because the available DSP can only at present provide about 6 pairs of
PWM channels. Furthermore, the DSP runs a sequential program in its micropro-
cessor, i.e., all the gate pulses cannot be updated with the same clock pulse [35].
Therefore, the DSP-based system requires an additional control strategy to com-
pensate the time delay [36].
1.2Major Objectives of the Book
The major objectives of this book are as follows:

1. to find a suitable power converter for the compact and lightweight direct grid
integration of renewable power generation systems with matured and low-price
semiconductor devices, a modern digital controller, and advanced magnetic
materials with high saturation flux density and low core losses;
1.2 Major Objectives of the Book 11
2. to investigate the different multilevel converter topologies, taking into account

the availability of power semiconductors, output power quality, control com-
plexity, and the cost of semiconductors so as to find a suitable multilevel con-
verter topology for medium-voltage applications;
3. to research the different number of MMC converter levels so as to introduce a
way to design a medium-voltage MMC converter with an optimal number of
converter levels;
4. to design and analyze a fully digital switching controller for a multilevel con-
verter within the most common software environment so as to reduce the devel-
opmental time and cost of the switching controller;
5. to design and analyze a high-frequency magnetic link with multiple secondary
windings so as to verify the new concept of generating multiple isolated and
balanced DC supplies for all of the H-bridge inverter cells of the MMC con-
verter; and
6. to design and construct a scaled down test prototype of a high-frequency
magnetic-link MMC converter to validate the new concept of step-up-trans-
former-less direct grid integration of renewable generation systems with
medium-voltage converters.
1.3Contribution of the Book
The major contributions of this book include the following:

1. an in-depth investigation into the commercially available medium- and large-
scale renewable power generation technologies and the research and develop-
ment of converter topologies leading to the conclusion that medium-voltage
converter may be the best possible option to reduce the weight and volume of
renewable power generation systems;
2. the comparison of different available multilevel converter topologies with
medium-voltage applications leading to the conclusion that the MMC converter
is the best possible candidate for medium-voltage applications;
3. the detailed study of a design process for the MMC converter which focuses on
the optimal selection of the number of converter levels;
(To the best knowledge of the authors, they are the first who have presented a
technique to select the optimal number of converter levels for medium-voltage
applications.)
4. the detailed study of a new design process for the switching controller to
reduce developmental time and cost;
(An alternative technique to design the switching controller to save develop-
ment time and cost are presented.)
5. the detailed study of high-frequency magnetic link with multiple secondary
windings to generate multiple isolated and balanced DC supplies for all the
H-bridge inverter cells of the MMC converter; and
12 1Introduction
(To the best knowledge of the authors, they are the first who have presented
high-frequency magnetic link to generate multiple isolated and balanced DC
supplies for the MMC converter.)
6. the development and testing within a laboratory of a scaled down high-frequency-
link MMC converter which converters 210V DC into 3-phase 1kVrms 50Hz AC.
(To the best knowledge of the authors, they are the first who have experimen-
tally verified a high-frequency magnetic-link MMC medium-voltage converter
for direct grid integration of renewable generation systems.)
1.4Organization of the Book
Chapter 2 presents a review of the traditional system, which requires a step-up trans-
former to interconnect the renewable generation systems to the grid, and the other
recently proposed converter topologies for transformer-less grid interconnection in
order to show a complete picture of the importance of size and weight reduction,
and performance improvement of power converter systems. In order to develop a
common converter for both wind and PV power generation systems, after a thorough
investigation of all possible converter topologies covering almost all types of existing
power converters, it is concluded that the multilevel converter topology with high-
frequency magnetic link would be the most feasible option for developing medium-
voltage converters for direct grid connection of wind and PV power systems.
Chapter 3 is dedicated to the method to find a suitable multilevel converter
topology, which can connect the renewable generation units directly to the
medium-voltage grid with mature semiconductor devices. Different multilevel con-
verter topologies, such as the neutral point clamped (NPC), flying capacitor (FC),
and MMC converters, have been considered for the design of an 11-kV converter
system. The comparison is made in terms of the number of semiconductors, sem-
iconductor cost and availability, THDs, filter size, and control complexity of the
converters. The performance is analyzed and compared through numerical analysis
in the MATLAB/Simulink environment. To generate the switching pulses, a level-
shifted carrier-based switching scheme is used for NPC topologies and a phase-
shifted carrier-based switching scheme is used for the FC and MMC converter
topologies with a carrier frequency of 12kHz and a modulation index of 0.80.98.
To couple the renewable energy source to the MMC converter, a common high-
frequency magnetic link with multiple secondary windings is considered and elec-
tromagnetic performance is reported in Chap. 4. The medium-frequency link is
used to generate the isolated balanced multiple DC supplies for the MMC con-
verter from a single low-voltage power source. Compared with the conventional
transformers operated at the power frequency (50 or 60Hz), the high-frequency
(in the range of a few kHz to MHz) transformers have much smaller and lighter
magnetic cores and windings and thus much lower costs.
The multilevel converter requires a number of switching and control PWM sig-
nals, which cannot be generated by the available DSP because the available DSP
1.4 Organization of the Book 13
can only at present provide about 6 pairs of PWM channels. In this instance, the
field programmable gate array (FPGA) is the natural choice for medium-voltage
multilevel converters. However, most of the available design techniques require
special software such as the HDL coder, system generator, PSIM and ModelSim,
which increases the development time and cost. In Chap. 5, a fully digital switch-
ing controller is developed for 3-phase 5-level converters. The SK 30GH 123
IGBTs are used to develop a prototype multilevel converter, and an XC3S500E
FPGA is used to develop the switching controller. The most common software
such as the MATLAB/Simulink and Xilinx ISE based design technique is used
which may reduce the developmental time and cost of the switching controller.
The simulation results serve as a preliminary validation of the proposed design
technique, which will be finally verified by the experimental results. The devel-
oped switching scheme can be used for any multilevel converter configuration
with minimum changes in software environment. Moreover, the proposed design
techniques and implementation issues may be useful for designing any other mod-
ern power converter.
To verify the feasibility of the high-frequency-link MMC medium-voltage con-
verter, in Chap. 6, a scaled down 1.73 kVA laboratory prototype test platform with
modular 5-level cascaded converter is developed, which converts 210V DC (recti-
fied generator voltage) into 3-phase 1kVrms 50Hz AC. The design and implemen-
tation of the prototyping, test platform, and the experimental results are analyzed
and discussed. It is expected that the proposed new technology would have great
potential for future renewable generation systems and smart grid applications.
The high number of levels means that medium-voltage attainability will be
possible to connect the renewable generation units to the medium-voltage grid
directly and it will also be possible to improve the output power quality. Since the
number of components and control complexity increase linearly with the num-
ber of levels, the optimal selection of the number of converter levels is important
for the best performance/cost ratio of the medium-voltage converter systems. In
Chap. 7, an 11-kV system and a 33-kV system are designed and analyzed tak-
ing into account the specified system performance, control complexity, and cost
and market availability of the power semiconductors. It is found that the 19-level
and 43-level converters are optimal for 11- and 33-kV systems, respectively. The
detailed designs and analysis of an 11-kV MMC converter and a 33-kV MMC
converter systems are presented.
Chapter 8 draws the conclusions for the whole book and proposes possible
future works for the further development of the technology.
References
1. Islam MR, Islam MR, Beg MRA (2008) Renewable energy resources and technologies prac-
tice in Bangladesh. Renew Sustain Energy Rev 12(2):299343
2. Li H, Xie M, Zhang T (2013) Promote the development of renewable energy: a review and
empirical study of wind power in china. Renew Sustain Energy Rev 22:101107
14 1Introduction
3. Sahu BK, Hiloidhari M, Baruah D (2013) Global trend in wind power with special focus on
the top five wind power producing countries. Renew Sustain Energy Rev 19:348359
4. Da Z, Xiliang Z, Jiankun H, Qimin C (2011) Offshore wind energy development in China:
current status and future perspective. Renew Sustain Energy Rev 15(9):46734684
5. Chen J (2011) Development of offshore wind power in China. Renew Sustain Energy Rev
15(9):50135020
6. Islam ABMS, Jameel M, Jumaat MZ, Shirazi SM, Salman FA (2012) Review of offshore
energy in Malaysia and floating Spar platform for sustainable exploration. Renew Sustain
Energy Rev 16(8):62686284
7. Zhixin W, Chuanwen J, Qian A, Chengmin W (2009) The key technology of offshore wind
farm and its new development in China. Renew Sustain Energy Rev 13(1):216222
8. Ion CP, Marinescu C (2013) Three-phase induction generators for single-phase power gen-
eration: an overview. Renew Sustain Energy Rev 22:7380
9. Ayhan D, Saglam S (2012) A technical review of building-mounted wind power systems and
a sample simulation model. Renew Sustain Energy Rev 16(1):10401049
10. Islam MR, Mekhilef S, Saidur R (2013) Progress and recent trends of wind energy technol-
ogy. Renew Sustain Energy Rev 21:456468
11. Iglesias RL, Arantegui RL, Alonso MA (2011) Power electronics evolution in wind tur-
binesa market-based analysis. Renew Sustain Energy Rev 15(9):49824993
12. Shafiullah GM, Oo AMT, Ali ABMS, Wolfs P (2013) Potential challenges of integrat-
ing large-scale wind energy into the power grida review. Renew Sustain Energy Rev
20:306321
13. Du Z, Gu W (2009) Aerodynamics analysis of wind power. In: Proceedings of the 2009
world non-grid-connected wind power and energy conference; 2426 Sept 2009, Nanjing,
China, pp 13
14. Islam MR, Guo YG, Zhu JG (2010) Steady state characteristic simulation of DFIG for wind
power system. In: Proceedings of the 6th international conference on electrical and computer
engineering, 1820 Dec 2010, Dhaka, Bangladesh, pp 151154
15. Islam MR, Guo YG, Zhu JG (2014) A review of offshore wind turbine nacelle: technical
challenges, and research and developmental trends. Renew Sustain Energy Rev 33:161176
16. Anholt-offshore-wind-farm (2014) [online]. Available at: http://www.offshorewind.biz/2013/
09/04/denmark-ribbon-cut-at-anholt-offshore-wind-farm/. Accessed on 13 May 2014
17. www.drmrenfrew.files.wordpress.com
18. Earth Policy Institute. Climate, energy and transportation, world cumulative wind turbine
installations (online). Available at: http://www.earth-policy.org. Accessed on 25 Mar 2013)
19. IEA Wind. IEA wind task 26; the past and future cost of wind energy (online) .Available at:
http://www.nrel.gov. Accessed on 20 Nov 2012
20. ABB. Wind Power Plants. Technical application papers no. 13 (Online). Available at: http://w
ww05.abb.com. Accessed on 13 Dec 2012
21. Jeppsson J, Larsen PE, Larsson A (2013) The Swedish Energy Agency, Vattenfall. Technical
description lillgrund wind power plant (online). Available at: http://www.vattenfall.se.
Accessed on 23 Mar 2013
22. Pepper L (2012) Monitoring and evaluation of Blyth offshore wind farminstallation and
commissioning (online). Available at: http://webarchive.nationalarchives.gov.uk. Accessed on
12 Sept 2012
23. Gerdes G, Tiedemann A, Zeelenberg S (2012) Case study: Modelling offshore wind farmsa
survey for the analysis of the experiences and lessons learned by developers of offshore wind
farms (Online). Available at: http://www.offshore-power.net. Accessed on 03 Dec 2012
24. Wind Turbine Installation Vessel Aeolus (2014) [Online]. Available at: http://www.vanoord.
com/activities/offshore-wind-equipment. Accessed on 10 May 2014
25. Maintenance of Offshore Wind Turbine (2013) [Online]. Available at: www.power-
technology.com. Accessed on 10 July 2013
26. See http://www.earth-policy.org/ for information about cumulative installed capacity of PV
systems
References 15
27. See http://www.siemens.de/ for information about photovoltaic power plants

28. Du Y, Lu DDC, James G, Cornforth D (2013) Modeling and analysis of current harmonic
distortion from grid connected PV inverters under different operating conditions. Sol Energy
94:182194
29. Islam MR, Guo YG, Zhu JG (2012) A transformer-less compact and light wind turbine gen-
erating system for offshore wind farms. In: Proceedings of the 2012 IEEE international con-
ference on power and energy, 25 Dec 2012, Kota Kinabalu, Malaysia, pp 605610
30. Swiss-solar-power-plant-transformers-roof (2014) [Online]. Available at: http://www.
sunwindenergy.com/news/swiss-solar-power-plant-transformers-roof. Accessed on 13 May
2014
31. Islam MR, Guo YG, Zhu JG (2014) Multilevel converters for step-up transformer-less direct
grid integration of renewable generation units with medium voltage smart microgrids. Large
scale renewable power generation: advances in technologies for generation, transmission and
storage. Springer, Berlin, Heidelberg, pp 127149
32. Islam MR, Guo YG, Zhu JG (2014) A multilevel medium-voltage inverter for step-up-trans-
former-less grid connection of photovoltaic power plants. IEEE J Photovoltaics 4(3):881889
33. Islam MR, Guo YG, Lin ZW, Zhu JG (2014) An amorphous alloy core medium frequency mag-
netic-link for medium voltage photovoltaic inverters. J Appl Phys 115(17):17E710-117E710-3
34. Islam MR, Guo YG, Zhu JG (2013) A medium-frequency transformer with multiple second-
ary windings for medium-voltage converter based wind turbine generating systems. J Appl
Phys 113(17):17A324-117A324-3
35. Islam MR, Guo YG, Zhu JG (2014) A high-frequency link multilevel cascaded medium-
voltage converter for direct grid integration of renewable energy systems. IEEE Trans Power
Electron 29(8):41674182
36. To HP, Rahman MF, Grantham C (2007). Time delay compensation for a DSP-based cur-
rent-source converter using observer-predictor controller. In: Proceedings of the seventh
international conference on power electronics and drive systems, 2730 Nov 2007, Bangkok,
Thailand, pp 10911096
Chapter 2
Power Converters for Small- to Large-Scale
Photovoltaic Power Plants
Abstract This chapter presents an extensive literature survey on various different

aspects of medium-voltage (MV) converter development for step-up-transformer-
less direct grid integration of photovoltaic (PV) power plants. The main objective
was to show how power electronic converter topologies, power electronic devices,
and control complexities have affected the development of the MV converter
and how to make an excellent choice of the converter topology for step-up-trans-
former-less grid integration through the MV converter. Besides the traditional sys-
tem, which requires a step-up transformer to connect the renewable power plants
to the grids, other recently proposed converter topologies for step-up-transformer-
less direct grid interconnection are also introduced in detail with the aim of pre-
senting a complete picture of power converter topologies for small- to large-scale
PV power plants.
Keywords Power converters Photovoltaic inverters Small to large scale Power

plants Remote area Medium voltage
2.1Introduction
The worlds energy demand is growing remarkably due to the fast growth of popu-
lation and economy in the developing countries. The energy sector is facing an
accelerating amalgam crisis of the worldwide established fossil and atomic energy
systems. Natural gas, coal, and crude oil are the main fossil fuels for the current
world energy supply. Crude oil is the most important fossil fuel among the three
main fossil flues. Figure2.1 shows the trend of world oil consumption [1]. Coal is
the second most important consumable fossil fuel. Figure2.2 shows the world coal
consumption [1]. Due to very high oil price, coal has been becoming an attractive
fossil fuel in the recent years. Figure2.3 shows the world consumption of natural
gas [1]. Almost a constant growth rate has been observed for the natural gas con-
sumption in the past decades.

18 2 Power Converters for Small- to Large-Scale Photovoltaic
90
Consumption (Million
80
barrels/day)
World Oil Consumption
70
60
50
1985 1990 1995 2000 2005
Year
Fig.2.1World oil consumption [1]
120
Quadrillion BTU
World Coal Consumption
100
80
60
1980 1985 1990 1995 2000 2005
Year
Fig.2.2World coal consumption [1]
3000
Consumption (Million
2500 World Gas Consumption

Tons OE)
2000
1500
1000
1980 1985 1990 1995 2000 2005
Year
Fig.2.3World natural gas consumption [1]
The increasing energy demand is not only diminishing the reserve of fos-
sil fuels, but also affecting the environment. Carbon dioxide (CO2) gas is gener-
ated from burning of fossil fuels, which significantly contributes to the increase
of average global temperatures, i.e., global warming. Figure2.4 shows the world
CO2 emission from fossil fuel burning [1]. Scientists worldwide are now seeking
2.1Introduction 19
CO2 Emissions (Milion Tons)

10000
8000 CO2 Emissions from Fossil Fuel

Burning
6000
4000
2000
0
1950 1960 1970 1980 1990 2000 2010
Year
Fig.2.4World CO2 emission form fossil fuel burning [1]
Biodiesel
38.6 %
Solar PV Average Annual World
35.2 % Energy Growth Rates
by Source
Wind
27.9 %
Geothermal
22.1 %
Fuel
Ethanol
15.1 %
Natural Coal
Hydro Gas
Oil Nuclear 3.9 %
2.3 % 2.4 %
1.1 % 0.7 %
Fig.2.5Average annual world energy growth rates by source [1]
solutions to these two enormous challenges (energy and environment) from renew-
able or clean energy sources, which are richly available in almost every country.
Many countries have set targets for renewable energy use to meet the increas-
ing energy demand and also to reduce the global warming effect. For example, the
target shares of total energy from renewables by 2020 in Sweden, Finland, Austria,
and Australia are 49, 38, 34, and 20%, respectively [2]. Therefore, average annual
growth rates of renewable sources are much higher than those of conventional
sources in recent years. Figure2.5 shows the average annual world energy growth
rates [1]. Solar photovoltaic (PV) represents the second highest growth rate due to
its abundance source and technological development of PV cell, e.g., reduction of
PV module cost. Average module cost was USD 100/W and USD 29/W in 1975
and 1980, respectively, and reduced to less than USD 3.5/W in 2004. Figure2.6
shows the average PV module cost [1].
100 1980 1998

Module cost per Watt (USD)
80 Less than
Less than
30 USD/Watt 5 USD/Watt
60
40
About 3.5
20 USD/Watt
5 USD/Watt
0
1975 1980 1985 1990 1995 2000 2005
Year
Fig.2.6Average photovoltaic module cost [1]
Since 2007, medium- and large-scale PV power plants have attracted a high
degree of attention and the power plants of more than 10MW in capacity have
now become a reality. These multi-megawatt PV power plants require large areas
of land, and thus, they are usually installed in remote areas, far from cities. For
power transmission, a step-up transformer is usually used in the PV inverter sys-
tem to feed in the solar energy into a medium-voltage (MV) grid (e.g., 636kV).
The transformer steps up the inverter output voltage from 300V AC to grid volt-
age level (e.g., 636kV). Although these special transformers are compact com-
pared with conventional distribution transformers, they are still large and heavy for
remote area PV applications. The large-size and heavy weight step-up transformer
may increase the system weight and volume and can be expensive and complex for
installation and maintenance. The MV inverter may offer the best possible solution
to interconnect the PV array to the MV grid directly [3]. Moreover, it may also
be possible to ensure electrical isolation through the inverter, which is important
for the interconnection of MV grid and PV array [4]. Recently, advanced mag-
netic materials, such as amorphous and nanocrystalline alloys, have attracted sig-
nificant attention to develop high-frequency magnetic links for MV inverters [5].
Compared with the power frequency transformer (operated at 50/60Hz), the high-
frequency magnetic links (in the range of a few kHzMHz) have much smaller and
lighter magnetic cores and windings and thus much lower cost. Therefore, the MV
inverter for step-up-transformer-less direct grid interconnection of PV systems has
become a favorable choice, since the installation of large-scale PV power plants
started commercially in 2007.
This chapter incorporates PV power generation technologies, including tradi-
tional power conditioning systems, two-level low-voltage converter topologies,
the limitations of power frequency step-up transformer-based grid integration of
renewable generation systems in remote area applications, and advanced converter
topologies for MV applications. Therefore, an extensive literature survey has been
conducted focusing on many different aspects of MV converter development for
2.1Introduction 21
step-up-transformer-less direct grid integration of PV power generation systems.

The main objective of this general review on power electronic converter topolo-
gies is to show how topologies, power electronic devices, and control complexities
have affected the development of the MV converter and to understand which con-
verter topology would be a natural choice for step-up-transformer-less grid inte-
gration through the MV converter.
2.2Solar Photovoltaic Arrays
The solar photovoltaic means producing electricity from sunlight. A PV cell

can be made by two semiconductor layers; one having positive charge and the
other having negative charge. When the sunlight shines on a PV cell, some of the
photons from the sunlight are absorbed by the semiconductor atoms, releasing
electrons from the cells negative layer. If there is an external circuit, the free elec-
trons flow to the positive layer and produce electric current in the external circuit.
Figure2.7 shows a photograph of PV cell.
One silicon PV cell can produce about 0.5V, which is too small to do valu-
able work. To increase the voltage, several individual PV cells are interconnected
together in package called a module. For example, a 12-V module will have 36
individual cells connected in series. In 24V applications, two 12-V modules can
be connected in series, where the current stays unchanged. Figure2.8 shows a pho-
tograph of 24-V PV module. Table2.1 summarizes the electrical characteristics of
solar module SX-10. When two modules are wired in parallel, their current is dou-
bled and the voltage stays unchanged. To achieve the desired voltage and current,
Fig.2.7A photograph of
PV cell, which may produce
0.5V DC
24-V PV module, where 72
cells are interconnected
Cell
Table2.1Electrical Parameters Rating

characteristics of solar
Nominal voltage 12V
module SX-10
Rated power 10W
Number of PV cells 36 (connected in
series)
Voltage at Pmax (Vmp) 16.8V
Current at Pmax (Imp) 0.59A
Short-circuit current (Isc) 0.65A
Open-circuit voltage (Voc) 21.0V
Temperature coefficient of Isc (0.0650.015)%/C
Temperature coefficient of Voc (8010)mV/C
Temperature coefficient of power (0.50.05)%/C
Fig.2.9A photograph Cell

of PV array with six 12-V
modules, where 36 cells
are interconnected in each Module
module
a number of modules can be connected in series and parallel. The seriesparallel

combination of modules is called a PV array. Figure2.9 shows a photograph of PV
array with six 12-V modules, where 36 cells are interconnected in each module.
2.2.1Solar Photovoltaic Array Modeling
When PV arrays are used to harvest solar energy, two important factors could limit
the implementation of PV systems, i.e., high cost and low efficiency in energy
conversion. The conversion efficiency of the current solar PV modules is typically
only about 1017% [6]. In PV systems, the PV array represents about 57% of
the total cost of the system, and the battery storage system corresponds to 30%
of the cost. Other system components such as inverters and maximum power
point tracker (MPPT) contribute to only 7% of the total cost [7]. Due to the low
2.2 Solar Photovoltaic Arrays 23
Fig.2.10Equivalent circuit Icell R scell

of a PV cell
Rshcell
D Vcell
IL
conversion efficiency and high cost of solar array, it is very desirable to operate the
PV panel at the maximum power point (MPP). An ideal solar cell can be modeled
by a current source in parallel with a diode. In practice, no solar cell is ideal and
hence a shunt resistance and a series resistance are added to the model as shown
in Fig.2.10, where Rscell is the intrinsic series resistance of usually a very small
value, and Rshcell is the equivalent shunt resistance of usually a very large value.
For a single silicon solar cell, the nonlinear IV characteristic can be presented
as [8, 9]
Vcell + Icell Rscell
Icell = IL Io exp (G(Vcell + Icell Rscell )) 1 (2.1)
Rshcell
where
G = AiqKT ,
q is the electronic charge (=1.6021019C),
Ai=Bi is the ideality factor (=1.92),
K is the Boltzmanns constant (=1.381023J/K),
T is the PV cell temperature,
Icell is the cell output current,
Vcell is the cell output voltage,
Io is the cell saturation current which can be presented as
3
T qEGO 1 1
Io = Ior exp (2.2)
Tr Bi K Tr T
IL is the light-generated current which can be presented as
Rad
IL = ISC + KISC (TC 28) (2.3)
1,000
The array temperature Tc is approximately given by [9]
TC = Tair + 0.3 Rad % (2.4)
Tr is the reference temperature (=301K),

Ior is the reverse saturation current at Tr (=19.9693106A),
Ios is the cell reverse saturation current, A
Tc is the cell temperature, C
KIsc is the short-circuit current temperature coefficient (=0.0017A/C)
Fig.2.11Simplified Icell R scell

equivalent circuit of a PV cell
D Vcell
IL
Rad is the cell illumination, W/m2 (1,000W/m2=100% illumination)

Isc is the cell short-circuit current at 28C and 1,000W/m2 (=2.52 A)
EGo is the band gap for silicon (=1.11eV)
Tair is the ambient temperature, C
Since PV arrays are built up with series and/or parallel connected combinations
of solar PV cells, for an array with nS nP cells, the current equation can be pre-
sented as [9, 10]

VPV + IPV Rs VPV + IPV Rs
IPV = np IL np Io exp G 1 . (2.5)
ns Rsh
where
IPV = nP Icell is the PV array output current,
VPV = nS Vcell is the PV array output voltage,
ns is the number of cells connected in series,
n p is the number of panels connected in parallel,
RS = RScell nnPS is the PV array series resistance, and
Rsh = Rshcell nnSP is the PV array shunt resistance.
The shunt resistance Rshcell is much greater than the series resistance Rscell,
which makes the last term of (2.1) much smaller than the other terms. The simpli-
fied cell equivalent circuit is shown in Fig.2.11.
The currentvoltage relation of simplified PV cell can be expressed as

Icell = IL Io exp (G(Vcell + Icell Rscell )) 1 (2.6)
and the array currentvoltage relation becomes

VPV + IPV Rs
IPV = np IL np Io exp G 1 . (2.7)
ns
2.2.2Solar Photovoltaic Array Characteristics
Solar energy sources have variable daily and seasonal patterns. For exam-
ple, monthly average global solar insolation at Dhaka City of Bangladesh var-
ies between 3.92 and 7.71kWh/m2/day. The maximum amount of insolation
Fig.2.12Monthly average
Average insolation (kWh/m2/day)

6
global solar insolation at
Dhaka City of Bangladesh; 5.5
recording period: 19881998
[11] 5
4.5
3.5
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
Fig.2.13Daily average
Average bright sunshine hours
9
bright sunshine hour at
Dhaka City of Bangladesh; 8
recording period: 19611980
7
[11]
6
3
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
is available in the months of MarchMay, and the minimum in December

January. Figure2.12 shows the monthly global solar insolation at Dhaka City of
Bangladesh [11]. Daily bright sunshine hours have also variable daily and sea-
sonal patterns. For example, the daily average bright sunshine hour at Dhaka City
of Bangladesh vary between 4.9 and 9.1h. The maximum is in the months of
FebruaryApril, and the minimum is in the JuneSeptember. Figure2.13 shows
the daily average bright sunshine hours at Dhaka City of Bangladesh [11]. The
output characteristics of the PV array are nonlinear and critically affected by the
solar radiation, temperature, and load conditions. The current versus voltage (IV)
and power versus voltage (PV) characteristics with various irradiances at 25C
temperature are shown in Figs.2.14 and 2.15, respectively [12, 13]. The simulated
IV and PV characteristics with various temperatures at 1,000W/m2 irradiance
are illustrated in Figs.2.16 and 2.17, respectively [12, 13].
2.2.3Solar Photovoltaic Array Maximum Power Point
It is also observed from the powervoltage curve of the solar PV module that on
the right hand side, when the voltage is almost constant, the slope of power versus
5
2
1000 W/m
Current, I (A)
4
800 W/m 2
3
600 W/m2
2
0
0 2 4 6 8 10 12 14 16 18 20 22
Voltage, V (V)
Fig.2.14Calculated IV curves at different irradiance ranging from 600 to 1,000W/m2; 25C

temperature was considered for the calculation [12, 13]
70
60 1000 W/m 2
800 W/m 2
50
Power, P (W)
600 W/m2
40
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20 22
Voltage, V (V)
Fig.2.15Calculated PV curves at different irradiance ranging from 600 to 1,000W/m2; 25 C

temperature was considered for the calculation [12, 13]
voltage is negative (P/V<0), whereas on the left hand side, the slope is posi-
tive (P/V>0), as illustrated in Fig.2.18 [12, 13]. Thus, the PV array has an
optimum operating point called the MPP, which varies depending on array tem-
perature and the present insolation level. The irradiance mainly affects the output
current, and the temperature mainly affects the terminal voltage of the PV array,
so that the effects of both factors have to be considered when designing a PV sys-
tem. However, the intermittent nature of PV sources, in terms of the power and
output voltage, is a major issue when connected to the grid. Therefore, solar PV
system requires a power conditioning circuit known as PV inverter that is capable
of extracting the maximum power from PV source and feeds the adjusted power to
the load and/or grid to their satisfaction.
5
o
Current, I (A) 75 C
4
o
50 C
3 o
25 C
2
0
0 5 10 15 20
Voltage, V (V)
Fig.2.16Calculated IV curves at different temperature ranging from 25 to 75C; irradiance

1,000W/m2 was considered for the calculation [12, 13]
70
60
50
Power, P (W)
o
75 C
40
o
30 50 C
o
20 25 C
10
0
0 5 10 15 20
Voltage, V (V)
Fig.2.17Calculated PV curves at different temperature ranging from 25 to 75C; irradiance

1,000W/m2 was considered for the calculation [12, 13]
Fig.2.18Maximum power
point (MPP) determination: dP/dV=0
The slope is positive on the
left side of the PV curve,
Power, P (W)
negative on the right side, and

0
dP
V>
zero at the MPP [12, 13]

/dV
/d
dP
MPP
<0
Voltage, V (V)
Fig.2.19Fundamental Stage 1 Stage 2

circuit of two-stage PV
inverter [14] L
PV array
AC line
Cs Cd
Boost & MPPT Inverter
Fig.2.20Most commonly L
iL io
used DC to DC boost
converter circuit
VL
+
+
-
+ Diode
Vi Vo
Cs Cd
- Transistor -
2.3Inverters in Small-Scale Solar PV Systems
2.3.1Two-Stage Solar PV Inverters for Small-Scale Systems

Usually, a PV inverter has two stages to shape the PV array output power for feed-
ing into the AC load. The first stage is responsible for boosting the PV array volt-
age and tracking the MPP, and the second stage inverts available DC power into
AC power. The most fundamental circuit topology of the two-stage PV inverter
using the full-bridge as shown in Fig.2.19 was proposed in [14].
Usually, the source-side controller tracks the maximum power and the load-
side controller controls the power factor of output current and maintains constant
DC voltage across capacitor Cd. When the transistor is turned on, the energy from
PV array is stored in the inductor L. When transistor is turned off, the inductor
stored energy is delivered to the DC-link capacitor Cd and the H-bridge inverter.
The booster circuit can have two distinct modes of operation, the continuous cur-
rent conduction and discontinuous current conduction, of significantly different
characteristics. The booster circuit, as shown in Fig.2.20, is also known as the
step-up DC to DC converter; its typical application is to convert low input voltage
to a high output voltage. When the transistor is turned on, the source voltage Vi is
applied across the inductor and the rising rate of inductor current is dependent on
the source voltage Vi and inductance L. Figure2.21 shows the equivalent circuit of
the DC to DC boost converter when the transistor is turned on.
When the transistor is turned off, the equivalent circuit is shown in Fig.2.22
and the inductor voltage becomes
vL = Vi Vo (2.8)
2.3 Inverters in Small-Scale Solar PV Systems 29
Fig.2.21Equivalent iL L
circuit of a DC to DC boost
converter when transistor is + + vL - +
turned on Vi Cs Co Vo
- -
Fig.2.22Equivalent iL L
circuit of a DC to DC boost
converter when the transistor + vL -
+ +
is turned off
Vi Cs Co Vo
- -
Fig.2.23Voltage across
inductor L for time period T
vi
A
toff
0
ton t
B
(vi-vo)
T
For steady-state operation, the integral of inductor voltage vL over one time period
T must be zero, i.e., the areas A and B in Fig.2.23 must be equal.
Therefore,
Vi ton + (Vi Vo ) toff = 0 (2.9)
Dividing both sides by T, and rearranging all terms, we have
Vo T 1
= = (2.10)
Vi toff 1D
Assuming the circuit is 100% efficient, i.e., the input power (Pi) and output power
(Po) are equal, (Pi=Po), or
Vi Ii = Vo Io (2.11)
one obtains
Io
= (1 D) (2.12)
Ii
iL
IL
0
t
T
t on toff
Fig.24Waveform of inductor current in the continuous conduction mode
D1
L D2
PV array
AC line
Cs Cd
Boost chopper Inverter
Fig.2.25Time-sharing chopper-based-modified circuit of two-stage PV inverter
L
D2
PV array
T1 T3
AC line
Cs Lr Cr
C
D1 T2
Fig.2.26Parallel resonant DC-link-based-modified two-stage PV inverter circuit
In the continuous operation mode, the inductor current never falls to zero in one
switching cycle, i.e., either the transistor or diode is conducting. Figure2.24
shows the inductor current in the continuous conduction mode.
Several modified topologies, such as a time-sharing dual-mode PV inverter [15]
in Fig.2.25 and a soft-switched DC to DC boost converter-based inverter [16] in
Fig. 2.26, were proposed to improve the efficiency of the fundamental circuit. In
the time-sharing dual-mode PV inverter, when the PV array output voltage is larger
than the DC-link voltage, the transistor is always in the off state and the input cur-
rent flows through the bypass diode D1 but does not flow through the boost inductor
L and free-wheeling diode D2. In this way, the proposed circuit can avoid the con-
duction losses of boost inductor, L, and the free-wheeling diode D2. When the input
voltage is smaller than the DC-link voltage, the bypass diode D1 behaves as open
circuit and the circuit operates as a fundamental two-stage PV inverter circuit. In the
soft-switched DC to DC boost converter-based PV inverter, there are two stages: the
converter stage and inverter stage. All the switches in both the converter and inverter
stages can be turned on and off with zero voltage switching, which can reduce the
switching losses significantly. The operating principle of the proposed soft switch-
ing inverter was presented with a few switching modes [16] as the following:
Mode 1 When the transistor in DC-link capacitor C is turned on with zero volt-
age switching, the DC-link capacitor begins to discharge and linearly
decreases the main inductor L current.
Mode 2 When the transistors T1, T2, and T3 are turned on with zero voltage
switching, the DC-link capacitor begins to discharge.
Mode 3 When T1 and T2 are turned on with zero voltage switching, the resonance
between resonance inductor, Lr, and resonance capacitor, Cr, is started and
the main inductor current is minimized. When the resonance is finished,
the current of resonant inductor, Lr, flows through diodes, D1 and D2.
Mode 4 When T1 and T2 are turned off with zero voltage switching, resonance
is started. The resonant capacitor Cr is charged by the current flowing
through inductors, L and Lr. The resonance stops when the voltage of res-
onant capacitor, Cr, equals the output voltage. The DC-link capacitor is
charged by L and Lr through the parallel diode of T3.
2.3.2Multiple-Stages Solar PV Inverters for Small-Scale

Systems
Due to the high value of stray capacitance of PV array, a galvanic connection

between the ground of the grid and the PV array may exist. In this case, danger-
ous leakage current called common-mode current may appear through the stray
capacitance between the PV array and the ground. The capacitance between PV
cells and the ground can reach very high value under certain conditions, such as
the PV array structure, and weather there are humidity and dust covering the PV
array. This capacitance can reach up to 150nF/kWp for crystalline-silicon cells and
1F/kWp for thin-film cells. The common-mode leakage current may increase the
system power loss, reduce the grid current quality, and cause personnel safety prob-
lems [17, 18]. Therefore, galvanic isolation is essential because it permits easy array
grounding, array isolation from grid in case of fault, and safety of personnel, and
can strongly reduce the leakage current between the PV array and the ground. The
connection of grid neutral and middle points of the DC link is one of the options to
reduce the ground leakage currents. In this context, Rahman and Zhong [19] pro-
posed a half-bridge topology-based PV inverter as shown in Fig.2.27. The step-up
Fig.2.27Basic circuit of
half-bridge topology-based
PV inverter L D
PV array
AC line
C1
Cs C2
Q
Boost & MPPT Inverter
Fig.2.28Sunny Boy
5000TL model half-bridge L
D
topology-based PV inverter
PV array
Cin c
L
D
PV array
AC line
C
Cin c
Inverter
L
D
PV array
Cin c
MPPT & booster
(boost) converter raises the voltage of the PV array and also serves the operation of
MPPT. Driven by the proper reference signals generated by the control algorithm
to modulate the pulse width of the switching signal of switch Q, the half-bridge
inverter inverts the DC power to 50 or 60Hz AC power. Figure2.28 illustrates the
half-bridge topology-based three-string PV inverter, Sunny Boy 5000TL, commer-
cially developed by SMA [20]. The half-bridge inverter-based topology requires
less switching devices, but its DC-link voltage needs to be twice the grid voltage
peak. The inverter topologies do not have electrical isolation between the array and
the grid, which is critical in case of fault and safety of personnel.
The insulation can be achieved through a power frequency transformer at
the grid side of the PV inverter which may also serve voltage step-up operation.
Filter
L
PV array
AC line
Cin
HFT-L
Inverter Rectifier Inverter
with MPPT
Fig.2.29High-frequency transformer-link (HFT-L)-based PV inverter power circuit
This power frequency transformer is heavy and large, increasing the PV inverter
installation cost and requiring regular monitoring and maintenance. Increasing
the operating frequency will lead to a compact and lightweight magnetic com-
ponent of isolation transformers [4, 5, 21]. Several medium- and high-frequency
(HF) transformer-based inverter topologies were developed and made available
commercially. Figure2.29 shows a high-frequency link-based inverter system
[10]. The DC PV array power is converted to 50 or 60Hz AC line power through
an isolated high-frequency transformer link. The DC voltage of the PV array is
firstly converted to high-frequency AC by a high-frequency inverter, which is then
transformer-coupled and converted to 50 or 60Hz AC through a high-frequency
rectifier, filter, and a full-bridge inverter. The multistage power conversion may
increase the cost of the converter and decrease the efficiency of the system, but
it can significantly reduce the weight and volume of the power conversion sys-
tem and minimize grid isolation issues. In 2010, Lu etal. [22, 23] proposed a pla-
nar high-frequency transformer-based PV inverter system as shown in Fig.2.30.
Higher output power from multiple PV arrays can be achieved by connecting each
PV array to its own DC to AC converter and single-phase transformer. The prima-
ries of the high-frequency transformer links are connected in an open delta ener-
gized by high-frequency voltage from DC to AC converters, where the converter
DC/DC converter
DC/AC
Inverter
0o
AC output
PV modules
DC/AC DC/AC
Inverter Inverter
120o
DC/AC
Inverter
240o
HF transformer
Fig.2.30High-frequency (HF) transformer-based PV inverter power circuit

L
PV array
Cin c
HFT
L
PV array
AC line
C in c
HFT
Inverter
L
PV array
Cin c
HFT
HF Inverters with booster Rectifiers
Fig.2.31High-frequency transformer (HFT)-based PowerLynx Powerlink PV 4.5kW inverter
output voltages are phase shifted each other by 120. The secondaries of the high-
frequency transformer links are connected in a wye configuration, and the output
is connected to a 50 or 60Hz DC to AC converter through a high-frequency three-
phase rectifier. The detailed design and analysis of the HF planar transformer were
reported in [24, 25]. The Original Equipment Manufacturer commercially devel-
oped the high-frequency transformer-based three-string PV inverter, PowerLynx
Powerlink PV 4.5kW, as shown in Fig.2.31 [26].
2.3.3Single-Stage Solar PV Inverter for Small-Scale

Systems
Compared to the single-stage one, the multistage power conversion is somewhat

more expensive and affects the efficiency of the PV inverter. In order to reduce
the volume and weight as well as the power conversion loss and cost, a hybrid PV-
battery-powered DC bus system was proposed in 2009 [27]. The DC to AC con-
version stage-less DC bus system is very applicable to electronic equipment and
appliances with high system efficiencies. The PV-battery-powered DC bus system
DC bus DC bus
High step-up High step-up AC AC

Rectifier
DC/DC converter DC/DC converter adaptor adaptor
Load # 1 Load # N
Battery Battery AC mains

PV PV
Module 1 Module N
Fig.2.32PV-battery-powered DC bus system
Fig.2.33Full-bridge with
MPPT-based circuit topology
of single-stage inverter PV array
AC line
Cin
Inverter
with MPPT
is shown in Fig.2.32. For AC systems, a single-stage PV inverter was proposed in

[28], and the circuit topology of single-stage inverter is shown in Fig.2.33.
The proposed inverter performs a dual function: MPPT and outputting a sinu-
soidal current, which makes the control circuit complex. In [29], an alternative
control technique was developed to reduce the complexity of the control circuit.
However, the common-mode issue was not considered in the proposed single-
stage inverter systems. The neutral point clamped (NPC) converter topology has
the opportunity to connect the grid neutral point to middle point of the DC link,
reducing the ground leakage currents. In this context, an NPC topology-based
single-phase PV inverter as shown in Fig.2.34 was presented in [30] and a three-
phase PV inverter system in Fig.2.35 was implemented in [31]. Since the pre-
sented circuits are run as buck converters, the PV array voltages should be greater
than the peak values of the output AC voltages. If Vrms is the inverter output AC
voltage and R is the reservation factor, the minimum array voltage can be calcu-
lated as

VA = 2Vrms R. (2.13)
Therefore, a few PV arrays in series connection are necessary to obtain the desired
voltage. From the available literature, several single-stage topologies have been
proposed based on either boost or buckboost configurations. An integrated
Inverter
PV array
C1
AC line
PV array
C2
Fig.2.34Circuit topology of single-stage inverter: NPC with grid neutral connected to the mid-
dle point of DC link
CPV
PV array
RG
C
PV array
CPV
RG
AC lines
Fig.2.35NPC topology-based three-phase PV inverter
(boostconverter and full-bridge inverter) PV inverter circuit topology shown in

Fig.2.36 was presented in [32]. The output power quality and the efficiency of the
inverter are limited by the fact that the boost converter cannot generate the output
voltage lower than the input voltage. A universal single-stage PV inverter shown in
Fig. 2.37 was presented in [33] that can operate as a buck, boost, or buckboost
converter. This inverter can operate with a wide range of input voltage, improving
the power quality and the efficiency. Using the integrated buckboost and inver-
sion functions, several modified configurations have been presented in [34, 35].
However, these topologies are only suitable for small-scale (e.g., <100kW) PV sys-
tems, where the PV array normally interconnects with a low-voltage public network.
Fig.2.36Single-stage
power circuit with boost
converter Lb
PV array
AC line
Cd
Inverter with
boost converter
Fig.2.37Single-stage
power circuit with universal
converter L
D1
PV array
AC line
Cd D2
Inverter with universal

converter
2.3.4ABB and Siemens Solar PV Inverters for Small-Scale

Systems
Although the galvanic isolation is not a requirement in Germany in the low-voltage

public network, a residual current monitoring unit needs to be installed for the grid
connection of transformer-less PV inverters. Most standards (e.g., IEEE 1547,
EN 61000-3-2, and IEC 61727) allow the DC current injection to vary between
0.5 and 1% of the rated current [36]. Therefore, the transformer-less inverter sys-
tem should have some extra protective devices to ensure the safety issues. ABB is
one of the leading manufacturers of PV inverters. In order to interconnect the PV
array to the public electricity network, ABB developed the transformer-less single-
phase string PV inverter model PVS300, suitable for small-scale PV systems (i.e.,
3.38.0kW). The PVS300 inverter converts the DC generated by PV arrays at a
voltage ranging 335800V into single-phase 230V AC that can be fed into the
public network through protective devices as shown in Fig.2.38 [37]. Due to the
high content of harmonics in the output power and the use of high switching fre-
quency of switching devices, the system consists of a heavy filter circuit and elec-
tromagnetic interference (EMI) filter, which may increase the losses and cost of
the system.
The SINVERT PVM inverters are transformer-less, 3-phase PV inverters devel-
oped by Siemens, to convert solar DC energy at a voltage ranging in 380850V into
3-phase 400V AC voltage for feeding into public electricity network. SINVERT
PVM inverters are available with power outputs from 10 to 20kW. The PV array
Surge Inverter
protection
Solar panels L
Public network
=
DC switch
EMI filter
LCL filter
AC switch
EMI filter
~ N
Control and monitoring
Fig.2.38ABB PVS300 string inverter design and grid connection

Solar panels
Public network
= L1
~ L2
L3
SINVERT PVM
inverter N
PE
Overvoltage
protection
M8
Fig.2.39Siemens SINVERT PVM string inverter design and grid connection
interconnection to utility grid through SINVERT inverter with safety devices is

shown in Fig.2.39 [38].
2.4Medium- and Large-Scale Solar PV Systems
For medium- and large-scale solar PV electricity generation, there are two well-
established inverter technologies: the centralized and string technologies as shown
in Figs.2.40 and 2.41, respectively.
In the centralized PV inverter technology, each string consists of a series of PV
modules to reach the voltage requirement without amplification and then a few
strings are parallel connected to a common inverter circuit. The number of strings
mainly depends on the power levels. This technology eliminates the amplifica-
tion stage, but possesses some major limitations, such as voltage mismatch loss
2.4 Medium- and Large-Scale Solar PV Systems 39
AC output
PV modules DC/AC
Inverter
PV modules
String diodes
PV modules
Fig.2.40Centralized PV inverter topology DC

DC
PV modules
DC
AC output
DC
DC/AC
Inverter
DC
DC
Fig.2.41String PV inverter topology
between PV modules and power losses due to common MPPT [26]. In order to
control every string individually, the string technology uses a DC to DC converter
for each string, which improves the system efficiency. In order to minimize the
voltage mismatch in the strings, a battery-integrated boost converter was pro-
posed to eliminate the voltage regulation stage [39]. The block diagram of battery-
integrated boost converter is shown in Fig.2.42, and the proposed converter-based
PV inverter system is shown in Fig.2.43. The detailed analysis of the proposed
converter is presented in [40].
The ABB central inverters are especially designed for medium-scale PV power
plants. The PVS800 version is a 3-phase inverter with a power capacity in the
Battery
D
L
Vout
S _
PV
Fig.2.42Battery-integrated boost converter

Battery 1
+
D
L1
Vout_1
C
S _
PV 1
+
Battery 2
AC output
L2
DC/AC
Vout_2 Inverter
C
S _
PV 2
+
Battery n
D
Ln
Vout_n
C
S _
PV n
Fig.2.43Battery-integrated boost converter-based PV inverter system
range of 100500kW. The PVS800 inverter topology allows a parallel connection

directly on the AC side, for grid connection through a step-up power frequency
transformer [41]. The transformer steps up the inverter output voltage from 300V
AC to grid voltage level (e.g., 636kV). The central inverter design and grid con-
nection are depicted in Fig.2.44. ABB has been delivering worldwide vacuum cast
coil dry-type transformers for PV applications. The cast coil dry-type transform-
ers are non-flammable and moisture proof. They feature a solid isolation system
which is discharge-free. However, the volume and weight of a 0.4/36kVA, 1 MVA
vacuum cast coil transformer are about 4.3m3 and 3,250kg, respectively, and no-
load and full-load losses are 3.1 and 11.5kW, respectively [42]. Moreover, dry-
type transformers can be sensitive to water, microcracks, temperature variations,
PVS800 Inverter 1
Switch
DC
EMC filter
DC switch
EMC filter
LCL filter
AC
PV array
Medium-voltage
Controller
network
transformer
Medium
step-up
voltage
PVS800 Inverter 2
Switch
DC
EMC filter
DC switch
EMC filter
LCL filter
AC
PV array
Controller
Fig.2.44ABB central inverter design and medium-voltage (MV) grid connection
and pollution which can block cooling ducts. Due to the use of traditional two-
level inverter, the harmonic content in the output power is high. Usually, high-
frequency pulse-width-modulated (PWM) gate signals are used to drive switching
devices to reduce the harmonic content. Due to high-harmonic content and high-
frequency stitching, the system requires a heavy LCL filter circuit and an electro-
magnetic compatibility (EMC) filter, which may increase the losses and cost of the
system.
Besides its low-voltage system, Siemens also developed the SINVERT
PVS inverter-based system for medium-scale PV plants. The AC output voltage
and power capacity of PVS version inverters are in the range of 288370V and
500630kW, respectively, as summarized in Table2.2. The 12.52MW central
inverters were designed by paralleling 24 PVS inverters through transformer and
switchgear at the grid side. The design and grid connection of 2-inverters-based
system is illustrated in Fig.2.45 [43]. Siemens developed GEAFOL castresin
transformers for grid connection of PV arrays. With GEAFOL, it is possible to
avoid the limitations associated with liquid-filled transformers while retaining the
Table2.2Capacity of PVS version inverters

Versions PVS500 PVS585 PVS600 PVS630
AC output voltage (V) 288 340 370 370
Power capacity (kW) 500 585 600 630
Min. input voltage (V) 450 530 570 570
SINVERT PVS 1
Medium voltage
Switch
Input 1
step-up
transformer
DC
EMC filter
EMC filter
Input 2
Filter
Input 3
AC
PV array
Medium-voltage
Controller
network
SINVERT PVS 2
Medium voltage
Switch
step-up
Input 1
transformer
DC
EMC filter
EMC filter
Input 2 Filter
Input 3
AC
PV array
Controller
Fig.2.45Siemens central inverter design and medium-voltage grid connection
properties such as operational safety, resistance to humidity, mechanical strength,

and compact design. However, no-load and full-load losses of a 0.4/30kV, 1 MVA
GEAFOL transformer are about 3.1 and 10kW, respectively, and the volume and
weight are about 3.5m3 and 2,990kg, respectively [44].
Although these special transformers are compact compared with conventional
distribution transformers, they are still large and heavy for PV applications. The
large-size and heavy-step-up transformer may increase the system weight and
volume and can be expensive and complex for installation and maintenance. The
medium-voltage inverter may be the best possible solution to interconnect the PV
array to the medium-voltage grid directly. Moreover, electrical isolation is impor-
tant for the interconnection of the medium-voltage grid and PV array. Since the
installation of large-scale PV power plants started commercially in 2007, the
medium-voltage inverter for interconnection of PV systems has been attracting
great attention.
In 2011, different MV multilevel inverter topologies were compared for the pos-
sible medium-voltage grid connection of wind turbine and PV systems [45, 46].
The modular multilevel cascaded (MMC) topology was considered as a possible
candidate for medium-voltage inverter systems. The MMC inverter requires multi-
ple isolated and balanced DC sources. In [47], a medium or HF link was proposed
to generate multiple isolated and balanced DC sources for the MMC from a single
source and in [48], a medium-frequency transformer link was developed to ver-
ify the feasibility of the new concept of voltage step-up using the MMC inverters.
Compared with the conventional transformers operated at the power frequency (50
Fig.2.46Quasi-Z source
C2
converter-based medium-
L1 L2
voltage PV inverter (one of 3 D
phases)
C3 C1
PV array 1 H-bridge 1
C2
L1 L2
D
Medium-voltage
network
C3 C1
C2
L1 L2
D
C3 C1
PV array n H-bridge n
or 60Hz), the medium-frequency transformer link has much smaller and lighter
magnetic cores and windings and thus much lower costs.
In 2012, by combining a quasi-Z source inverter into an MMC as shown in
Fig.2.46, a medium-voltage PV inverter was proposed in [49].
The proposed inverter does not have isolation between PV array and medium-
voltage grid. Multiphase isolated DC to DC converter-based MMC inverter topol-
ogy as shown in Fig.2.47 was proposed in [50, 51]. In the proposed configuration,
the voltage balancing is the challenging issue, since each H-bridge cell is con-
nected to a PV array through a DC to DC converter. A common DC link may be
one of the possible solutions to minimize the voltage imbalance problem, and a
single DC-link-based inverter in Fig.2.48 was presented in [52, 53]. Although this
design may reduce the voltage balancing problem in the grid side, the generation
of common DC-link voltage from different PV arrays makes the inverter operation
complex and limits the range of MPPT operation.
As an alternative approach to minimize the voltage imbalance problem with a
wide range of MPPT operation, a common magnetic link was proposed [3]. The
boost converter is considered for the MPPT operation. The array DC power is con-
verted to a high-frequency AC through a high-frequency inverter. The inverter also
DC
C2 C1
DC
DC
Medium-voltage
network
C2 C1
DC
DC
C2 C1
DC
PV array n H-bridge n
Fig.2.47Multiphase isolated converter-based medium-voltage PV inverter (one of 3 phases)
Common DC-link
DC
DC
DC
DC
DC
PV array 1
AC
DC
H-bridge 1
DC
Medium-voltage
DC
DC
PV array 2
DC
AC
network
DC
DC
H-bridge 2
PV array 3
DC
DC
DC
AC
DC
DC
H-bridge k
PV array n
Fig.2.48Common DC-link-based medium-voltage PV inverter (one of 3 phases)

Module A-1
MPPT
L D
Cdc
Cdc Cin
Q
Module A-2
Phase-A
Cdc
PV array-1 High-frequency
inverter
MPPT
Cdc
L D
Module A-N
Cdc Cin
Q
Module B-1
Medium voltage ac (e.g., 636 kV)

inverter
Cdc
A
Module B-2
Phase-B
MPPT B
Cdc
C
L D
Cdc Cin
Q
Cdc
n
inverter
Common magnetic-link
Module B-N
Module C-1
Cdc
Module C-2
MPPT
Phase-C
Cdc
L D
Cdc Cin
Q
Cdc
PV array-N High-frequency
inverter
Module C-N
Fig.2.49Common magnetic-link-based 3-phase medium-voltage PV inverter

ensures constant output voltage. The inverter is connected to a primary winding

of a multiwinding high-frequency magnetic link. Each secondary winding works
as an isolated source and is connected to an H-bridge cell through a bridge recti-
fier. The number of primary windings depends on the number of PV arrays, and
the number of secondary windings depends on number of levels of the inverter.
The detailed power circuit of common magnetic-link-based PV inverter system is
shown in Fig.2.49 [3]. In medium- or large-scale PV power plants, several PV
arrays are operated in parallel. The multiple-input multiple-output magnetic link
can incorporate the parallel operation of multiple PV arrays, where each PV array
is connected to a primary winding through a booster and high-frequency inverter
[54]. The magnetic link also provides electrical isolation between the PV array
and the grid, which can thus inherently overcome the common-mode and voltage
imbalance problems and ensure a wide range of MPPT operation and safety of
operating personnel.
2.5Summary
Besides the traditional systems, which require a step-up transformer to connect

the renewable generation systems to the grid, many other recently proposed con-
verter topologies for step-up-transformer-less direct grid interconnection are also
reviewed in detail with the aim of presenting a complete picture of power con-
verter systems. In order to develop a common converter for both the wind and PV
power generation systems, appropriate converter topologies should be investi-
gated. In this chapter, most of the existing power converters are reviewed. In order
to reduce the system volume and weight as well as to improve the efficiency, a
number of transformer-less single-stage inverter topologies have been proposed in
the past decades for small-scale PV power generation systems. The common-mode
or leakage current and personal safety are really critical issues with these inverter
systems. Most of the commercially available system uses the traditional two-level
inverter in which the switching devices are driven by high-frequency PWM gate
signals. Due to high-harmonic contents and high-frequency switching, the sys-
tem requires a heavy LCL filter circuit and an EMC filter, which may increase the
losses and cost of the system.
Today, the industrial trend is to move away from these heavy and large-size
passive components to compact and lightweight systems that use more and more
semiconductor devices in modular construction. It is found that multilevel con-
verters with medium-frequency links would be a feasible option to develop a MV
converter for the direct grid integration of PV power plants. Although different
multilevel converter topologies have been developed in the last few decades, most
of them are not suitable for medium-voltage applications. Because of some spe-
cial features, such as the number of components scaling linearly with the number
of levels, and the identical individual modules of completely modular construc-
tion enabling high-level number attainability, the MMC converter topology is
2.5Summary 47
considered as a highly feasible candidate for medium-voltage applications. The

MMC converter requires multiple isolated and balanced DC sources. The medium-
frequency magnetic link may be the best option to generate multiple isolated and
balanced DC sources for the MMC inverter from a single power source. Compared
with the power frequency transformers, the medium-frequency links have much
smaller and lighter magnetic cores and windings and thus much lower costs.
References
1. Earth policy institute Climate, energy and transportation, world cumulative wind turbine
installations [online]. Available at: http://www.earth-policy.org. Accessed on 25 Mar 2013
2. Islam MR, Guo YG, Zhu JG (2012) 11-kV series-connected H-bridge multilevel con-
verter for direct grid connection of renewable energy systems. J Int Conf Elec Mach Syst
1(2):211219
Electron 29(8):41674182
5. Islam MR, Guo YG, Lin ZW, Zhu JG (2014) An amorphous alloy core medium fre-
quency magnetic-link for medium voltage photovoltaic inverters. J Appl Phys
115(17):17E710-117E710-3
6. Sharma VK, Colangelo A, Spagna G (1995) Photovoltaic technology: basic concepts, siz-
ing of a standalone photovoltaic system for domestic applications and preliminary economic
analysis. Energy Convers Manage 36(3):161174
7. Santos JL, Antunes F, Chehab A, Cruz C (2006) A maximum power point tracker for PV sys-
tems using a high performance boost converter. Sol Energy 80(7):772778
8. Hua C, Shen C (1998) Study of maximum power tracking techniques and control of DC/DC
converters for photovoltaic power system. In: 29th annual IEEE power electronics specialists
conference, Fukuoka, Japan, 1722 May 1998, pp 8693
9. Jaboori MG, Saied MM, Hanafy AAR (1991) A contribution to the simulation and design
optimization of photovoltaic systems. IEEE Trans Energy Conversion 6(3):401406
10. Bose BK, Szczesny PM, Steigerwald RL (1985) Microcomputer control of a residential pho-
tovoltaic power conditioning system. IEEE Trans Ind Appl 21(5):11821191
11. Islam MR, Islam MR, Beg MRA (2008) Renewable energy resources and technologies prac-
tice in Bangladesh. Renew Sustain Energy Rev 12(2):299343
12. Islam MR, Guo YG, Zhu JG, Jin J (2011) Design and fabrication of a microcontroller
based maximum power point tracker for renewable energy systems. J Appl Supercond
Electromagnet 2(1):1723
13. Islam MR, Guo YG, Zhu JG, Rabbani MG (2010) Simulation of PV array characteristics and
fabrication of microcontroller based MPPT. In: 6th international conference on electrical and
computer engineering (ICECE), Dhaka, Bangladesh, 1820 Dec, pp 155158
14. Sugimoto H, Dong H (1997) A new scheme for maximum photovoltaic power tracking con-
trols. In: Power conversion conference, Nagaoka, Japan, 36 Aug 1997, pp 691696
15. Ogura K, Nishida T, Hiraki E, Nakaoka M, Nagai S (2004) Time-sharing boost chopper cas-
caded dual mode single-phase sinewave inverter for solar photovoltaic power generation sys-
tem. In: IEEE 35th annual power electronics specialists conference, Aachen, Germany, 2025
June 2004, pp 47634767
16. Young-Ho K, Jun-Gu K, Young-Hyok J, Chung-Yuen W, Yong-Chae J (2010) Photovoltaic
parallel resonant DC-link soft switching inverter using hysteresis current control. In:
Twenty-fifth annual IEEE applied power electronics conference and exposition, Palm
Springs, Canada, 2125 Feb 2010, pp 22752280
17. Nema S, Nema RK, Agnihotri G (2011) Inverter topologies and control structure in photovol-
taic applications: a review. J Renew Sustain Energy 3(1):012701-1012701-23
18. Patrao I, Figueres E, Espin FG, Garcera G (2011) Transformerless topologies for grid-con-
nected single-phase photovoltaic inverters. Renew Sustain Energy Rev 15(7):34233431
19. Rahman MF, Zhong L (1997) A new, transformerless, photovoltaic array to utility grid
interconnection. In: 1997 international conference on power electronics and drive systems,
Singapore, 2629 May 1997, pp 139143
20. SMA Solar Technology AG. Sunny Boy 5000TL multi-string inverter [online]. Available at:
http://www.sma.de. Accessed on 25 Feb 2013
21. Agheb E, Bahmani MA, Hoidalen HK, Thiringer T (2012) Core loss behavior in high
frequency high power transformers-II: arbitrary excitation. J Renew Sustain Energy
4(3):033113-1033113-11
22. Lu J, Stegen S, Butler D (2010) High frequency and high power density transformers for DC/
DC converter used in solar PV system. In: 2010 2nd IEEE international symposium on power
electronics for distributed generation systems, Hefei, China, 1618 June 2010, pp 481484
23. Lu J, Butler D (2011) High frequency transformers for DC/DC converter used in solar PV
system. J Energy Power Eng 5:536541
24. Lu J, Dawson F (2011) Characteristics for high frequency planar transformer with a novel
comb-shaped shield. IEEE Trans Magn 47(10):44934496
25. Panchal C, Lu J (2011) High frequency planar transformer for universal contact-less battery
charging platform. IEEE Trans Magn 47(10):27642767
26. Kjaer SB, Pedersen JK, Blaabjerg F (2005) A review of single-phase grid-connected inverters
for photovoltaic modules. IEEE Trans Ind Appl 41(5):12921306
27. Lu DDC, Agelidis VG (2009) Photovoltaic-battery-powered DC bus system for common
portable electronic devices. IEEE Trans Power Electron 24(3):849855
28. Liang TJ, Kuo YC, Chen JF (2001) Single-stage photovoltaic energy conversion system. IEE
Proc Electric Power Appl 148(4):339344
29. Chen Y, Smedley KM (2004) A cost-effective single-stage inverter with maximum power
point tracking. IEEE Trans Power Electron 19(5):12891294
30. Gonzalez R, Gubia E, Lopez J, Marroyo L (2008) Transformerless single-phase multilevel-
based photovoltaic inverter. IEEE Trans Ind Electron 55(7):26942702
31. Cavalcanti MC, Farias AM, Olivetra KC, Neves FAS, Afonso J (2012) Eliminating leakage
currents in neutral point clamped inverters for photovoltaic systems. IEEE Trans Ind Electron
59(1):435443
32. Junior LG, Brito MAGD, Sampaio LP, Canesin CA (2011) Single stage converters for low
power stand-alone and grid-connected PV systems. In: 2011 IEEE international symposium
on industrial electronics, Gdansk, Poland, 2730 June 2011, pp 11121117
33. Prasad BS, Jain S, Agarwal V (2008) Universal single-stage grid-connected inverter. IEEE
Trans Energy Conversion 23(1):128137
34. Wang C (2004) A novel single-stage full-bridge buck-boost inverter. IEEE Trans Power
Electron 19(1):150159
35. Patel H, Agarwal V (2009) A single-stage single-phase transformer-less doubly grounded
grid-connected PV interface. IEEE Trans Energy Conversion 24(1):93101
36. Meneses D, Blaabjerg F, Garca O, Cobos JA (2013) Review and comparison of step-up
transformerless topologies for photovoltaic AC-module application. IEEE Trans Power
Electron 28(6):26492663
37. ASEA Brown Boveri (ABB). Solar photovoltaic string inverter [online]. Available at: http://w
ww.abb.com. Accessed on 13 May 2013
38. SIEMENS. SINVERT PVM operating instruction 07/2010 [online]. Available at: http://www.
siemens.com/. Accessed on 13 April 2014
39. Du Y, Lu DDC (2010) Analysis of a battery-integrated boost converter for module-based
series connected photovoltaic system. In: 2010 international power electronics conference,
Sapporo, Japan, 2124 June 2010, pp 694698
References 49
40. Du Y, Lu DDC (2011) Battery-integrated boost converter utilizing distributed MPPT configu-
ration for photovoltaic systems. Sol Energy 85(9):19922002
41. ASEA Brown Boveri (ABB). Solar photovoltaic central inverter [online]. Available at: http://
www.abb.com. Accessed on 10 Mar 2014)
43. SIEMENS. SINVERT PVS solar photovoltaic inverter [online]. Available at: http://www.
siemens.com/. Accessed on 13 Feb 2014)
44. ASEA Brown Boveri (ABB). GEAFOL transformer [online]. Available at: http://www.abb.
com. Accessed on 02 July 2013)
45. Islam MR, Guo YG, Zhu JG (2011) Performance and cost comparison of NPC, FC and
SCHB multilevel converter topologies for high-voltage applications. In: 2011 international
conference on electrical machines and systems, Beijing, China, 2023 Aug 2011, pp 16
46. Islam MR, Guo YG, Zhu JG, Dorrell (2011) Design and comparison of 11kV multilevel
voltage source converters for local grid based renewable energy systems. In: Proceedings of
the 37th annual conference on IEEE industrial electronics society, Melbourne, Australia,710
Nov 2011, pp 35963601
47. Islam MR, Guo YG, Zhu JG (2011) H-bridge multilevel voltage source converter for direct
grid connection of renewable energy systems. In: 2011 IEEE PES innovative smart grid tech-
nologies Asia (ISGT), Perth, Australia, 1316 Nov 2011, pp 17
Phys 113(17):17A324-117A324-3
49. Sun D, Ge B, Peng FZ, Haitham AR, Bi D, Liu Y (2012) A new grid-connected PV system
based on cascaded H-bridge quasi-Z source inverter. In: 2012 IEEE international symposium
on industrial electronics, Hangzhou, China, 2831 May 2012, pp 951956
50. Choi H, Zhao W, Ciobotaru M, Agelidis VG (2012) Large-scale PV system based on the mul-
tiphase isolated DC/DC converter. In: 3rd IEEE international symposium on power electron-
ics for distributed generation systems, Aalborg, Denmark, 2528 June 2012, pp 801807
51. Zhao W, Choi H, Konstantinou G, Ciobotaru M, Agelidis VG (2012) Cascaded H-bridge
multilevel converter for large-scale PV grid-integration with isolated DC-DC stage. In: 3rd
IEEE international symposium on power electronics for distributed generation systems,
Aalborg, Denmark, 2528 June 2012, pp 849856
52. Kouro S, Fuentes C, Perez M, Rodriguez J (2012) Single DC-link cascaded H-bridge multi-
level multistring photovoltaic energy conversion system with inherent balanced operation. In:
38th annual conference on IEEE industrial electronics society, Montreal, QC, Canada, 2528
Oct 2012, pp 49985005
53. Rivera S, Wu B, Kouro S, Wang H, Zhang D (2012) Cascaded H-bridge multilevel converter
topology and three-phase balance control for large scale photovoltaic systems. In: 3rd IEEE
international symposium on power electronics for distributed generation systems, Aalborg,
Denmark, 2528 June 2012, pp 690697
54. Islam MR, Lei G, Guo YG, Zhu JG (2014) Optimal design of high-frequency magnetic-links
for power converters used in grid connected renewable energy systems. IEEE Trans Magn.
doi:10.1109/TMAG.2014.2329939
Chapter 3
Power Converter Topologies
for Grid-Integrated Medium-Voltage
Applications
AbstractAlthough several converter topologies have been used in low-voltage

applications, most of the topologies are not suitable in medium-voltage applica-
tions. The selection of converter topology in medium-voltage applications is really
a critical problem and highly affects the converter performance and cost. The main
aim of this chapter was to find out a suitable converter topology, which can inter-
connect the renewable generation units directly to the medium-voltage grid with
mature semiconductor devices. Different multilevel converter topologies, such as
neutral point clamped (NPC), flying capacitor (FC), and modular multilevel cas-
caded (MMC), have been considered and compared for the design of an 11kV
converter system. The comparison is made in terms of the number of semiconduc-
tors, semiconductor cost and availability, total harmonic distortions (THDs), filter
size, and control complexity of the converters. The performance is analyzed and
compared in the MATLAB/Simulink environment. To ensure quality performance,
a level-shifted carrier-based switching scheme is used for the NPC topologies and
a phase-shifted carrier-based switching scheme is used for the FC and MMC con-
verter topologies with a carrier frequency of 12kHz and modulation index of
0.80.9.
Keywords Power converters Neutral point clamped (NPC) Modular multilevel

cascaded (MMC) Flying capacitor (FC) Topology selection Medium-voltage
Two-level Switching and control MATLAB/Simulink
3.1Introduction
In terms of technology development, a continuous race to develop higher volt-

age and higher current power semiconductors for utilization in high-power sys-
tems still goes on. Many recent generations of devices are suitable for low-voltage
applications while medium-voltage semiconductors are still under development.
The highest voltage rating of the commercially available insulated gate bipo-
lar transistor (IGBT) is 6.5kV, which is suitable for 2.88kV or lower-voltage

52 3 Power Converter Topologies for Grid
converter system with traditional 2-level converter topology. In order to intercon-

nect the 2-level converter-based renewable generation system to medium-voltage
grid, a power frequency step-up transformer is commonly used [1, 2]. Due to the
high percentage of harmonic distortion in the output-generated power, the 2-level
converter-based system requires a large-size line filter. The use of high-voltage-
rated devices may increase the switching and conduction losses of the 2-level con-
verter. Moreover, the price of power semiconductor devices increases rapidly with
their voltage as well as power ratings. Therefore, the traditional 2-level converter
topologies are not suitable for medium- and high-voltage networks.
The cascade connection of lower-rated semiconductors may be a cost-effective
solution for medium-voltage applications. The multilevel converter topologies rep-
resent the simplest and most efficient way to connect the semiconductor devices
in cascade form. The cascade connection gives the ability to use low-voltage-rated
matured and low-cost semiconductor devices. Medium- and high-voltage high-
power converters can be designed by using available semiconductor devices with the
multilevel converter topology. In comparison with conventional 2-level converters,
multilevel converters present lower switching losses, lower-voltage stress on switch-
ing devices, and better harmonic performance. These remarkable features allow the
direct connection of renewable energy systems to the grid without using large, heavy,
and costly power transformers. In addition to this, they serve to minimize the input
and output filter requirements. Although several multilevel converter topologies have
been used in low-voltage applications, most of them are not suitable in medium-volt-
age networks as the number of auxiliary components scales quadratically with the
number of levels. Because of their special features (e.g., the number of components
scales linearly with the number of levels, and individual modules are identical and
completely modular in construction hence enabling high-level number attainability),
the modular multilevel cascaded (MMC) converter topology can be considered as a
possible candidate for medium-voltage applications [3]. Several power semiconduc-
tor vendors such as Semikron, ABB, IXYS, and Mitsubishi Electric produce switch-
ing devices in modular forms, and all the devices of an H-bridge inverter in a single
pack with a gate drive circuit, which reduces not only the complexity and overall size
of the converter, but also the converter developmental time and semiconductor costs.
The modularity also provides a means to modify or rectify the converter circuit eas-
ily without changing other sections of the converter.
The main aim of this chapter was to find out a suitable converter topology,
which can interconnect the renewable generation units directly (without using the
traditional step-up transformer and line filter) to the medium-voltage grid with
low-voltage low-cost matured semiconductor devices. Besides the multilevel con-
verter topologies, the traditional 2-level converter topologies have also been con-
sidered with their switching schemes, power handling capacity, and performance.
Different multilevel converter topologies, such as neutral point clamped (NPC),
flying capacitor (FC), and MMC, have been considered for the design of an 11kV
converter system. The comparison is made in terms of the number of semiconduc-
tors, semiconductor cost and availability, total harmonic distortions (THDs), filter
size, and control complexity of the converters. The performance is analyzed and
3.1Introduction 53
compared in the MATLAB/Simulink environment. To ensure quality performance,

a level-shifted carrier-based switching scheme is used for the NPC topologies and
a phase-shifted carrier-based switching scheme is used for the FC and MMC con-
verter topologies with a carrier frequency of 12kHz and modulation index of
0.80.9. The modeling of circuit configurations, modulation schemes, and switch-
ing techniques in MATLAB/Simulink are presented in detail.
3.2Two-Level Converters
3.2.1Single-Phase Converter
3.2.1.1Half-Bridge Converter
Figure3.1 illustrates the circuit diagram of half-bridge converter consisting of two
switching devices, e.g., insulated gate bipolar device (IGBT) and a three-wire DC
power supply. The switching devices can be controlled in a smart way for progres-
sively wider pulses at the positive peak and narrower pulses at the negative peak,
which may help to filter out high-frequency switching harmonics. The unipolar
pulse width modulation (PWM) scheme with a sinusoidal reference and a triangu-
lar carrier can be used to operate switching devices. The comparator of the modu-
lation scheme compares the reference signal with the carrier signal and generates
Boolean output of 0 or 1. This generated gate pulse may operate the top IGBT of
the half-bridge converter. The bottom IGBT can be operated by the inverted gate
pulse of the Boolean output. Figure3.2 shows the modulation scheme of half-
bridge converter. The freewheeling diodes permit current flow even if all switches
are open and lagging currents flow in case of inductive loads.
The two switching devices of a half-bridge converter should not be on at the
same time, to avoid the direct short circuit between the terminals of DC power
supply. The peak load voltage, VL, of half-bridge converter can be expressed as
Vs
VL = mi (3.1)
2
where mi is the modulation index and Vs the DC power supply voltage. Table3.1
summarizes the operation of half-bridge converter circuit. The output voltage
waveform is shown in Fig.3.3.
Fig.3.1Circuit diagram of
half-bridge voltage source Vs + Top
converter _C1 IGBT
2
+
0 Load
Vs
_ + VL _
Vs +
2 _C2 Bottom
IGBT
0.5
Signals
0
Gate pules
Reference
-0.5 Carrier
-1
0 2 4 6 8 10 12
Time (ms)
Fig.3.2Modulation scheme of half-bridge PWM converter
Table3.1Switching scheme of half-bridge converter

Top IGBT Bottom IGBT Source voltage Load voltage
On Off Vs VL = +mi V2s
Off On VL = mi V2s
Off Off VL = 0
On On Must be
avoided
200
100 Reference
Voltage (V)
0
Inverter
output
-100
-200
0 2 4 6 8 10 12 14 16
Time (ms)
Fig.3.3Output voltage waveform of half-bridge PWM converter
If T is the time period and Tp the time when the top switch is in the on state
during a cycle, the rms output voltage, VL(rms), of the half-bridge PWM converter
can be deduced as
3.2 Two-Level Converters 55

1
T
VL(rms) = VL2 dt
T 0

1 Tp V 2 T
V 2
2 s 2 s
= mi dt + mi dt
T 0 4 Tp 4

(3.2)
2 T T
m V2 p
i s
= dt + dt
4T 0 Tp

mi Vs 1 Tp 1
= [t] + [t]TTp
2 T 0 T

mi Vs 1
= T p 0 + T Tp
2 T
mi Vs
=
2 (3.3)
Figure 3.4 plots the frequency spectrum of output voltage, which contains about
146% THD with a carrier frequency of 1kHz. Although the half-bridge converter
requires less number of switching devices, the requirements of double voltage-rated
DC power supply and large-size line filter may limit the use of this converter. The
peak inverse voltage (PIV) of the switching devices is equal to the total DC power
supply voltage, Vs. Maximum 1.44kV converter can be designed with the commer-
cially available IGBTs. Table3.2 summarizes the possible maximum output volt-
ages of half-bridge converters with commercially available switching devices.
3.2.1.2Full-Bridge Converter
Figure 3.5 illustrates the circuit diagram of full-bridge converter consisting of

four switching devices with two on each half-bridge arm. Due to the similarity of
0.5
Normalized amplitude
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10
Frequency (kHz)
Fig.3.4Frequency spectrum of output voltage of half-bridge PWM converter

Table3.2Maximum rms output voltage of half-bridge converters with modulation index at 0.8
Voltage rating of IGBT Vcom@100FIT (V) Voltage rating of DC Converter rms output
(V) supply (V) voltage (V)
6,500 3,600 3,600 1,440
4,500 2,250 2,250 900
3,300 1,800 1,800 720
2,500 1,200 1,200 480
1,700 900 900 360
Vs + T1 T3
_C1
2
+
Load
Vs
_
Vs + + VL _
2 _C2 T2
T4
Fig.3.5Circuit diagram of full-bridge PWM converter
Table3.3Switching scheme of full-bridge converter

T1 T2 T3 T4 Source voltage Load voltage
On On Off Off Vs VL = +mi Vs
Off Off Off Off 0
Off Off On On VL = mi Vs
On On On On Must be
avoided
structure with capital letter H, it is also called H-bridge converter. The switching
device pairs T1T2 and T3T4 can be switched on and off alternately with each
pair providing the opposite polarity of voltage across the load. The two switch-
ing devices of the same arm should not be on at the same time, to avoid direct
short circuit between the terminals of the DC power supply. The switching devices
can be controlled in a smart way for progressively wider pulses at the center and
narrower pulses at the edges of half-cycle output voltage waveform, which may
help to filter out high-frequency switching harmonics. The modulation scheme as
depicted in Fig.3.2 can also be used to switch the switching pairs. The Boolean
output by comparison of carrier signal and positive half-cycle of the reference sig-
nal may switch T1 and T2 simultaneously. The Boolean output by comparison of
carrier signal and negative half-cycle of the reference signal may switch T3 and T4
simultaneously. Table3.3 shows the switching scheme of a full-wave converter.
The output voltage waveform of full-bridge converter has three voltage levels,
e.g., 200, 0, and 200V in the voltage waveform as shown in Fig.3.6. Due to the
higher number of voltage levels, the THD is much lower than that of half-bridge
200
100 Reference
Voltage (V)
0
Inverter
-100 output
-200
10 15 20 25
Time (ms)
Fig.3.6Output voltage waveform of full-bridge PWM converter
0.5
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10
Frequency (kHz)
Fig.3.7Frequency spectrum of output voltage of full-bridge PWM converter
converter. At 1kHz PWM switching frequency, the full-bridge converter may pro-
vide an output voltage with less than 80% THD. Figure3.7 shows the frequency
spectrum of output voltage of the full-bridge converter.
If the output PWM voltage waveform as shown in Fig.3.6 is approximated by
square wave as shown in Fig.3.8, the rms value of the output voltage can be calcu-
lated as

1 T 2

VL(rms) = V dt
T 0 L

1 T2
T
2 2
= mi Vs2 dt + mi Vs2 dt (3.4)
T 0 T
2

T
2 2
= mi2 Vs2 dt
T 0
= mi Vs (3.5)
200
100 Approximated
Voltage (V)
T/2
0
T/2
-100 PWM output
-200
10 15 20 25
Time (ms)
Fig.3.8Output voltage of full-bridge PWM converter with square wave approximation
Table3.4Maximum rms output voltage of full-bridge converters with modulation index at 0.8
Voltage rating of IGBT Vcom@100FIT Voltage rating of DC Converter rms output
(V) (V) supply (V) voltage (V)
6,500 3,600 3,600 2,880
4,500 2,250 2,250 1,800
3,300 1,800 1,800 1,440
2,500 1,200 1,200 960
1,700 900 900 720
Compared with the half-bridge converter, almost double-rated voltage converter

can be designed by using the full-bridge converter topology with same voltage-
rated IGBTs. Table3.4 summarizes the possible maximum converter ratings that
can be designed with the available IGBTs. Although the full-bridge converter
requires two more switching devices, which may increase converter power loss
and cost, the smaller THD can reduce the size of line filter significantly, which
may reduce the converter power loss and cost and make the converter compact and
lightweight.
3.2.2Three-Phase Converter
Three-phase converters are more common than single-phase half- or full-bridge

converters. Figure3.9 illustrates the power circuit of a 3-phase bridge converter,
which consists of six switching devices with two for each phase. Large capacitors
are usually connected at the input terminal to make the supply voltage constant
and also minimize the harmonics fed back to the source especially when the DC
Vs + T1 T3 T5
_C1
2
+ Va
a Vb
Vs b
_ c Vc
Vs +
2 _C2 T4 T6
T2
Fig.3.9Circuit diagram of 3-phase 2-level converter
0.5 Ref. A Ref. B Ref. C

Signals
Carrier
-0.5
-1
0 2 4 6 8 10 12 14 16
Time (ms)
Fig.3.10Modulation scheme of 3-phase 2-level converter
supply is through a rectifier circuit. The triangular carrier signal is compared with
the sinusoidal modulating signal. When the modulating signal is greater than the
carrier, pulse 1 is high (1) and pulse 2 is low (0). Single carrier and three modulat-
ing or reference signals (120 phase shift between each other) may generate six
gate pulses to drive the converter. Figure3.10 shows the PWM switching scheme
of the 3-phase bridge converter. The Boolean output through comparison of Ref A
with carrier signal may switch T1, and its inverted output may switch T2. The gate
pulse generated from Ref B and carrier may switch T3, and its inverted gate pulse
may switch T4. Similarly, Ref C and carrier may generate gate pulses for T5 and
T6. The line-to-line voltage waveform of 3-phase bridge converter is almost the
same as that of the full-bridge converter. Figure3.11 shows the line-to-line out-
put voltage waveform of 3-phase bridge converter. Similar to the single-phase full-
bridge converter, the line-to-line voltage of the 3-phase bridge converter contains
about 80% THD. Figure3.12 shows the frequency spectrum of the output line-
to-line voltage. The maximum output voltage generation capacity is also similar
to the single-phase full-bridge converter, e.g., maximum 2.88kV (rms) converter
200
100
Voltage (V) Reference
Output voltage
-100
-200
0 2 4 6 8 10 12 14 16
Time (ms)
Fig.3.11Output line voltage of 3-phase 2-level converter
0.5
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10
Frequency (kHz)
Fig.3.12Frequency spectrum of output line voltage of 3-phase 2-level converter
with 6.5kV IGBTs. Although the 3-phase bridge converter requires 33% addi-
tional switching devices, it can handle about 73% more power than that of the sin-
gle-phase full-bridge converter. Therefore, the 3-phase converter could be a smart
choice for medium- and high-power applications. If Vll is the converter line-to-
line voltage and Il the line current, the power handling capacity of the 3-phase
2-level converter can be calculated as

PT = 3Vll Il . (3.6)
3.2.3Space Vector Pulse Width Modulation (SV-PWM)
The main aim of any modulation technique is to obtain desired output having a
maximum fundamental component with minimum harmonics and less switch-
ing losses. The space vector pulse width modulation (SV-PWM) method is an
advanced PWM method, and it is possibly the best among all the PWM techniques
for high-power applications. Compared to sinusoidal pulse width modulation
(SPWM), SV-PWM offers 15% higher DC-voltage utilization and 33% fewer
communications per cycle. Moreover, it can obtain a better harmonic performance
and less switching loss.
In each cycle of SV-PWM, the desired output voltage is approximated by a
time average of three voltage vectors, in which two are nonzero voltage vectors
adjacent to the reference and the third is a zero vector. The strategy of sequencing
these three switching vectors within each cycle will affect the current ripple, the
switching losses and the spectrum of the output voltages and currents. Moreover,
the switching sequence is not unique. Since the SV-PWM offers superior perfor-
mance with respect to other modulation techniques, it is important to establish the
sequencing strategy which is well suited for the systems.
In SV-PWM, the converter is controlled through the concept of converter states.
Each converter state corresponds to a certain combination of switches. There are
two switches in each leg of a 2-level 3-phase converter. The switches in one leg
cannot be both on or off at the same time (the former leads to a short circuit of
the associated phase, and the latter results in an open circuit). There are three con-
verter legs in a 3-phase system, so the total number of converter states is eight.
Figure 3.13 shows the eight possible combinations of on and off patterns
of 3-phase 2-level converter. The on and off states of the lower switches are the
inverted states of the upper ones. The phase voltages corresponding to the eight
combinations of switching patterns can be calculated. This transformation results
in six nonzero voltage vectors and two zero vectors. The nonzero vectors form the
axes of a hexagon containing six sectors (V1 V6) as shown in Fig.3.13. The
angle between any adjacent two nonzero voltage vectors is 60. The zero vec-
tors are at the origin and give a zero voltage vector to the output. The envelope of
the hexagon formed by the nonzero vectors is the locus of the maximum output
Fig.3.13Space vector q axis

diagram of 3-phase 2-level
converter V3 V2
(010) (110)
Sector 2
3
Se
r
cto
cto
Se
(000)
1
V0 V7
(111) V1
V4
(011) (100) d axis
r4
r6
cto
cto
Se
Se
Sector 5
V5
(101)
V6
(001)
Table3.5Eight switching Voltage vector Switching vector Line voltage

states of 3-phase 2-level
T1 T3 T5 Vab Vbc Vca
converter
V0 0 0 0 0 0 0
V1 1 0 0 Vs 0 Vs
V2 1 1 0 0 Vs Vs
V3 0 1 0 Vs Vs 0
V4 0 1 1 Vs 0 Vs
V5 0 0 1 0 Vs Vs
V6 1 0 1 Vs Vs 0
V7 1 1 1 0 0 0
voltage. If versus is the DC supply voltage of 3-phase 2-level converter, the length
of six active space vectors can be calculated as
2
Vs (3.7)
3
When the upper switch on a phase leg is on, i.e., T1, T3, or T5 is 1, the correspond-
ing lower switch should be off, i.e., T2, T4, or T6 is 0. In the 2-level SV-PWM,
V0 and V7 generate the same line voltage (which is zero) and hence are redun-
dant states. The eight possible switching states and corresponding phase as well as
the line voltages are shown in Table3.5. One can choose one or both of the zero
voltage vectors, as suited, to reduce the number of switching or manipulate har-
monics. These two zero vectors are the only redundant space vectors for a 2-level
SV-PWM. As the number of levels of the SV-PWM implementation increases, the
number of redundant space vectors increases, which can be exploited to achieve
the desired performance characteristics.
Usually, when creating space vectors, the three time varying quantities are sinu-
soids of the same amplitude and frequency that have 120 phase shifts between
each other. When this is the case, the space vector at any given time maintains
its magnitude. As time increases, the angle of the space vector increases, causing
the vector to rotate with frequency equal to the frequency of the sinusoids. Each
of the converter states can be mapped into the dq plane by applying Parks trans-
formation to the phase voltages as given. The vectorial representation of a given
3-phase system as shown in Fig.3.14 can be carried out by the relation (3.9) to
(3.15) [4]. Figure3.15 describes the technique to calculate the d-axis component
corresponding to 3-phase voltages. The q-axis component can be calculated from
(3.12), which can be deduced from Fig.3.16.
Vd = Van Vbn cos 60 Vcn cos 60 (3.8)
1 1
= Van Vbn Vcn (3.9)
2 2
Vq = 0 + Vbn cos 30 Vcn cos 30 (3.10)

3 3
= 0Van + Vbn Vcn (3.11)
2 2

1 21 21 Van
Vd Vbn (3.12)
=
Vq 0 23 23 Vcn

|Vref | = Vd2 + Vq2 (3.13)

Vq
= tan 1 (3.14)
Vd
Fig.3.14The dq plane q axis
b Vq Vref

Vd a d axis
Fig.3.15d-axis component Vbn

calculation
600
Van Vd
600
Vcn
Fig.3.16q-axis component
calculation Vq
Vbn
300
900
Van
0
30
Vcn
Fig.3.17Algorithm to
Read
select the sector
Yes
0 0 < 60 0 Sector1
No
Yes
600 < 1200 Sector2
No
Yes
1200 < 1800 Sector3
No
Yes
1800 < 2400 Sector4
No
Yes
240 0 < 300 0 Sector5
No
Sector6
Equation (3.14) is used to calculate , and the reference vector is located accord-
ing to an algorithm as shown in Fig.3.17.
The only difference among PWM schemes that use adjacent vectors is the
selection of the zero vector(s) and the sequence in which the vectors are applied
within the switching cycle. The degrees of freedom we have in the selection of a
switching pattern are as follows:
selection of the zero vector whether or not we would use V7 (111) or V0 (000) or
both sequencing of the vectors, and
splitting of the duty cycles of the vectors without additional commutations.
Considering the degrees of freedom, there are four most commonly used switching
patterns:
seven switching states,
three switching states (with only one zero vector),
three switching states (with both zero vectors), and
four switching states.
3.2.3.1Switching Pattern of Seven Switching States
The number of commutations and switching states in one sampling period is six or
seven, respectively, as shown in Fig.3.18. Both zero vectors (V7 and V0) are used
Vo V1 V2 V7 V2 V1 Vo
To /4 T1 /2 T2 /2 To /2 T2 /2 T1 /2 To /4
Ts
Fig.3.18Seven-state switching pattern at sector 1
200
100
Voltage (V)
-100
-200
5 10 15 20
Time (ms)
Fig.3.19Output line-to-line voltage with seven-state switching pattern
in each sector, where V7 is placed in the middle and V0 at both side of the pattern.
For example, if the reference is located in the first sector, the possible switching
sequence could be V0V1V2V7V2V1V0, etc. The switching scheme has been
implemented in MATLAB/Simulink environment. The output line-to-line voltage
of 3-phase 2-level converter with this switching scheme is shown in Fig.3.19. Due
to the symmetry in the switching waveform this scheme shows the lowest THD.
Compared with SPWM, seven switching states switching scheme gives about
45% less harmonic distortion. Figure3.20 shows the frequency spectrum of line-
to-line voltage, where about 55% THD was measured. Figure3.21 plots the line
current of 3-phase 2-level converter with seven switching states switching scheme-
based SV-PWM.
Normalized amplitude 0.2
0.1
0
0 2 4 6 8 10
Frequency (kHz)
Fig.3.20Frequency spectrum of output voltage with seven-state switching pattern
20
10
Current (A)
-10
-20
5 10 15 20 25 30
Time (ms)
Fig.3.21Line current with seven-state switching pattern
3.2.3.2Three Switching States Switching Pattern (Only One Zero

Vector)
Since the selection of the nonzero vectors is based on the desired output voltage
vector and the phase and magnitude of the current are determined by the load, it
is not always possible to avoid switching the phase carrying the highest current.
In such a case, the phase carrying the second highest current is not switched on
so that the switching losses are still reduced. In this scheme, the number of com-
mutations in one sampling period is four and there are only three switching states
in Ts. With this strategy, the switching sequence is repeated in each cycle as long
as the reference vector is within the same sector. For the case when a reference is
in sector 1, the switching sequence is given in Fig.3.22. If V7 is selected as the
zero vector, each switching cycle will have two commutations in the inverter leg b
and two in leg c (a total of 4), while leg a will have the upper switch on during the
entire cycle. If the zero vector is V0, the number of commutations (i.e., transitions)
V1 V2 V7 V1 V2 V7
T1 T2 To T1 T2 To
Ts Ts
Fig.3.22Three-state switching pattern (only one zero vector)
200
100
Voltage (V)
-100
-200
5 10 15 20 25
Time (s)
Fig.3.23Output line-to-line voltage with three-state switching pattern (only one zero vector)
per cycle will be the same, but the inverter leg that will cease to commutate in
sector 1 will be leg c. This redundancy of zero vectors allows for a reduction of
commutation losses. Selection of the zero vector should be done so as to avoid
commutation in the inverter leg carrying the largest current. The output line-to-
line voltage of 3-phase 2-level converter with three switching states switching
pattern (only one zero vector) is shown in Fig.3.23. Figure3.24 shows the fre-
quency spectrum of output line-to-line voltage. Compared with the seven-state
switching scheme, three switching states switching pattern (only one zero vector)
gives significantly higher harmonics, i.e., about 20% more. Still it is about 15%
lower than that of the SPWM scheme. Figure3.25 plots the line current, which is
0.1
0
0 2 4 6 8 10
Frequency (kHz)
Fig.3.24Frequency spectrum of output line-to-line voltage with three-state switching pattern

(only one zero vector)
20
10
Current (A)
-10
-20
20 25 30 35 40 45
Time (ms)
Fig.3.25Line current with three-state switching pattern (only one zero vector)
more distorted compared with the line current in Fig.3.21 with the same circuit
parameters.
3.2.3.3Three Switching States Switching Pattern (Both Zero Vectors

but Alternatively)
One can modify the switching sequence between the voltage vectors, so that each
transfer from one state to the other involves only one commutation. Figure3.26
shows the switching pattern of three-state switching (both zero vectors but alterna-
tively). This requires the use of both zero vectors (V7 and V0) in a given sector and
a reversal of the switching sequence every cycle. In this scheme, the zero vectors
V7 and V0 are used alternatively in adjacent cycles, so that the effective switching
frequency is halved. However, the sampling period is still Ts.
V1 V2 V7 V2 V1 Vo
T1 T2 To T2 T1 To
Ts Ts
Fig.3.26Switching pattern of three-state switching (both zero vectors alternatively)
200
100
Voltage (V)
-100
-200
5 10 15 20 25
Time (ms)
Fig.3.27Output line-to-line voltage of 3-phase 2-level converter with three-state switching

(both zero vectors alternatively)
The benefit is a reduction of commutations (to three). The switching losses for
this scheme are expected to be ideally 50% as compared to those of the seven-
state switching scheme, but the THD is significantly higher due to the existence of
the harmonics at half of the sampling frequency. About 68% THD was measured
on the line-to-line output voltage. In this scheme, the number of commutations in
one sampling period is three and there are only three switching states in Ts. The
output line-to-line voltage of 3-phase 2-level converter with three-state switch-
ing (both zero vectors alternatively) is shown in Fig.3.27. Although the harmonic
0.1
0
0 2 4 6 8 10
Frequency (kHz)
Fig.3.28Harmonic spectrum of line-to-line voltage with three-state switching (both zero vec-
tors alternatively)
20
10
Current (A)
-10
-20
20 25 30 35 40 45
Time (ms)
Fig.3.29Line current of 3-phase 2-level converter with three-state switching (both zero vectors
alternatively)
performance of this switching pattern is similar with the three-state switching pat-
tern (only one zero vector), the number of commutation is higher (3 times in a
period). Figure3.28 shows the harmonic spectrum of output line-to-line voltage
with three-state switching (both zero vectors alternatively). The line current wave-
form is also almost the same as in Fig.3.25. Figure3.29 shows the line current
waveform with three-state switching (both zero vectors alternatively).
3.2.3.4Four Switching States Switching Pattern
Both zero vectors, V7 and V0, are used to reduce the number of commutations per
switching cycle. The switching sequence is reversed after each switching cycle
within a sector. For example, if the reference is in the first sector, the possible
switching sequence can be V0V1V2V7V7V2V1V0, as shown in Fig.3.30.
Vo V1 V2 V7 V7 V2 V1 Vo
To/2 T1/2 T2/2 To/2 To/2 T2/2 T1/2 To/2
Ts Ts
Fig.3.30Four-state switching pattern
200
100
Voltage (V)
-100
-200
5 10 15 20
Time (s)
Fig.3.31Output line-to-line voltage of 3-phase 2-level converter with four-state switching

pattern
The main advantage of this approach is that it requires only three commutations
per cycle and gives symmetrical pulses.
The benefit is a reduction of commutations (to three). The switching losses for
this scheme are expected to be ideally 50% as compared to those of the seven-
state switching scheme, but the THD is significantly higher due to the existence
of the harmonics at half of the sampling frequency. In this scheme, the number
of commutations in one sampling period is three and there are only four switch-
ing states in Ts. The output line-to-line voltage of 3-phase 2-level converter with
four-state switching scheme is shown in Fig.3.31. The harmonic performance of
this switching scheme is slightly better than those of three-state switching pattern
(only one zero vector) and three-state switching (both zero vectors alternatively).
0.1
0
0 2 4 6 8 10
Frequency (Hz)
Fig.3.32Harmonic spectrum of line-to-line voltage with four-state switching pattern
20
10
Current (A)
-10
-20
20 25 30 35 40 45
Time (s)
Fig.3.33Line current of 3-phase 2-level converter with four-state switching pattern
About 65% THD was measured. Figure3.32 shows the frequency spectrum of
line-to-line voltage. The line current waveform is also slightly less distorted than
those in Figs.3.25 and 3.29. Figure3.33 shows the line current waveform with
four-state switching pattern.
3.2.3.5Evaluation of Switching Patterns
Table3.6 summarizes the performance of four SV-PWM switching patterns. It can

be clearly seen that a pattern with high THD has low losses and vice versa. The
switching losses are approximately proportional to the magnitude of the current
being switched, and hence, it would be advantageous to avoid switching the con-
verter leg carrying the highest instantaneous current. Compared with the seven-
state switching scheme (if 100% switching loss is considered with this scheme),
the three-state switching pattern (only one zero vector), three-state switching (both
zero vectors alternatively), and four-state switching pattern gives about 33, 50,
Table3.6Comparison of switching patterns

Switching patterns Seven states Three states with Four
V0 or V7 V0 and V7 states
Commutations in Ts 6 2 3 3
Switching loss (%) 100 33 50 50
THD (%) 55 70 68 65
and 50% switching losses, respectively. In this way, the three-state and four-state
switching patterns are the best choice for power converters.
3.3Multilevel Converters
The multilevel converters are power electronic circuits that can provide output
voltage with more than two levels and be operated in inverter or rectifier mode.
The 3-level converter is the lowest level number multilevel converter. Although the
number of levels of output phase voltage of 3-level converter is the same as that
of line-to-line voltage of 2-level full-bridge converter, the line-to-line voltage of
3-level converter consists of five-voltage levels. There are three basic topologies in
the multilevel converter: NPC, FC, and MMC. In 1975, the concept of multilevel
converter topology was proposed [5], and in the last few decades, several multi-
level converter topologies have been patented [6, 7]. In order to achieve high volt-
age using low-voltage switching devices, the multilevel converter topology uses
a number of switching devices with low-voltage DC sources. The proper control
of the switching devices superimposes these multiple DC sources in a staircase
(time shifted) form in order to achieve high voltage at the output. The renew-
able energy sources, such as wind turbine generators, solar PV arrays, and fuel
cells, can be used as the multiple DC-voltage sources. The NPC and FC convert-
ers require single DC supply, but the MMC converter requires multiple isolated
and balanced supplies. High-frequency magnetic link could be a possible option
to generate multiple DC supplies for MMC converter from single source. In mul-
tilevel converters, the rated voltage of the switching devices depends on the rating
of the DC-voltage sources to which they are connected. Therefore, the rated volt-
age of switching devices is much lower compared with a 2-level converter. The
power circuits of one phase of different 3-phase multilevel converters are shown
in Fig.3.34. According to different switching states, it is possible to achieve high-
voltage level on the output voltage by adding up the DC-voltage sources compared
with the 2-level inverters.
According to the output voltage waveform of two level, the quality of output
voltage depends only on the sampling frequency (on X-axis), so that a higher sam-
pling frequency means better voltage waveform, or closer to the sine wave refer-
ence. On the other hand, increasing the switching frequency would increase the
switching loss and cause electromagnetic compatibility (EMC) problems. The
(a) (b)
SA7 SA7
C C
SA5 SA5
DA1
CA4
SA3 SA3 (c)

DA2
H-bridge cell 1
CA2
SA1 SA1
SA1 SA3
C C
DA3
CA1
a CA5 a Vdc SA2 SA4 a
Vdc Vdc
SA2 SA2
DA4
C C
SA4 SA4
CA3 H-bridge cell 2
DA5 SA5 SA7
SA6 SA6
CA6
DA6 Vdc SA6 SA8 n

SA8 SA8
C n C n
Fig.3.34Five-level power converter circuit topologies [2], a NPC converter topology, b FC

converter topology, and c MMC converter topology
higher switching frequency means higher dv/dt, which would decrease the stabil-
ity and reliability of the switching devices. Moreover, the semiconductor devices
also have limited range of switching frequencies. The switching frequency of the
most commonly available power IGBTs is at 10kHz or lower. For example, the
recommended switching frequency of Semikron 1,200V 123 NTP, 126 Trench,
12E4HD, and 12V IGBTs are 9, 5, 9, and 12kHz, respectively. The 7-level con-
verter output voltage waveform is depicted in Fig.3.35. In the voltage waveform,
there is an additional parameter to control the quality of the output voltage wave-
form, i.e., the number of voltage levels. Both the X- and Y-axes can be controlled
to improve the converter performance effectively. The switching frequency can be
reduced to a low value with an increase to the number of voltage levels, making
the converter more stable and reliable and making it possible to use low-voltage-
rated matured and low-cost semiconductor devices.
Synthesizing a staircase output voltage, which is closer to the sinusoidal volt-
age reference compared with the 2-level converter output, allows for a reduction in
the harmonic content in voltage waveforms, which would lead to the size and cost
reduction of the line filter circuit of power conversion systems. Each level of the
staircase consists of a few PWM pulses, as shown in Fig.3.36.
Like the 2-level converter, the carrier signal frequency of the PWM genera-
tor can be adjusted to control the quality of the output waveform. However, the
number of levels of the staircase waveform mostly affects the quality of the out-
put voltage waveform. Figure3.37 plots the THDs of 11 and 33kV converters
with different numbers of converter levels. For an 11kV converter, 19-level or
3.3 Multilevel Converters 75
15
Line voltage Reference
10
Voltage (kV)
5
-5
-10
-15
20 30 40 50 60 70 80
Time (ms)
Fig.3.35Line-to-line output voltage waveform of 5-level converter
Level 5
200
Level 4
100
Voltage (V)
Level 3
0
PWM output
(before filter circuit)
Level 2
-100
-200
Level 1
0 5 10 15 20 25 30
Time (ms)
Fig.3.36Staircase output phase voltage waveform of 5-level converter
higher level converter may provide output power with less than 5% THD, which
inherently satisfies the IEEE and IEC standards. For a 33kV system, 21-level or
higher level converter may satisfy the 5% THD requirements without using line
filter. Although the number of semiconductor devices increases with the number
of converter levels, the high-level converter enables the use of low-voltage devices,
which are not only low cost but also matured forms of technology. Table3.7 lists
the market prices of some available power IGBTs.
Therefore, the semiconductor cost of multilevel converters is lower than that of
the equivalent 2-level converters operated at the same power rating. Figure3.38
plots the per-unit prices of switching devices. It can be seen from Fig.3.38 that
the 2.5kV or lower-rated IGBTs cost much lower than the high-voltage ones.
The highest voltage rating of the commercially available IGBT is 6.5kV, and this
is suitable for 2.88kV or lower-voltage converter systems with the traditional
Table3.7Market prices of some available power IGBTs [8]

Voltage (V) Current (A) Brand Price
(AUD)
600 400 Powerex 157.91
300 Powerex 118.28
1,200 3,600 Eupec 2,072.31
2,400 Eupec 1,582.39
1,400 Infineon 1,100.02
600 Powerex 414.51
400 Powerex 200.44
300 Powerex 207.28
200 Eupec 151.64
1,700 3,600 Eupec 2,624.48
2,400 Eupec 2,017.17
1,200 Eupec 1,106.11
600 Infineon 425.00
450 Eupec 339.60
300 Eupec 238.36
225 Eupec 195.21
150 Semikron 168.00
3,300 1,500 Eupec 3,007.82
1,200 Eupec 2,431.88
400 Eupec 1,369.31
6,500 600 Eupec 3,997.60
400 Eupec 3,078.18
200 Eupec 1,918.98
10
33 kV 11 kV
8 converter converter
THD (%)
6
5 % limit (IEEE 1547, IEC61727)
Unchanged
4
2
10 20 30 40 50
Number of levels
Fig.3.37THDs of 33 and 11kV multilevel converters [3]
2-level converter topology. Table3.8 summarizes the voltage ratings of com-

mercially available IGBTs, where Vcom@100FIT is the device commutation volt-
age for a device reliability of 100 failures in time (FIT) due to cosmic radiation.
The multilevel converter topology enables the possibility to develop medium- and
Table3.8Voltage handling capacity of available power IGBTs

Rated device voltage (V) Vcom@100FIT
(V)
6,500 3,600
4,500 2,250
3,300 1,800
2,500 1,200
1,700 900
1,200 800
900 600
600 300
Table3.9Voltage handling capacity of power converters

Device voltage rating (V) Number of levels Line voltage (kV,
rms)
6,500 2 2.88
5 10.18
11 25.45
19 45.82
21 50.91
3,300 2 1.44
5 5.09
11 12.72
19 22.91
21 25.45
1,200 2 0.64
5 2.26
11 5.65
19 10.18
21 11.31
600 2 0.240
5 0.848
11 2.12
19 3.81
21 4.24
high-voltage converters using commercially available semiconductor devices.

Table3.9 summarizes the voltage handling capacity of 2-level and multilevel con-
verters. With 2-level converter topology, converters of a maximum of 2.88kV rms
can be designed using commercially available devices, but this limitation can be
easily overcome with multilevel topologies.
Series connection of switching devices is a well-established method to
develop a medium-voltage converter with 2-level converter topology. In order to
reduce the effects of non-unique device characteristics and mismatch of driver
circuit operations, the method uses switching aid circuits, such as snubber and/
or active gate control circuits, which increases the switching loss. The switch-
ing loss mostly depends on the communication voltage of the switching devices,
which can be much lower with multilevel converter topologies than that of the
2-level converters. Moreover, multilevel converters require no switching aid cir-
cuits, which improve the commutation times. Therefore, multilevel converters give
lower switching losses than those of the 2-level converters. A power loss compari-
son between the MMC converter and 2-level converter was analyzed, and it was
reported that the switching losses of MMC converters were much lower than those
of the 2-level series-connected converters operating under the same conditions [9].
Among the variety of topologies for multilevel converter structures reported in
the recent decades, the NPC, FC, and MMC converter with separate DC sources
are three commonly used structures. Most of the other topologies for multilevel
converters have been designed through the modification or combination of these
three basic topologies. A brief description of these three multilevel converter
topologies is presented in Sects. 2.12.3.
3.3.1Neutral Point Clamped Converter
In order to increase the converter voltage rating without increasing the voltage
rating of switching devices, the NPC converter topology was introduced in 1981
[10]. A 3-phase 7-level NPC converter circuit is shown in Fig.3.39. A phase leg
of a 7-level converter consists of six pairs of switching devices and five pairs of
diodes. Each switch in a pair works in complimentary mode, and the respective
diode pair provides access to the midpoint. There are seven switching states to
synthesize the 7-level phase voltage. If it is assumed that each auxiliary diode volt-
age rating is the same as the active switching device voltage rating, the number of
diodes required for each phase leg will be (2m4).
3
V)
.7 k .4 k
A
Per unit price (k AUD)
at 1 0
age t at
vol
t ren
2 d cur
Rate ate
d
R
1 Price against rated voltage

Price against rated current
0
0 1 2 3 4 5 6
Rated voltage or current (kV or kA)
Fig.3.38THDs of 33 and 11kV multilevel converters [3]

An m-level NPC converter has an m-level output phase voltage and a

(2m 1)-level output line voltage. Figure3.40 shows the output line-to-line
voltage waveform of 7-level converter. There are seven-voltage levels to the
peak value and 13 levels to peak-to-peak value. The difference between two suc-
cessive voltage levels is equal to the voltage across each DC-link capacitor. The
three phases of the NPC converter share a common DC bus voltage, which has
been subdivided by (m 1) capacitors into equally m-levels. Each phase leg of
the NPC converter requires (2m2) active switching devices, and the commuta-
tion voltage of each active switching device is (m1) times lower than the total
DC-bus voltage. For example, in total, there are six DC-link capacitors in a 7-level
NPC converter and the voltage across each capacitor is Vdc/6.
In 1998, the NPC converter-based high-power AC motor drive system was
proposed [11]. In 2002, another application of the NPC converter for high-power
medium-voltage variable speed motor drives was proposed [12]. The NPC con-
verter has also attracted significant attention because it serves as an interface
between a high-voltage DC transmission line and an AC transmission line [13].
SA11 SB11 SC11
C
SA9 SB9 SC9
D1 D1 D1
SA7 SB7 SC7

D2 D2 D2
C
SA5 SB5 SC5
D3 D3 D3
SA3 SB3 SC3

D4 D4 D4
SA1 SB1 SC1
C
D5 D5 D5
A
B n
DC C
SA2 SB2 SC2

Load
D6 D6 D6
C SA4 SB4 SC4
D7 D7 D7
SA6 SB6 SC6
D8 D8 D8
SA8 SB8 SC8
C D9 D9 D9
SA10 SB10 SC10
D10 D10 D10
SA12 SB12 SC12

C
Fig.3.39Three-phase 7-level NPC converter circuit topology

The NPC converter topology has the ability to connect the neutral point to the
middle point of the DC link thereby reducing the ground leakage currents and ena-
bling this topology to form back-to-back connection. Accordingly, in the recent
years, the back-to-back NPC converters have also been largely utilized in the grid
interfacing of renewable energy sources such as wind turbine generators and solar
PV arrays. In total, there are (6m 12) auxiliary diodes required for a 3-phase
m-level converter.
When the number of converter levels is sufficiently high, the number of diodes
required will make the system impractical to implement. This is due to that the
diode reverse recovery of these huge clamping diodes becomes a critical fac-
tor. The reverse recovery problem can be reduced by using the newly developed
silicon carbide diode which has negligible reverse recovery time. However, this
technology is still under development. Only 40A or lower-rated silicon carbide
diodes are currently available on the market. All the converters in this chapter are
designed with 250A rated diodes, as the silicon carbide technology is still not
available for these designs. On the other hand, these special diodes significantly
increase the semiconductor cost of the converter (a 3-phase 43-level NPC con-
verter requires 246 diodes). For example, the unit price of a 600V 40A silicon
carbide diode is about 25 USD. Therefore, the reverse recovery problem is still
a challenging issue for designing medium- or high-voltage converters with NPC
converter topology. In addition to the requirement for several auxiliary diodes,
the capacitor voltage balancing is one of the notable drawbacks of NPC topol-
ogy. Unbalanced capacitor voltages may cause distortions in output waveform
and damage switching devices due to the over voltage breakdown of switching
devices. For this reason, many researchers are turning their attention to develop-
ing extra circuits and control algorithms [14], but these circuits and algorithms are
for no more than five levels. Moreover, the boost switch may limit the voltage and
power rating of the converter. Therefore, this topology is not suitable for medium-
or high-voltage applications, where a higher number of levels are required.
3.3.2Flying Capacitor Converter
In 1992, FC converter was proposed [15]. The overall structure of the FC con-
verter is comparable with that of the NPC converter except that the converter uses
auxiliary capacitors instead of auxiliary diodes. The circuit topology of a 3-phase
7-level FC converter is shown in Fig.3.41.
A phase leg of the 7-level converter consists of six pairs of switching devices
and five branches of capacitors. An m-level FC converter has an m-level output
phase voltage and a (2m 1)-level output line voltage. The three phases of the
FC converter may share a common DC-bus voltage similar to the NPC converter,
which has been subdivided by (m1) capacitors into equal m-levels. Each phase
leg of the FC converter requires (2m2) active switching devices, and the com-
mutation voltage of each active switching device is (m 1) times lower than
Fig.3.40Output line-to- 7
6
s
line voltage of 7-level NPC
el
5
ev
converter 10 4
el
Voltage (kV)
g
3
ta
ol
2
V
1
0
-10
5 10 15 20 25 30 35
Time (ms)
the total DC-bus voltage. In addition to the DC-link capacitors, the FC converter
requires large number of auxiliary capacitors. If the voltage rating of the capaci-
tors is equal to that of the active switching devices, the number of auxiliary capac-
itors required for each phase leg will be (m23m+2)/2 [16]. There are several
possible switching states to synthesize a desired output. The optimal switching
states may ensure good performance and reduce power losses of the converter as
well.
Due to some features (e.g., the control of real and reactive power and the capa-
bility to ride through short-duration outage and voltage sags), the FC-voltage
source converter has found its application in power systems. In 2002, an FC mul-
tilevel-voltage source converter-based unified power-flow controller was proposed
[17]. In 2007, a 3-level FC multilevel converter-based high-voltage direct current
power transmission system was presented [18]. However, the number of auxil-
iary capacitors required is quadratically related to the number of levels. When the
number of converter levels is sufficiently high, the number of capacitors required
makes the system impractical to implement. The large number of capacitors not
only increases the converter size and weight but also significantly reduces the life-
time of the converter. Due to the regulation of the capacitor voltages, the FC multi-
level converter requires complex control strategies. Moreover, the pre-charging of
clamping capacitors with the required voltage level may increase the complexity
and cost, and decrease the performance of the converter.
3.3.3Modular Multilevel Cascaded Converter
In the NPC and FC converters, modification of circuit configuration, e.g., increase

or decrease of the number of levels is complicated, which affects the entire circuit
of the converter. On the other hand, the MMC converter is completely modular
in structure, where it is easy to increase or decrease the number of levels of the
converter. The modification of number of converter levels does not affect the entire
circuit. In MMC converters, each phase leg consists of a series of modular single-
phase full-bridge or H-bridge inverter cells connected in series. A 3-phase 7-level
SA11 SB11 SC11
SA9 SB9 SC9

C11 C11 C11
SA7 SB7 SC7
C C7 C7 C7
SA5 SB5 SC5
C4 C4 C4
SA3 SB3 SC3

C12 C12 C12
C2 C2 C2
SA1 SB1 SC1
C C8 C8 C5 C8 C5
A
DC C1
C5 B n
C1
C1
SB2 C
C13 SA2 C13 C13 SC2
C9 C9 C9
Load
C SA4 SB4 SC4
C3 C3 C3
C14 SA6 C14 C6 SB6 C14 SC6

C6 C6
SA8 SB8 SC8

C C10 C10 C10
SA10 SB10 SC10

C15 C15 C15
SA12 SB12 SC12

C
Fig.3.41Three-phase 7-level FC converter circuit topology
MMC converter circuit is illustrated in Fig.3.42, which is Wye connected. The

3-phase MMC converter can also be connected in delta.
Each isolated and balanced DC source is connected to a single-phase full-
bridge or H-bridge inverter cell. The requirement of multiple isolated and balanced
DC supplies may limit the application of MMC converter. For example, a 3-phase
7-level MMC converter requires nine DC supplies. Recent research has proposed
advanced magnetic material-based high-frequency magnetic link toward the solu-
tion of this problem [19, 20]. The four switching devices of each H-bridge inverter
cell can generate three different voltage outputs: +Vdc, 0, and Vdc by connect-
ing the isolated DC source to the AC output. The proper control of the switch-
ing devices superimposes the voltage outputs of all the H-bridge cells of a phase
in a staircase form in order to achieve high voltage at the converter output. Each
H-bridge inverter cell contributes two voltage levels in the output voltage. Similar
to the NPC and FC converters, an m-level MMC converter also has an m-level out-
put phase voltage and a (2m1)-level output line-to-line voltage. As the NPC and
FC multilevel converters, the MMC converter requires (2m2) active switching
devices per phase leg. The commutation voltage of each active switching device
is (m1) times lower than the total DC-bus voltage. There are several possible
switch combinations to synthesize the 7-level phase voltage.
Because of their special features (e.g., the number of components scales lin-
early with the number of levels, and individual modules are identical and com-
pletely modular in construction hence enabling high-level number attainability),
the MMC converter topology has been considered the best possible candidate for
medium- to high-voltage high-power applications. In 1990, a single-phase MMC
converter structure was presented for plasma stabilization applications [21]. In
1997, Robicon Corporation presented a medium-voltage high-power 3-phase
MMC converter for motor drive applications [22]. The complicated multi-winding
phase-shifted power-frequency transformer was used to deliver electric power to
all the floating H-bridge inverter cells. In 1996, an MMC-based static synchronous
compensator for reactive power control was presented [23]. In 1999, the converter
was proposed for heavy-duty electric and hybrid-electric vehicles that have large
electric motor drives [24]. A series of isolated batteries were used to deliver elec-
tric power to all the floating H-bridge inverter cells.
3.3.4Switching Schemes of Multilevel Converters
A phase shift or level shift can be introduced between the carrier signals of con-
tiguous cells to produce a phase-shifted or level-shifted switching pattern between
them. In this way, a stepped multilevel waveform is generated. For the m-level
converter, (m1) phase-shifted or line-shifted carriers with the same frequency
fca are required. If fm is the frequency of the reference signal, the frequency modu-
lation index of the converter can be calculated as
fca
mf = . (3.15)
fm
3.3.4.1Phase-Shifted Carrier-Based Switching Scheme
The phase-shifted carriers are specially conceived for FC [25] and MMC [22]
converters. Since each FC cell is a 2-level converter, and each H-bridge cell is a
3-level inverter, the traditional bipolar (using one carrier signal that is compared
to the reference to decide between two different voltage levels, typically the
positive and negative busbars of a voltage source converter) and unipolar PWM
techniques can be used, respectively. Due to the modularity of these topologies,
each cell can be modulated independently using the same reference signal in a
phase. If Ac the peak-to-peak amplitude of the carrier signals and Am the peak-
to-peak amplitude of the reference signal, the amplitude modulation index can be
calculated as
Am
map = . (3.16)
Ac
Figure 3.43 shows the block diagram of the phase-shifted switching scheme for
the 7-level converter. This modulation scheme requires three reference signals;
H-bridge cell H-bridge cell H-bridge cell
SA1 SA3 SB1 SB3 SC1 SC3
Vdc Vdc Vdc

Vdc Vdc Vdc

B n
C
Load
Vdc Vdc Vdc

Fig.3.42Three-phase 7-level modular multilevel cascaded converter topology
120 shifted by each other and one for a particular phase leg. The number of car-
rier signals depends on the number of converter levels; in total, there are (m1)/2
carrier signals in an m-level converter. The carrier phase shifting for that particular
cell or pair can be calculated as [26]
360
ps = . (3.17)
(m 1)
The MATLAB/Simulink model to generate 3-phase-shifted carrier signals for a
3-phase 7-level converter is illustrated in Fig.3.44. These three-generated carriers
may be used to drive the left-side H-bridge arms of three H-bridge inverter cells
on a phase leg. Figure3.45 plots the three-generated carrier signals. The inverted
carrier signals may be used to generate gate pulses to drive the right-side H-bridge
leg of three H-bridge inverter cells on a phase leg. Figure3.46 plots the inverted
carrier signals. In total, there are six carrier signals and three reference signals in
the modulation scheme of 3-phase 7-level converter. Figure3.47 plots all the car-
rier and reference signals of a 3-phase 7-level converter.
Each compare unit generates one switching signal for the top switching device
of a half-bridge cell or a pair. The inverted form of this switching signal drives
the bottom switching device. Figure3.48 shows the gate pulse generation tech-
nique for the top switching device. The technique to generate gate pulse for the
bottom switching device is illustrated in Fig.3.49. For the left half-bridge cell, one
Reference signal
generator (3-phases)
SA1
>= 0 SB1
SC1
Comp
0o shifted carrier SA2
signal generator SB2
NOT
SC2
Inv
SA3
>= 0 SB3
Inv SC3
Comp
SA4
SB4
NOT
SC4
Inv
SA5
>= 0 SB5
SC5
Comp
NOT
SC6
Inv
SA7
>= 0 SB7
Inv SC7
Comp
SA8
SB8
NOT
SC8
Inv
SA9
>= 0 SB9
SC9
Comp
NOT
SC10
Inv
SA11
>= 0 SB11
Inv SC11
Comp
SA12
SB12
NOT
SC12
Inv
Gate pulses
Fig.3.43Phase-shifted switching scheme for the 3-phase 7-level FC and MMC multilevel
converters
1
sin asin 2/pi 1
theta
1
0 -K-
3
sin asin 2/pi 2
60 -K-
2
Scope1
sin asin 2/pi 3

120 -K-
3
Fig.3.44MATLAB/Simulink model to generate phase-shifted carrier signals for 7-level converter

d
ifted
shifte
0 o sh
0.5
o
Carrier signals
120
0
hifted
-0.5
60 o s
-1
0 2 4 6 8 10
Time (ms)
Fig.3.45Three-generated phase-shifted carrier signals
1
Inverted carrier signals
0.5
-0.5
-1
0 2 4 6 8 10
Time (s)
Fig.3.46Three-inverted phase-shifted carrier signals
0.5
Signals
-0.5
-1
0 2 4 6 8 10
Time (ms)
Fig.3.47Three reference signals and six phase-shifted carrier signals-based modulation scheme
of 3-phase 7-level converter
Gate pulse PS1

1
Signals 0.5
-0.5
-1 Carrier Reference
0 5 10 15 20
Time (ms)
Fig.3.48Generation of gate pulse PS1 with phase-shifted carrier
Gate pulse PS2

1
0.5
Signals
-0.5
-1 Carrier Reference
0 5 10 15 20
Time (ms)
Fig.3.49Generation of gate pulse PS2 with phase-shifted carrier
is asserted when the reference signal value is greater than or equal to the carrier
signal value, and the other is asserted when the reference signal value is less than
the carrier signal value. For the right half-bridge cell, one is asserted when the
inverted carrier signal value is greater than or equal to the reference signal value,
and the other is asserted when the inverted carrier signal value is less than the ref-
erence signal value. Figure3.50 shows the MATLAB/Simulink model to generate
the gate pulses for a 3-phase 7-level converter.
For the FC converter, the advantage of the even power distribution is that once
the FCs are properly charged (initialized to their corresponding values), no imbal-
ance will be produced due to the self-balancing property of this topology and as a
result there is no need to control the DC-link voltages [27, 28]. Another interest-
ing feature is that the total output voltage has a switching pattern with k (number
of the power cells) times the frequency of the switching pattern of each cell. This
multiplicative effect is produced by the phase shifts of the carriers. Hence, lower
THD is obtained at the output, using the k times lower frequency carriers. With
the phase-shifted carrier-based modulation scheme, the control signal assignment

to the appropriate semiconductor of the MMC converter is easy and this remains
simple even when the level number increases to higher values.
3.3.4.2Level-Shifted Carrier-Based Switching Scheme
In the level-shifted scheme, the carriers are arranged in vertical shifts instead of
the horizontal shift. Each carrier is set between two voltage levels, and hence, it is
named as level shifted. Since each carrier is associated with two levels, the same
principle of bipolar PWM can be applied which takes into account the fact that the
control signal has to be directed to the appropriate semiconductors in order to gen-
erate the corresponding levels [29]. The carriers span the whole amplitude range
that can be generated by the converter.
The level-shifted carrier-based modulation scheme is especially useful for NPC
converters, since each carrier can be easily associated with two power switches
of the converter [30]. Moreover, the level-shifted carrier-based modulation scheme
provides less distorted line voltages because all carriers are in phase in contrast
to the phase-shifted carrier-based modulation scheme. Although the level-shifted
carrier-based modulation scheme may provide reasonable performance, the com-
plexity of control signal assignment to the appropriate semiconductors makes it
unpopular for MMC multilevel converter applications. Further, the level-shifted
carrier-based modulation scheme generates more distorted input current in the
MMC converter and a capacitor voltage imbalance in the FC converter [30].
Figure3.51 plots the reference and carrier signals for a 3-phase 7-level NPC con-
verter. In the level-shifted scheme, the carriers on a particular phase leg are level
shifted by
Am
ls = . (3.18)
(m 1)
The peak-to-peak amplitude of carrier signals can be calculated by
Am
Acl = . (3.19)
(m 1)
Figure3.52 shows the block diagram of the level shifted switching scheme for the
3-phase 7-level converter.
Each comparator unit generates one switching signal for the top switch-
ing device of a pair. The inverted form of this switching signal drives the bottom
switching device. Figure3.53 shows the gate pulse generation technique for the
top switching device. The technique to generate gate pulse for the bottom switch-
ing device is illustrated in Fig.3.54. For the top pair, one is asserted when the ref-
erence signal value is greater than or equal to the carrier signal value, and the other
is asserted when the reference signal value is less than the carrier signal value.
For the second top pair, one is asserted when the shifted carrier signal value is
3
1 TT1
3
Ust 3
3
3
>=
3
NOT TT2
3
0 Com
Inv
3 TT3
3
3
3
>=
-1 3
1 Com1 NOT TT4
3
Inv1
1
3 TT5
3
3
3
>=
3
2 Com2 NOT TT6
3
Inv2
-1
2 theta 2
3 TT7
wt 3
3
3
>=
Com3
3
3 NOT TT8
3
Inv3
3 TT9
3
3 3
3
>=
3
4 Com4 NOT TT10
3
Inv4
Carriers
-1
Generator
3
3 TT11
3
3
>=
3
5 Com5 NOT TT12
3
Inv5
Fig.3.50MATLAB/Simulink model to generate gate pulses for 3-phase 7-level converter
greater than or equal to the reference signal value, and the other is asserted when
the shifted carrier signal value is less than the reference signal value. In multilevel
converters with level-shifted carriers, the amplitude modulation index can be cal-
culated as
Am
mal = . (3.20)
(m 1)Acl
0.5
Signals
-0.5
-1
0 5 10 15 20 25 30
Time (ms)
Fig.3.51Three reference signals and four level-shifted carrier signals
3.3.4.3Reference Signals in Switching Schemes
Different types of major reference signals used in the traditional converters

can also be used in the multilevel converter system, e.g., sinusoidal, third-har-
monic injected sinusoidal, 60 modulated sinusoidal and trapezoidal. Each has
its unique advantages and disadvantages. It is possible to reduce the switching
losses by reducing the number of switching occurrences in each period. Flatting
the top of the reference signal waveform in the third-harmonic injected sinusoi-
dal, 60 modulated sinusoidal and the trapezoidal not only allows the possibil-
ity of switching loss reduction but also increased the range of linear modulation.
These four reference signals with phase-shifted carrier signals generate four
modulation schemes: (a) the phase-shifted carriers with sinusoidal references
known as the sine pulse width modulation (SPWM), (b) the phase-shifted carriers
with third-harmonic injected sinusoidal references known as the third-harmonic
injected pulse width modulation (THPWM), (c) the phase-shifted carriers with
60 modulated sinusoidal references known as the 60 pulse width modulation
(SDPWM), and (d) the phase-shifted carriers with trapezoidal type references
known as the trapezoidal pulse width modulation (TRPWM). Figure3.55 shows
the MATLAB/Simulink model to generate 120 shifted sinusoidal three reference
signals. Parameters such as modulation index, frequency, and phase displacement
can be controlled through this block. Figure3.56 plots the sinusoidal reference
and carrier signals.
The MATLAB/Simulink model as shown in Fig.3.57 can be used to gener-
ate the third-harmonic injected sinusoidal reference signals. The amplitude of the
third-harmonic component can be controlled to obtain the desired quality output.
The flatting top of the reference signal reduces the number of commutation in a
cycle, which helps reduce the switching losses of the converter. Figure3.58 shows
the reference and carrier signals of THPWM scheme.
The amplitude-limited triangular signals can be used as the trapezoidal refer-
ence signal, which can easily be generated by the MATLAB/Simulink model as
Carrier signal SA11

generator
>= 0 SB11
5/6 SC11
Comp
SA2
SB2
NOT
SC2
Inv
SA9
>= 0 SB9
3/6 SC9
Comp
SA4
SB4
NOT
SC4
Inv
SA7
>= 0 SB7
1/6 SC7
Comp
SA6
SB6
NOT
SC6
Inv
SA5
>= 0 SB5
1/6 SC5
Comp
SA8
SB8
NOT
SC8
Inv
SA3
>= 0 SB3
3/6 SC3
Comp
SA10
SB10
NOT
SC10
Inv
SA1
>= 0 SB1
5/6 SC1
Comp
SA12
Reference signal SB12
NOT
generator (3-phases) SC12
Inv
Gate pulses
Fig.3.52Level-shifted modulation technique for a 3-phase 7-level NPC converter

0.5
Signals Carrier
0
Gate pulse PS7
-0.5
Reference
-1
0 5 10 15 20
Time (ms)
Fig.3.53Generation of gate pulse PS7 with level-shifted carrier
0.5
Carrier
Signals
0
Gate pulse PS2
-0.5
Reference
-1
0 5 10 15 20
Time (ms)
Fig.3.54Generation of gate pulse PS2 with level-shifted carrier
2*pi
m
Freq 1
sin Ust
12:34
[0 -2*pi/3 2*pi/3]
Phase -K-
deg. ->rad
Fig.3.55MATLAB/Simulink model to generate sinusoidal reference signals

0.5
Signals
-0.5
-1
0 5 10 15 20
Time (ms)
Fig.3.56SPWM modulation scheme for 3-phase 5-level converter
2*pi
m
Freq
sin
12:34
t
1
Ust
[0 -2*pi/3 2*pi/3]
3
Phase -K-
-K-
deg. ->rad sin
Fig.3.57MATLAB/Simulink model to generate third-harmonic injected sinusoidal reference

signals for 3-phase 5-level converter
shown in Fig.3.59. The parameters of the reference signals such as the frequency,
maximum amplitude, and phase displacement can easily be controlled in the
Simulink environment. This modulation scheme also uses flatting top references,
which significantly reduce the switching losses of the converter. Figure3.60 plots
the trapezoidal reference signals.
Sixty degree pulse width modulation scheme also use sinusoidal references but
their amplitudes are limited to 60, clamped the magnitude of the reference signal,
sin at a constant value while is higher than 60. The limiting of the peak value
actually makes the top of the reference signals flattened. The MATLAB/Simulink
model to generate sinusoidal reference signals can also be used to generate 60
reference signals. An extra amplitude limiter may be added in the model as shown
in Fig.3.61. Figure3.62 plots the reference and carrier signals for SDPWM
scheme.
The modulation schemes are applied on the 7-level to 19-level MMC converter
systems, and performance is analyzed in the MATLAB/Simulink environment.
0.5
Signals
0
-0.5
-1
0 5 10 15 20
Time (ms)
Fig. 3.58Third-harmonic injected pulse width modulation schemes for 3-phase 5-level
converter
2*pi
m
Freq
1
sin asin Ust
12:34
[0 -2*pi/3 2*pi/3]
Phase -K-
deg. ->rad
Fig.3.59MATLAB/Simulink model to generate trapezoidal reference signals for 3-phase

5-level converter
0.5
Signals
-0.5
-1
0 5 10 15 20
Time (ms)
Fig.3.60Trapezoidal pulse width modulation schemes for 3-phase 5-level converter

2*pi
m
Freq 1
sin Ust
12:34
t
[0 -2*pi/3 2*pi/3]
Phase -K-
deg. ->rad
Fig.3.61MATLAB/Simulink model to generate 60 sinusoidal reference signals for 3-phase

5-level converter
0.5
Signals
-0.5
-1
0 5 10 15 20
Time (ms)
Fig.3.62Sixty degree pulse width modulation schemes for 3-phase 5-level converter
Of these four modulation schemes, the THPWM-based SPWM scheme gives the
best harmonic performance, as shown in Fig.3.63. The SDPWM and TRPWM
schemes have higher lower-order harmonic content than the SPWM and THPWM
schemes. In addition, the SPWM has shown a higher reduction rate for the high-
level number.
3.4Selection of Multilevel Converter Topology

forMedium-Voltage Applications
The objective of this section is to find a suitable multilevel converter topology

to design a medium-voltage converter using matured low-voltage semiconductor
technology for renewable generation systems. Parameter identification is a vital
part of the medium-voltage converter design process. The operating conditions and
basic converter data are summarized in Table3.10. If Vll(rms) is the grid line-to-line
Table3.10Converter Technical data Five-level 11-level

specification
Converter line voltage 11kV 11kV
Minimum DC-link voltage 3,882V 1,553V
Nominal DC-link voltage 4,044V 1,618V
Phase current 250A 250A
Apparent output power 4.76MVA 4.76MVA
Converter carrier frequency 12kHz 12kHz
Output frequency 50Hz 50Hz
voltage and m the number of levels of the converter, the minimum DC-link voltage
of each H-bridge inverter cell or each DC-link capacitor can be calculated as
Vll(rms)
Vdc(min) = 2 . (3.21)
(m 1)
To determine the nominal DC-link voltage of each H-bridge inverter cell or each
DC-link capacitor, a voltage reserve of 4% is assumed, i.e.,
Vdc(nom) = 1.04 Vdc(min) . (3.22)
If Ip(rms) is the converter phase current, the apparent output power (for Wye con-
nection) can be calculated from

Sc = 3Vll(rms) Ip(rms) . (3.23)
Semiconductor utilization is a very important issue in the design process of a

medium- or high-voltage converter system as this is a considerable cost. The
DC-link voltage needs to be considered when selecting IGBTs and diodes voltage
ratings, and a cosmic ray effect assessment may also be necessary. In addition to
the output capacity and voltage ratings of the converter, the availability of IGBT
and diode modules in the market needs to be considered in the design process. For
the 5-level converter, the commutation voltage of each H-bridge cell or DC-link
capacitor is 4,044V. The highest-rated IGBT in the market is 6.5kV, which rec-
ommends the use of the 3.6kV or lower-voltage applications. Therefore, two
4.5kV series-connected IGBTs are assumed as a single switch for all the 5-level
converter topologies. Table3.11 summarizes the design of the power semicon-
ductors for the converter specification in Table3.10, with a carrier frequency of
12kHz. To enable a converter output phase current of 250A, a 400A current
rating is chosen for the power semiconductors. Assuming, the maximum capacitor
voltage ripple, Vc to be 5% of the DC-link voltage and the DC output current,
Idc to the maximum amplitude of the phase current, Ip(rms) [31], we can calculate
the capacitance of the FC as
Ip(rms)
C= . (3.24)
nc Vc fc
where nc is the number of series-connected FC cells.
3.4 Selection of Multilevel Converter Topology for Medium-Voltage Applications 97
Table3.11Components of different multilevel converters [16]

Number of levels 5 11
Topology NPC FC MMC NPC FC MMC
IGBTs 48 48 48 60 60 60
Rated voltage (kV) 4.5 4.5 4.5 3.3 3.3 3.3
Vcom (V) 2,022 2,022 2,022 1,618 1,618 1,618
Vcom@100FIT (V) 2,250 2,250 2,250 1,800 1,800 1,800
DVUF (%) 90 90 90 90 90 90
Diodes 18 54
Capacitors 18 135
In order to select a suitable converter topology for a medium-voltage converter,

the following basic multilevel converter topologies are considered: a 5-level NPC,
a 5-level FC, a 5-level MMC, an 11-level NPC, an 11-level FC, and an 11-level
MMC. The comparison is made in terms of the number of semiconductors, semi-
conductor cost, THDs, filter size, and control complexity of the converters. The
performance is analyzed and compared in the MATLAB/Simulink environment.
To generate switching pulses, a level-shifted carriers-based switching scheme is
used for the NPC converter topologies and a phase-shifted carriers-based switch-
ing scheme is used for the FC and MMC converter topologies with a carrier fre-
quency of 12kHz and a modulation index of 0.80.98.
3.4.1Detailed Design and Analysis of 5-Level and 11-Level

NPC Converters
With a 5-level converter topology, the commutation voltage of each DC-link

capacitor is 4,044V, i.e., each voltage level contributes 4,044V to the output
voltage waveform. The output line voltage waveform of a 5-level NPC converter
consists of nine voltage levels and moderately matches the reference sine signal.
Figure 3.64 plots the line voltage of a 5-level NPC converter without using the
line filter circuit. The THDs of the line voltage are calculated at 17.26%. The fre-
quency spectrum is depicted in Fig.3.65.
With the 11-level converter topology, the communication voltage of each
DC-link capacitor is 1,618V, i.e., each voltage level contributes 1,618V to the
output voltage waveform. The output line voltage waveform of an 11-level NPC
converter consists of twenty-one voltage levels, and this aligns strongly with the
reference sine signal. Figure3.66 plots the line voltage of an 11-level NPC con-
verter without using the line filter circuit. The THDs of the line voltage are calcu-
lated at 7.07%. The frequency spectrum is depicted in Fig.3.67.
The output voltage waveform of an 11-level NPC converter coincides to a
large degree with the reference sine wave as compared to other 5-level converter
14
Study on 11 kV converter
12
TRPWM SPWM
SDPWM THPWM
10
THD (%)
6
5 % limit
2
7 8 10 12 14 15 16 18 19
Number of levels
Fig.3.63Calculated THD at different level numbers ranging from 7-level to 19-level [1]
Fig.3.64Line voltage of
15
ence
5-level neutral point clamped
ge
volta
converter 10
Refer
Voltage (kV)
Line
5
0
D
% TH
-5
17.26
-10
-15
20 30 40 50 60 70 80
Time (ms)
Fig.3.65Frequency
spectrum of 5-level neutral 0.04

point clamped converter
0.02
0
0 5000 10000 15000
Frequency (Hz)
outputs. All 11-level converter output voltage waveforms are very close to the ref-
erence sine wave while the NPC converter performance is better than the others.
Furthermore, from the output figures, it is clear that increasing the number of levels
means improving the converter performance. It is assumed that each blocking diode
voltage rating is the same as the active device voltage rating. A total of 54 diodes
are required for an 11-level NPC converter. The large number of diodes affects
Fig.3.66Line voltage 15
ge
e
enc
volta
of 11-level neutral point
10 1
fer
clamped converter
Re
Line
Voltage (kV)
5
D
% TH
-5
7.07
-10 -1
-15
20 30 40 50 60 70 80
Time (ms)
the reverse recovery of the clamping diodes, and this is a major design challenge
in high-voltage high-power systems. A list of the number of power components
required for each converter topology is shown in Table3.11. As already stated, the
availability of IGBT and diode modules is also considered when designing the con-
verter. For the 5-level NPC, FC and MMC converter topology, each IGBT switch is
formed from the series connection of two 4.5kV IGBTs so the number of IGBTs
is 48. To enable a converter output phase current of 250A, the simulation result is
used to determine the current rating of the power semiconductors.
3.4.2Detailed Design and Analysis of 5-Level

and 11-Level FC Converters
A total of 6 and 45 clamping capacitors are required for each phase of the 5-level
and 11-level FC converters, respectively. These large numbers increase the con-
verter size and cost. In addition, they reduce the overall lifetime of the converter.
The capacitor voltage balancing problem also becomes a challenging issue with
such a high number of components. An 11kV 5-level FC converter also requires 4
DC-link capacitors with a rated voltage of 4,044V. Each voltage level contributes
4,044V to the line- or phase-voltage waveforms. The line voltage waveform of a
5-level FC converter is shown in Fig.3.68. Due to capacitor voltage imbalance, the
voltage waveform of the FC converter is severely distorted compared with that of
the NPC converter. Therefore, the harmonic performance is not as good as that of
the NPC converter. Figure3.69 illustrates the harmonic spectrum of the 5-level FC
converter. In addition to 135 auxiliary capacitors, an 11kV 11-level FC converter
requires 10 DC-link capacitors with a rated voltage of 1,618V.
Each voltage level contributes 1,618V to the line- or phase-voltage waveforms.
The line voltage waveform of an 11-level FC converter is shown in Fig.3.70.
Although, the 11-level converter output voltage waveform matches the reference
sine signal, the performance is not as good as that of the NPC converter. The output
line voltage frequency spectrum of an 11-level FC converter is shown in Fig.3.71.
Fig.3.67Frequency
spectrum of 11-level neutral 0.04
point clamped converter
0.02
0
0 5000 10000 15000
Frequency (Hz)
Fig.3.68Line voltage
15
of 5-level flying capacitor
ence
ge
converter 10
volta
Refer
Voltage (kV)
Line
0
D
-5 % TH
17.80
-10
-15
20 30 40 50 60 70 80
Time (ms)
Fig.3.69Frequency
spectrum of 5-level flying 0.04

capacitor converter
0.02
0
0 5000 10000 15000
Frequency (Hz)
Fig.3.70Line voltage of 15
Reference
11-level flying capacitor

Line voltage
converter 10 1
Voltage (kV)
00
D
% TH
-5
7.28
-10 -1
-15
20 30 40 50 60 70 80
Time (ms)
3.4.3Detailed Design and Analysis of 5-Level

and 11-Level MMC Converters
There are no blocking diodes or clamping capacitors in the MMC converter topol-
ogy. The number of components scales linearly with the number of levels. Hence,
the overall number of components is much lower than that of other multilevel
converter topologies. The individual modules are similar and totally modular in
construction, which makes it easy to implement for any number of levels. The
higher number of attainable levels provides more scope for reducing harmonics.
The high number of levels means that it is possible to connect the converter to
the medium-voltage grid directly. However, unlike the NPC and FC converters, the
MMC converter requires multiple isolated and balanced DC supplies. For exam-
ple, a 3-phase 11-level MMC converter requires 15 isolated and balanced DC sup-
plies. As the number of voltage levels increases, the voltage rating of these DC
supplies decreases to a low level and this is favorable for some applications. The
voltage rating of these DC supplies actually determines the voltage ratings of the
active switching devices for each H-bridge inverter cell. The voltage balancing of
multiple DC supplies is important for good output voltage waveforms. The output
line voltage waveform of a 5-level MMC converter is shown in Figs.3.72, and
3.73 shows its frequency spectrum. As calculated, an 11kV 11-level MMC con-
verter generates a 9-level line voltage waveform with 18.13% THDs. The DC sup-
plies voltage rating of an 11kV 11-level MMC converter is 1,618V. The available
Fig.3.71Frequency
spectrum of 11-level flying 0.04

capacitor converter
0.02
0
0 5000 10000 15000
Frequency (Hz)
5-level modular multilevel
ence
ge
10
volta
cascaded converter
Voltage (kV)
Refer
5
Line
0
D
% TH
-5
18.13
-10
-15
20 30 40 50 60 70 80
Time (ms)
Fig.3.73Frequency
spectrum of 5-level modular 0.06
multilevel cascaded converter
0.04
0.02
0
0 5000 10000 15000
Frequency (Hz)
11-level modular multilevel
Reference
Line voltage
cascaded converter 10
Voltage (kV)
5
0
-5 THD%
7.70
-10
-15
20 30 40 50 60 70 80
Time (ms)
3.3kV IGBT may be used to design the 11kV 11-level converter because this
IGBT is recommended for 1,800V maximum applications.
With these IGBTs, 90% DVUF can be ensured and this is important to con-
firm the high performance/cost of the power converter. Figure3.74 plots the line
voltage waveform of an 11-level MMC converter. There are 20 steps in the peak-
to-peak line voltage waveform, and each step contributes 1,618V to the line volt-
age of 32,360V peak-to-peak. The 11kV 11-level MMC converters give 7.70%
THDs, which is the highest among the three multilevel converter topologies. The
frequency spectrum of the line voltage of an 11-level MMC converter is depicted
in Fig.3.75.
3.4.4Performance Analysis and Topology Selection
The MMC converter does not require any auxiliary devices whereas NPC and FC
converters require a huge number of auxiliary diodes and capacitors. Although
at the 3-level, the number of auxiliary devices of the FC converter is compara-
ble with that of the NPC converter, at a higher number of levels, it is a few times
higher. For example, 3-phase 3-level FC and NPC converters require three and
six auxiliary devices, respectively, and 3-phase 17-level FC and NPC converters
Fig.3.75Frequency
spectrum of 11-level modular 0.04
multilevel cascaded converter
0.02
0
0 5000 10000 15000
Frequency (Hz)
Table3.12Cost and performance of different multilevel converters [16]

Total comp. 66 66 48 114 195 60
THD (%) 17.26 17.80 18.13 7.07 7.28 7.70
Complexity 19 29 29 41 55 55
Total cost (AU$) 90,961 117,408 82,027 115,663 109,545 82,159
require 360 and 90 auxiliary devices, respectively. Figure3.76 shows a compari-

son of the auxiliary device requirement in the FC and NPC converters. Table3.12
summarizes the THDs for different multilevel converter topologies.
Among these three converter topologies, the NPC converter topology gives
the best harmonic performance. The harmonic performance of the MMC con-
verter topology is not as good as that of the NPC converter topology. The har-
monic content decreases rapidly with the increase in the number of levels. This
means that by increasing the levels of the converter, it is possible to keep the out-
put voltage THDs to be less than, or equal to 5% (according to the standards of
IEEE1547 and IEC61727). Due to the modularity and the lower number of com-
ponent requirements, the MMC converter would probably be the natural choice
for high-level converter applications. Among these three converters, the 11-level
MMC converter is the low-cost high-performance converter, and it is suitable for
the direct connection to the medium-voltage grid. The price data quoted for the
semiconductor devices, and capacitors were collected from the Galco Industrial
Electronics and Farnell catalogs where devices were chosen from the same family
so that it was possible to fit in with requirements.
The IGBTs chosen have integrated freewheel diodes, and hence, the diodes do
not appear in the costing. The current rating of most devices is selected on the
basis of simulation results. Table3.12 also summarizes the estimated cost of dif-
ferent converter systems. The number of semiconductors increases with the
number of levels but the change in terms of cost is small because the price of
the lower-rated device is comparatively much lower. Because of the lower volt-
age and current requirements, the total semiconductor cost of the 11-level MMC
converter is lower than that of all other topologies. The number of arithmetic
Table3.13Comparison of different multilevel converter topologies [16]

Performance 0.92 0.97 1 0 0.02 0.06
Complexity 0 0.28 0.28 0.61 1 1
Cost 0.25 1 0 0.95 0.78 0.003
Total index 1.17 2.25 1.28 1.56 1.80 1.063
Fig.3.76Number of Number of auxiliary devices

auxiliary devices in NPC and
y
FC converters [32] 300 liar
uxi
o f a in FC y
er s liar
mb itor er uxi
200 Nu apac vert o f a PC
c co n er n N
mb s i er
Nu diode vert
100 con
0
2 4 6 8 10 12 14 16 18
Number of levels
Fig.3.77Overall
comparisons of multilevel 2.25
converters [32]
1.80
1.56
1.17 1.28
11-level MMC
1.06
11-level NPC
5-level MMC
11-level FC
5-level NPC
5-level FC
and logic operations (ALOs) for the switching section is calculated through the
MATLAB/Simulink environment. The number of ALOs is used to compare the
complexity of the converters, as summarized in Table3.12.
Based on the performance, control complexity, and semiconductor cost, an
overall comparative study is carried out among the converter topologies men-
tioned. If y is the given value, ymin the minimum value, and ymax the maximum
value in a group of data, the normalized index value can be calculated as
y ymin
Id = . (3.25)
ymax ymin
Table3.13 summarizes the index values of Table3.12. The lower the complexity,
cost and THDs, the better the converter, i.e., a lower index value indicates a better
converter performance/cost. The total index values of all converters are plotted in
Fig.3.77. For the lower-level number converter, the NPC converter topology may
be the best choice. But for the higher-level number converter, the MMC converter
topology may be the best choice. Moreover, in the MMC converter, the individual
modules are identical and completely modular in construction, and thus, in case
of a fault in one of these modules, it is possible to replace it quickly and easily.
Therefore, the MMC is the proper converter topology to design a medium-voltage
converter for the step-up transformer-less direct grid connection of renewable gen-
eration systems (Fig.3.77).
3.5Summary
Among the three basic multilevel converter topologies, the NPC converter topol-
ogy gives the best harmonic performance. The number of auxiliary diodes in an
NPC converter linearly increases with the number of converter level. This large
number of diodes affects the reverse recovery of the clamping diodes which is a
major design challenge in medium- or high-voltage applications. Therefore, the
NPC converter topology is not suitable for medium- or high-voltage applications
where a higher number of converter level is required.
The harmonic performance of the FC converter topology is slightly better than
that of the MMC converter. In the FC converter, the number of required auxiliary
capacitors is quadratically related to the number of converter levels. When the
number of converter levels is high, the number of required capacitors will make
the system impractical to implement. The FC converter requires an extra con-
trol circuit to pre-charge the clamping capacitors with the required voltage level,
which decreases the system performance and increases the complexity and cost
of the converter. These clamping capacitors increase the weight and volume of the
converter and significantly reduce its lifetime. Therefore, the use of FC converter
is still within the 5-level converter topology.
Unlike the NPC and FC converter, the MMC converter does not require any
auxiliary diodes or capacitors. The MMC converter would be the natural choice to
design medium-voltage power converters for step-up transformer-less grid integra-
tion of renewable generation units, because there is no requirement for any aux-
iliary devices, and individual modules are identical and completely modular in
construction. The MMC converters can enable high-level number attainability as
well as significant reduction in the semiconductor cost of the converter.
References
Electron 29(8):41674182
4. Broeck HW, Skudelny HC, Stanke GV (1988) Analysis and realization of a pulse width mod-
ulator based on voltage space vectors. IEEE Transactions Ind Appl 24(1):142150
5. Baker BK, Bannister LH (1975) Electric power converter. US Patent 3 867 643, Feb 1975
6. Peng FZ, Lai JS (1997) Multilevel cascade voltage-source inverter with separate dc source.
US Patent 5 642 275, June 1997
7. Aiello MF, Hammond PW, Rastogi M (2001) Modular multi-level adjustable supply with
series connected active inputs. US Patent 6 236 580, May 2001
9. Massoud MA, Finney SJ, Williams BW (2004) Multilevel converters and series connection
of IGBT evaluation for high-power, high-voltage applications. In: Proceedings of the second
international conference on power electronics, machines and drives, University of Edinburgh,
UK, 31 Mar2 Apr 2004, pp 15
10. Nabae A, Takahashi I, Akagi H (1981) A new neutral-point clamped PWM inverter. IEEE
Trans Ind Appl 17(5):518523
11. Manjrekar MD, Lipo TA (1998) A hybrid multilevel inverter topology for drive applications.
In: Proceedings of the thirteenth annual applied power electronic conference exposition,
Anaheim, California, 1519 Feb 1998, pp 523529
12. Corzine K, Familiant Y (2002) A new cascaded multilevel H-bridge drive. IEEE Trans Power
Electron 17(1):125131
13. Lai JS, Peng FZ (1996) Multilevel convertersa new breed of power converters. IEEE Trans
Ind Appl 32(3):509517
14. Nami A, Zare F, Ghosh A, Blaabjerg F (2010) Multi-output DC-DC converters based on
diode clamped converters configuration: topology and control strategy. IET Trans Power
Electron 3(2):197208
15. Meynard TA, Foch H (1992) Multi-level conversion: high voltage choppers and voltage-
source inverters. In: Proceedings of the 23rd annual IEEE power electronic specialised con-
ference, Madrid, 29 June3 July 1992, pp 397403
grid integration of renewable generation units with medium voltage smart microgrids. In:
Large scale renewable power generation: advances in technologies for generation, transmis-
sion and storage. Springer, Berlin Heidelberg, pp 127149
17. Xu L, Agelidis VG (2002) Flying capacitor multilevel PWM converter based UPFC. IEE
Proc Electric Power Appl 149(4):304310
18. Xu L, Agelidis VG (2007) VSC transmission system using flying capacitor multilevel con-
verters and hybrid PWM control. IEEE Trans Power Delivery 22(1):693701
115(17):17E710-117E710-3
Phys 113(17):17A324-117A324-3
21. Marchesoni M, Mazzucchelli M, Tenconi S (1990) A nonconventional power converter for
plasma stabilization. IEEE Trans Power Electron 5(2):212219
22. Hammond PW (1997) A new approach to enhance power quality for medium voltage AC
drives. IEEE Trans Ind Appl 33(1):202208
23. Peng FZ, Lai JS, McKeever JW, VanCoevering J (1996) A multilevel voltage-source inverter
with separate DC sources for static var generation. IEEE Trans Ind Appl 32(5):11301138
References 107
24. Tolbert LM, Peng FZ, Habetler TG (1999) Multilevel converters for large electric drives.
IEEE Trans Ind Appl 35(1):3644
25. Meynard TA, Foch H (1998) Electronic device for electrical energy conversion between
a voltage source and a current source by means of controllable switching cells. US Patent
5737201 A, Apr 1998
26. Islam MR, Guo YG, Jafari M, Malekjamshidi Z, Zhu JG (2014) A 43-level 33kV 3-phase
modular multilevel cascaded converter for direct grid integration of renewable generation
systems. In: Proceedings of IEEE innovative smart grid technologies conference (ISGT)
Asia, Kuala Lumpur, Malaysia, 2023 May 2014, pp 16
27. Wilkinson RW, Meynard TA, Mouton HT (2006) Natural balance of multi cell converters: the
general case. IEEE Trans Power Electron 21(6):16581666
28. Meynard TA, Foch H, Thomas P, Cournault J, Jakob R, Nahrstaedt M (2002) Multicell con-
verters: basic concepts and industry applications. IEEE Trans Ind Electron 49(5):955964
29. Carrara G, Gardella S, Marchesoni M, Salutari R, Sciutto G (1992) A new multilevel PWM
method: a theoretical analysis. IEEE Trans Power Electron 7(3):497505
30. Rodriguez R, Franquelo LG, Kouro S, Leon JI, Portillo RC, Prats MAM, Perez MA (2009)
Multilevel converters: an enabling technology for high-power applications. Proc IEEE
97(11):17861817
31. Krug D, Bernet S, Fazel SS, Jalili K, Malinowski M (2007) Comparison of 2.3kV medium
voltage multilevel converters for industrial medium voltage drives. IEEE Trans Ind Electron
54(6):29792992
32. Islam MR, Guo YG, Zhu JG (2013) Multiple-input-multiple-output medium frequency-
link based medium voltage converter for direct grid connection of photovoltaic arrays In:
Proceedings of international conference on electrical machines and systems (ICEMS2013),
Busan, Korea, 2629 Oct 2013, pp 202207
Chapter 4
Design and Characterization
of High-Frequency Magnetic Links Used
in Power Electronic Converters
AbstractNowadays, it is common to use high-frequency magnetic links in

designing grid-connected power electronic converters, which can provide elec-
trical isolation without increasing system volume and weight. The modular mul-
tilevel cascaded (MMC) converter topology has received significant attention in
designing medium or high-voltage converters, but it requires balanced multiple
isolated DC sources. To couple a renewable energy source to the MMC converter,
a high-frequency magnetic link with multiple secondary windings can be an excel-
lent option and is explored in this chapter. The high-frequency magnetic link is
used to generate the isolated balanced multiple DC supplies for all of the H-bridge
inverter cells of the MMC converter from a single, low-voltage power source.
Compared with the conventional power frequency transformers operated at 50
or 60Hz, the high-frequency magnetic links (in the range of a few kHz to MHz)
have much smaller and lighter magnetic cores and windings, and thus much lower
costs. The selection of core material and the design of magnetic link are multipha-
sic problems, which highly affects the efficiency of the magnetic link. This chap-
ter presents the design and characterization of high-frequency magnetic links with
advanced magnetic materials.
KeywordsHigh-frequencyMagnetic linksGrid isolationVoltage imbalance

Design Electromagnetic
optimization characteristicsAmorphous alloy
Nanocrystalline
4.1Introduction
It is a common fashion to use high-frequency magnetic links in power electronic

converters. According to the structure of the converter circuits, the high-frequency
magnetic links can provide a few essential functions, such as step-up or step-
down of the voltage levels, isolation of load from source, and generation of iso-
lated multiple sources from single or multiple sources. The modular multilevel
cascaded (MMC) converter requires multiple isolated DC sources that must be

110 4 Design and Characterization of High-Frequency
balanced, and therefore, its application is not straightforward, especially in wind

power generation systems [1]. For example, a 3-phase 13-level converter, as shown
in Fig.4.1, requires eighteen isolated and balanced DC supplies. In this context,
high-frequency magnetic links with multiple secondary windings could be the
possible option to generate isolated and balanced multiple DC supplies for all the
H-bridge inverter cells of MMC converter from a single DC or AC source, e.g.,
a PV array or wind turbine [2, 3]. Increasing the operation frequency can greatly
reduce the physical volume of the transformer that has already been widely used
in the switched mode power supplies. Operated at 1.2kHz, the weight and size of
a 3-MW, 1.2-kHz transformer can be less that 8% of an equivalent 50-Hz unit [4].
In 2011, Pereda and Dixon [5] proposed a high-frequency magnetic link to gen-
erate multiple DC supplies for all of the auxiliary H-bridge cells of asymmetric
cascaded multilevel inverters. In the asymmetric cascaded multilevel converter,
Cell A1 Cell B1 Cell C1
Vdc SA2 SA4 Vdc SB2 SB4 Vdc SC2 SC4
a
b
c
SA5 SA7 SB7 SC5 SC7 n
SB5
Vdc SA6 SA8 Vdc SB8 Vdc SC6 SC8

SB6
Vdc Vdc Vdc

Fig.4.1Power circuit of a 3-phase 13-level MMC converter

4.1Introduction 111
all cascaded modules have different power ratings. This diminishes the modular-
ity and thus limits the high-level attainability and increases the cost of the con-
verter. In this converter system, the high-frequency magnetic link generates
multiple sources from a single source with different voltages and power ratings.
Moreover, in the asymmetric cascaded multilevel converter system, only the auxil-
iary H-bridges are connected through the high-frequency magnetic link. The main
H-bridges are supplied directly from the source, which means that there is no elec-
trical isolation between the grid and load. Therefore, the use of this converter is
limited to the isolated winding motor applications [6]. For step-up-transformer-
less grid-connected renewable generation systems, the medium-voltage converter
should have the capability to ensure electrical isolation and minimize the common
mode issues. In this case, all H-bridge modules (auxiliary and main) require con-
necting through the high-frequency magnetic link.
In 2013, a high-frequency magnetic link with multiple identical secondary wind-
ings was proposed to interconnect wind turbine generator with MMC converter [7].
The identical secondary windings of the common magnetic link generate multiple
isolated sources for the MMC converter. Figure4.2 shows the basic block diagram
of common high-frequency magnetic-link-based medium-voltage converter system.
Although the common magnetic link may ensure equal voltages at the secondary ter-
minals, the topology diminishes modularity of the power conversion system, which
is very important for medium- or high-voltage high-power system. The implementa-
tion of high-power high-frequency inverter is the main challenging issue. Moreover,
the leakage inductance may also limit the power capacity of the high-frequency
magnetic link. In this regard, identical multiple four windings (single primary and
three secondaries) high-frequency magnetic links can be connected in cascaded form
to have same current passing through all primary windings of all magnetic links.
Fig.4.2Common magnetic-link-based medium-voltage converter for wind power generation

systems [8]

Module-1 a b c n
A
B H-bridge H-bridge H-bridge
P inverter
A B inverter C inverter
with rectifier with rectifier with rectifier
C
HF magnetic-link Phase-A Phase-B Phase-C
Module-2
Generator A
+ H-bridge
Power B H-bridge H-bridge
HF P inverter
G Frequency
Rectifier Cd inverter with rectifier with rectifier with rectifier
C
Module-N
A
P inverter
C
Fig.4.3Cascaded high-frequency magnetic-link-based medium-voltage converter for wind

power generation systems
Figure 4.3 shows the basic block diagram of cascaded high-frequency magnetic-
link-based MMC converter. The cascaded magnetic links give freedom to use mul-
tiple low-power magnetic cores other than a single high-power core, which helps to
increase the modularity and reduce the leakage inductances. Still this topology uses
single high-frequency inverter to energize all high-frequency magnetic links, whose
implementation is really critical when the power level is high.
In order to keep the modularity of whole power conversion system, the high-
frequency magnetic link as well as high-frequency inverter should also be mod-
ular. Each identical magnetic link can be excited by identical high-frequency
inverter. All high-frequency inverters may be energized by the same power source,
which helps generate equal voltage at the secondary terminals and ensure bal-
anced DC supplies for the MMC converter. In this topology, each module consists
of a high-frequency inverter, a four windings magnetic link, and three H-bridge
inverters with high-frequency rectifiers. Each module may provide three-level
line-to-line voltage. Figure4.4 shows a modular converter for medium-voltage
grid connection of wind turbine generators. With this topology, it is very simple to
change the converter power level by addition or subtraction of modules.
In power frequency transformers, the most widely used magnetic materi-
als are silicon-steel sheets, nickel-steel sheets, and nanocrystalline ribbons. With
higher excitation frequencies, the skin and proximity effects, hysteresis losses,
and dielectric losses will increase significantly. Therefore, various better magnetic
materials with high saturation flux densities and low specific power losses have
been developed in the last few decades. Among these materials, the amorphous
4.1Introduction 113

Module-1 a b c n
A
HF
P A inverter B inverter C inverter
inverter
C
Module-2
Generator A
+ B H-bridge H-bridge H-bridge
HF
G inverter
Cd with rectifier with rectifier with rectifier
C
Rectifier
Module-N
A
HF
inverter
C
Fig.4.4Modular medium-voltage converter for wind power generation systems
and nanocrystalline are the most commonly used core materials in recent high-
frequency applications. This reduced weight and volume medium-frequency
transformer can be integrated in the MMC converter structure to overcome all the
existing limitations, such as the requirement for multiple isolated DC supplies,
balancing of DC-link voltages, grid isolation, and common mode issues [9].
This chapter presents the development and characterization of a high-frequency
magnetic link with multiple secondary windings to couple the renewable energy
source to the MMC converter. The high-frequency magnetic link is used to gener-
ate the isolated balanced multiple DC supplies for all of the H-bridge inverter cells
of the MMC converter from a single low-voltage power source. Compared with
the conventional power frequency transformers, the high-frequency magnetic links
have much smaller and lighter magnetic cores and windings, and thus much lower
costs. The high-frequency magnetic-link-based multilevel converter systems will
have the following advantages: (i) capable of generating multiple isolated sources
from a single or multiple sources, (ii) inherent DC-link voltage balancing, (iii)
inherent grid isolation capability, and (iv) a wide range of MPPT operation.
4.2Design of High-Frequency Magnetic Links
In the transformer design, the winding electromotive force (emf) is proportional

to the number of turns, frequency, and the magnetic flux linked to the winding, or
e(emf) Nf (4.1)
where N is the number of turns, f the frequency of excitation current, and the
magnetic flux linking to the winding. For a given power capacity, as the operat-
ing frequency increases, the required cross-sectional area of the magnetic core
and the number of turns of the primary and secondary windings can be dramati-
cally reduced. The following sections present the design and analysis of two
high-frequency magnetic links of single primary and six secondary windings.
The magnetic links can be used to generate six isolated balanced DC supplies of a
three-phase 5-level MMC converter. The concept can be used to model other mag-
netic links for other power converter topologies by only changing the number of
primary and secondary windings according to power converter requirements.
4.2.1Flowchart of High-Frequency Magnetic-Link Design
Taking into account the operating frequency of commercially available power semi-
conductor devices, one can find a frequency window in the range of 120kHz for
the proposed system. According to the power inverter rating, the magnetic-link
specifications, such as the rated power, frequency, excitation current, and voltage,
can be calculated. From the specifications of magnetic link and the data sheets of
core materials, magnetic-link initial parameters can be calculated. These parameters
can be used as the initial values of optimization process. The volume and weight
of the transformer were optimized by selecting proper parameters. Different fac-
tors are considered during the design process, such as the winding dimensions, hole
reserve for natural cooling, maximum temperature limits, maximum power loss,
availability of core material stripe dimensions (standard available widths in the range
of 2.550mm and thickness of 20m), parasitic parameters, skin and proximity
effects, and possibility to induce identical voltage in multiple secondary windings.
Single-layer windings have low AC/DC resistance ratios, which would increase
significantly the winding and core dimensions. For simplicity of the winding pro-
cess, a toroidal core structure is adopted. Finally, the inner and outer diameters,
and height of the core are chosen as 6.5, 10.5, 25mm, respectively. The flowchart
of the proposed methodology for the multiple secondary windings high-frequency
magnetic-link design is depicted in Fig.4.5. As shown, the design process involves
multiphysics problems with some critical decision making tasks.
4.2.2Core Material Selection
Along with the advent of new power semiconductor devices, different soft magnetic
materials are conceived with a high magnetic saturation and low core loss to reduce
the weight and volume of conventional power transformers. The grain-oriented sili-
con-steel sheet, which are commonly used as the core material for power frequency
transformers, are not suitable for medium-frequency applications because of the
4.2 Design of High-Frequency Magnetic Links 115
Fig.4.5Flowchart of high- Start

frequency magnetic-link
design process [10]
Initialization
Calculation of
number of turns
Calculation of
winding dimensions
Calculation of core
dimensions
Calculation of
losses
Considerations of
temperature and
parasitic effect
No
Optimized?
Yes
Optimized
parameters
Stop
heavy eddy current loss [11]. The soft ferrites have been widely used in medium-
and high-frequency inductors and transformers due to their low price and general
availability. Because of the low saturation flux density [11] (only 0.30.5T), which
results in the transformers large size, they are not suitable for large power applica-
tions. The amorphous alloy and nanocrystalline magnetic materials have excellent
magnetic characteristics, such as high permeability, high saturation flux density, and
relatively low core losses at medium-/high-frequency range [1214]. Metglas is a
commercially available amorphous material manufactured by Hitachi Metals, Japan.
Figure 4.6 shows a photograph of Metglas stripe. The Metglas magnetic alloys
2705M and 2714A are cobalt-based materials with maximum flux density of 0.77
and 0.57T, and the specific core losses are about 6W/kg and 3W/kg, respectively,
at 10kHz sinusoidal excitation of 0.3T. The alloy 2,826MB is ironnickel-based
material with maximum flux density of 0.88T and specific core loss of 30W/kg at
Fig.4.6Metglas amorphous alloy 2605S3A stripe of 20m thickness and 25mm width
Table4.1Properties of Metglas alloys [15]

Parameters 2605S3A 2705M 2826MB 2714A 2605SA1
Maximum flux density (T) 1.41 0.77 0.88 0.57 1.56
Core loss at 10kHz, 0.3 T (W/kg) 7 6 30 3 20
Basic material Iron Cobalt Iron nickel Cobalt Iron
Resistivity (cm) 138 136 138 142 130
Curie temperature (oC) 358 365 353 225 395
Magnetostriction (ppm) 20 <0.5 12 <0.5 27
Mass density (g/cm3) 7.29 7.80 7.90 7.59 7.18
Ribbon thickness (m) 18 22 29 15 25
Minimum ribbon width (mm) 2.5 2.5 2.5 2.5 5
Maximum ribbon width (mm) 50 50 12.7 50 213
10-kHz sinusoidal excitation of 0.3T. The Metglas alloys 2605SA1 and 2605S3A
are iron-based material with maximum flux density of 1.56 and 1.41T, and the spe-
cific core losses are about 20 and 7W/kg, respectively, at 10-kHz sinusoidal excita-
tion of 0.3T. Table4.1 summarizes the properties of five different Metglas alloys,
which are commercially available.
There are a few vendors providing commercially manufactured nanocrys-
talline magnetic materials, such as Hitachi Metals, Vacuumschmelze, MH&W
International, and AMMET, etc. Commercially available nanocrystalline mag-
netic materials are Finemet, Vitroperm, and Nanoperm, which are manufactured
by Hitachi Metals, Vacuumschmelze, and MH&W International, respectively.
Figure 4.7 shows a photograph of Vitroperm nanocrystalline stripe. Table4.2
summarizes the properties of nanocrystalline materials. The precursor material
of Finemet is amorphous metal obtained by rapid quenching the molten metal,
Fig.4.7Compound
VITROPERM stripe [16]
Table4.2Properties of nanocrystalline magnetic materials

Parameters Finemet Vitroperm Nanoperm
Maximum flux density (T) 1.23 1.20 1.20
Core loss at 100kHz, 0.3T (W/kg) 85 80 110
Basic material Fe, Si, B Fe, Cu, Nb, Si, B Fe, Cu, Nb,
Si, B
Resistivity (cm) 120 115 115
Curie temperature (oC) 570 600 600
Magnetostriction (ppm) <1 <0.5 <0.5
Mass density (g/cm3) 7.30 7.35 7.35
Ribbon thickness (m) 18 18 17
consisting of Fe, Si, B, and small amounts of Cu and Nb. By applying heat treat-
ment to the alloy at higher temperature than its crystallization temperature, this
alloy forms nanocrystalline structure (grain size of approximately 10nm).
Although the nanocrystalline material has lower specific core loss than
Metglas, its saturation flux density (about 1T) is much lower than that of Metglas,
which is up to 1.56T. Taking into account the flux density, specific core loss, cost,
and availability, the Metglas alloys 2605S3A and 2605SA1 stripe of 20 and 30m
thickness, respectively, have been chosen as the core materials. In order to develop
test cores, the core material Metglas alloys 2605S3A and 2605SA1 ribbons were
collected from Metglas Inc.
The coefficients, k, m, and n in Steinmetz Eq.(4.2) deduced from the data-
sheet for this material are experimental results under sinusoidal voltage excita-
tion, where f is the frequency in kHz and B the magnitude of flux density in T.
Table 4.3 summarizes the datasheet coefficients of Steinmetz equation. Usually,
the high-frequency magnetic links are operated under non-sinusoidal voltage exci-
tations, i.e., square-wave voltage [17]. Therefore, for this design, new coefficients
Table4.3Coefficients of Excitation voltage type Coefficients

Steinmetz equation with
k m n
sinusoidal wave excitation of
Metglas alloy 2605SA1 Sinusoidal (from datasheet) 6.50 1.51 1.74
Table4.4Coefficients of Core material Coefficients

Steinmetz equation with
k m n
square-wave high-frequency
excitation 2605S3A 6.571 1.422 1.999
2605SA1 6.256 1.580 1.619
are calculated by measurements under square-wave voltage excitations. The newly

derived coefficients and datasheet coefficients are listed in Table4.4.

Pcore = k f m Bn (4.2)
Using these two sets of coefficients, the specific core losses of Metglas alloy
2605SA1 have been compared. About 2030% extra loss is observed due to non-
sinusoidal excitation waveform. Figure4.8 shows the specific core loss of Metglas
alloy 2605SA1 under square-wave voltage excitation. Compared with Metglas
alloy 2605SA1, 2605S3A gives much lower core loss. Figure4.9 plots the specific
core loss of Metglas alloy 2605S3A.
4.2.3Calculation of Number of Turns
Coils of high-frequency magnetic links are usually wound with Litz wires. High
number of turns may increase the volume, weight, cost, and winding complexity of
the high-frequency magnetic links. On the other hand, sufficient number of turns is
10000
Metglas 2605SA1
11 kHz
16 kHz
1000
21 kHz
Core loss (W/kg)
26 kHz
10
0
10
6 kHz
1 1 kHz
0.1
0.1 0.2 0.3 0.4 0.5 1 2
Flux density (T)
Fig.4.8Specific core loss in terms of flux density and frequency of Metglas alloy 2605SA1
under square-wave excitation
10000
Metglas 2605S3A 11 kHz
1000 16 kHz
Core loss (W/kg) 21 kHz
100 26 kHz
10
1
6 kHz
1 kHz
0.1
0.01
0.1 0.2 0.3 0.4 0.5 1 2
Flux density (T)
Fig.4.9Specific core loss in terms of flux density and frequency of Metglas alloy 2605S3A
under square-wave excitation
required for proper electromagnetic inductions. Therefore, selection of an optimal

number of turns is important for the design of high-frequency magnetic links.
By Faradays law, the voltage, v, and flux, , of a transformer can be related by
d
v(t) = N (4.3)
dt
where N is the number of turns. When the transformer works with a square-wave
voltage, according to (4.3), a triangular flux is required to generate the square-
wave voltage as shown in Fig.4.10, where T is the period of excitation voltage,
Vmax the maximum excitation voltage, and max the maximum flux.
The triangular flux can be expressed mathematically as
max T
T
T /4 t 4 0 t 2
(t) = max . (4.4)
T /4 t 3T
4
T
2 t T
The expression of the voltage can be deduced by substituting (4.4) into (4.3) as

T
N Tmax
/4 0 t 2
v(t) = . (4.5)
N Tmax T
/4 2 t T
Fig.4.10Voltage and flux in V

a square-wave transformer Vmax
t
max
T
Let
max
Vmax = N (4.6)
T /4
and (4.5) can be rewritten as

Vmax 0 t T2
v(t) = . (4.7)
Vmax T2 t T
The rms value of the excitation voltage can be calculated as

1 T T
Vrms = 2
Vmax + (Vmax )2 (4.8)
T 2 2

2 T
Vrms = 2
Vmax = Vmax (4.9)
T 2
Vrms = Vmax (4.10)

If f is the frequency of the excitation voltage, Bmax the maximum flux density,
and A the cross-sectional area of the transformer core, from (4.6) and (4.10), the
expression of the number of turns can be deduced as
max
Vrms = Vmax = N (4.11)
T /4
Vrms = 4fNmax (4.12)
Vrms = 4fNABmax (4.13)

Since max = Bmax A, (4.12) can be rewritten as
Vrms
N = . (4.14)
4fABmax
If the transformer excitation frequency is 10kHz, core size is 5cm2 (2.52cm),
and the flux density is 1T, the minimum number of turns required by the primary
winding of the transformer can be calculated as
Vrms(pri) 210
Np = = = 10.50
4fABmax 4 10 103 5 104 1
and the required minimum number of turns of each secondary winding as
Vrms(sen) 374
Ns = = = 18.70.
4fABmax 4 10 103 5 104 1
Finally, the designs are considered with 14 turns for primary windings and 25
turns for each secondary winding.
4.2.4Winding Wire Selection
The current density within the conductors becomes highly uneven as the excita-
tion frequency increases, due to mainly the skin and proximity effects. With the
increase of excitation frequency, the current density will reduce inside the con-
ductor and increase near the surface. This is called the skin effect. As a result,
although the total amount of current within the conductor is not affected, the cur-
rent density distribution across the conductor becomes non-uniform, and the AC
resistance significantly higher than the DC resistance. The skin effect can be eval-
uated by the skin depth, which is defined as the radial distance from the surface of
the conductor to where the value of current density is 37% smaller than its value
at the surface [18]. The relationship between the skin depth and the operating fre-
quency can be deduced as

= . (4.15)
f
where is the permeability of the material, the resistivity, and f the frequency.
Figure4.11 depicts the skin depth versus frequency of a copper conductor.
If Rdc is the DC resistance of a winding carrying a current of I, the skin effect
loss in a Litz winding can be calculated by [18]

Pskin = Rdc I 2 1 ( ). (4.16)
4
where
d
= (4.17)
2
and
ber( )bei ( ) bei( )ber ( )
1 ( ) = (4.18)
ber ( )2 + bei ( )2
where

ber( ) = Re J( ej3/4 ) , (4.19)
Fig.4.11Skin depth versus

excitation frequency of a 2
copper conductor
Skin depth (mm)
1.5
0.5
0 5 10 15 20
Frequency (kHz)

bei( ) = Im J( ej3/4 ) , (4.20)
ber( ) + bei( )
ber ( ) = (4.21)
2
and
ber( ) + bei( )
bei( ) = (4.22)
2
Figure4.12 shows the skin effect loss in a round conductor for different penetra-
tion ratios, d/.
On the other hand, the AC current in a wire also generates a magnetic field that
enters the adjacent conductors and induces emfs inside these conductors, resulting
in non-uniform current distribution in the conductors. This is called the proximity
effect, which highly depends on the excitation frequency [18]. Although the total
current of the conductor does not change; the current density in the conductor will
be reduced near the adjacent wire and reinforced on the opposite side through a
redistribution of current density. Proximity effects can be classified into two types,
internal proximity effectsthe effects on each of the strands due to the field gen-
erated by the strands, and external proximity effectsthe effects on an isolated
round conductor within an external field. If ls is the total length of a single strand,
ns the total number of strands, rb the radius of the bundle, and the conductivity,
the power loss due to the internal proximity effects can be calculated by [18]
n s I 2 ls
Pp_int = 2 ( ). (4.23)
4 rb2
where
ber2 ( )ber ( ) + bei2 ( )bei ( )
2 ( ) = . (4.24)
ber ( )2 + bei ( )2
The power losses due to an external magnetic field, He, can be calculated by [18]
Fig.4.12Power losses 1.8

due to skin effect in a round
Skin effect loss (W/m)
conductor for different 1.6

penetration ratios
1.4
1.2
1
0 1 2 3 4 5
Penetration ratio
Fig.4.13Power losses due
Proximity effect loss (W/m)

to proximity effects in a 3
round conductor for different
penetration ratios
2
tsc
fe
ef
al
rn
ts
1
te
c
fe
Ex
ef
al
rn
te
In
0
0 0.5 1 1.5 2 2.5
Penetration ratio
2
Pp_ext = ns He2 2 ( ). (4.25)

Figure4.13 shows the power loss in a round conductor due to proximity effects.
The skin and proximity effects will increase the AC losses in high-frequency
windings. A special type of wire has been conceived named the Litz wire, a con-
ductor consisting of insulated strands twisted or braided together. Such a design
equalizes the flux linkages of individual strands causing the current to spread
uniformly throughout the conductor, and the AC to DC resistance ratio tends to
approach unity. In a conductor, the AC/DC resistance ratios, Kr, depends strongly
on the number of layers, U, the conductor diameter, d, and the operating fre-
quency. The AC resistance of the U-th layer for solid round wires can be repre-
sented as [19]
Rac
Kr = = M( ) + (2U 1)2 D( ) . (4.26)
Rdc 2
where
0.886d
= (4.27)

sinh( ) + sin( )
M( ) = (4.28)
cosh( ) cos( )
sinh( ) sin( )
D( ) = (4.29)
cosh( ) + cos( )
and
0.071
= (4.30)
f
For a particular core shape and defined fill factor, there should be an optimal selec-
tion of the Litz wire strand number and diameter to realize the minimum loss.
Figure 4.14 shows the AC/DC resistance ratio of a U-layer winding at a 10-kHz
19
7
U=
U=
1.8
AC/DC resistance ratios
10
4
U=16
U=
U=
1.6
U=13
U is the number of
1.4
layers
1.2
U=1
1
0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Diameter (mm)
Fig.4.14Graphical representation of the AC/DC resistance ratio Kr, at 10kHz, showing the
effect of conductor diameter and number of layers on AC resistance
excitation current. It shows that at medium or high frequencies, associated with a

small skin depth and proximity effect, the number of layers as well as the conduc-
tor diameter should be kept as small as possible. Moreover, the insulated strands
should be twisted or braided together to equalize the flux linkages throughout the
conductors. To achieve this and so as to reduce the winding loss, a winding of
Litz wire with small number of layers should be always used in a medium-/high-
frequency link.
The size of transformer-winding copper wire depends on the primary and sec-
ondary side currents, ip and is, e.g., 6 and 1A, respectively. The current density,
J, is chosen as 4A/mm2, appropriate for a frequency of 10kHz. The number of
strands of the Litz wires for the primary and secondary windings can be chosen as
13 and 3, respectively. The required minimum cross-sectional area of the primary
winding single strand can be calculated as
ip
ap = 13
= 0.115 mm2 (4.31)
J
and single-strand diameter can be calculated as

4ap
dp = = 0.383 mm. (4.32)

Finally, for the primary windings, the Litz wires with a diameter of 0.4mm are con-
sidered. Theoretically, the overall area of 13 strand wire is 130.115=1.495mm2.
In practical, when 13 insulated strands are twisted or braided together, the overall
diameter of the Litz wire is 2mm and the cross-sectional area of the primary wind-
ing wire 3.2mm2. Also, the required minimum cross-sectional area of the secondary
winding single strand can be calculated as
is
as = 3
= 0.083 mm2 (4.33)
J
and the single-strand diameter can be calculated as

4as
ds = = 0.325 mm. (4.34)

Finally, for the secondary windings, the Litz wires with a diameter of 0.4mm are con-
sidered. Theoretically, the overall area of 3 strand wire is 30.083=0.249mm2. In
practical, when 3 insulated strands are twisted or braided together, the overall diame-
ter of the Litz wire is 1mm and the cross-sectional area of the primary winding wire
0.79mm2. Hence, the design can be considered with the wire cross-sectional areas of 4
and 1.2mm2 for the primary and secondary winding, respectively. The area required by
the primary and secondary windings is calculated as
Aw = Np dp + Ns ds (4.35)
Considering a toroid hole reserve factor of 8 for all the windings, the minimum
required toroid hole area can be calculated by
ATH = 8(Np dp + Ns ds ) (4.36)
4.2.5Parasitic Calculation
The leakage flux and stray capacitance may affect the performance of a high-
frequency transformer as well as the operation of the PWM converter. These two
components, leakage inductances and stray capacitances, are usually called para-
sitic parameters.
4.2.5.1Stray Capacitances
The stray capacitances of a medium-frequency transformer due to the electrical

coupling between any two conducting elements are highly structure dependent and
distributed in nature. It is hard to model the effects of stray capacitances accu-
rately because of their distributive nature, i.e., they can be formed between two
turns in the same winding or two different windings, two layers in the same wind-
ing or two different windings, windings and core, and windings and ground, etc.
Since the stray capacitances can significantly affect the performance of the trans-
former as well as the power electronic converters, the consideration of stray capac-
itances is important for the design of a medium-frequency transformer.
The proposed multiwinding transformer can be approximately represented by the
two-winding transformer model. For medium-frequency applications, the lumped
models of stray capacitances are sufficient and have been generally employed. The
Fig.4.15A two-winding Cpso

transformer with stray
capacitances Rp Rs
Vp Cpo Cso Vs
Np:Ns
Fig.4.16Equivalent circuit C ps
of a two-winding transformer
with stray capacitances; all L lp L ls
Rp Rs
parameters referred to the
primary side [22] Vp Cp Rm Ep Vs
Cs
-shaped network of three lumped capacitances approach [20, 21] gives a clear
view of the coupling capacitances using the concepts of self- and mutual capaci-
tances. This approach represents the overall effects of all stray capacitances by three
lumped capacitors, such as the self-capacitances of the primary and secondary wind-
ings, and the mutual capacitance between the primary and secondary windings [22].
Figure 4.15 illustrates the circuit of a two-winding transformer with stray capaci-
tance effects, where Rp and Rs are the resistances of the primary and secondary
windings and Cpo, Cso, and Cpso are the self-capacitances of the primary and second-
ary windings, and the mutual capacitance between two windings, respectively.
If Llp is the leakage inductance of the primary winding, Lls the equivalent leak-
age inductance of secondary winding referred to the primary side, Ep the mag-
netization back emf, and Rm and Lm are the equivalent core loss resistance and
magnetization inductance, respectively, the equivalent circuit of a two-winding
transformer referred to the primary side with stray capacitance effects can be rep-
resented by a network as shown in Fig.4.16 [22].
The equivalent capacitances, Cp, Cs, and Cps, can be calculated from

Ns
Cp = Cpo + 1 Cpso (4.37)
Np
2
Ns Ns Ns
Cs = Cso + 1 Cpso (4.38)
Np Np Np
and
Ns
Cps = Cpso . (4.39)
Np
If Rp and Rs can be ignored and the primary and secondary windings are short
circuited, it is possible to measure Cpso approximately as the capacitance between
the primary and secondary. In order to calculate the Cpso, an external inductor with
known inductance value, Lex can be used to make a resonance circuit and then
from two resonance frequencies, f1 and f2, the Cpso can be calculated from
1 1
Cpso = . (4.40)
(2f2 )2 Lex (2f1 )2 Lex
Using (4.39) and (4.40), Cps can be calculated.The approximate values of capaci-
tances Cp and Cs can be calculated from the short-circuit test of Fig.4.15 with the
following relations
Cp = Ceqp Cps (4.41)
and
Cs = Ceqs Cps . (4.42)
where the capacitances Ceqp and Ceqs are the approximated equivalent capaci-
tances seen from the primary and secondary sides with the other side short cir-
cuited, which makes the circuit just like a single-coil inductor. The equivalent
capacitances Ceqp and Ceqs can be calculated from the step response of the circuit
under the condition of one-side short circuited. If Vc is the capacitor-charging
voltage and Ic is the average capacitor current over a time period, t, the stray
capacitance of the coil inductor can be calculated from
tIc
Ceq = . (4.43)
Vc
4.2.5.2Leakage Inductances
Usually, the leakage inductance is defined as the ratio of the leakage flux linking a
winding, l, and the current passing through the winding, i, which is appropriate in
the case of linear magnetic cores. In order to develop the dynamic circuit model of
a high-frequency transformer, differential leakage inductance may be used due to
its non-linear magnetic properties, which can be defined as [22]
dl
Ldl = i. (4.44)
d
The differential leakage inductances of the primary and secondary windings can
be defined by
dlp (ip )
Ldlp = . (4.45)
dip
and
dls (is )
Ldls = . (4.46)
dis
If imp and ims are the magnetization currents seen from the primary and second-
ary sides, respectively, the mutual inductances between the primary and secondary
windings can be defined by
dmp dm
Ldmp = = Np . (4.47)
dimp dimp
and
dms dm
Ldms = = Ns . (4.48)
dims dims
where m is the mutual flux.
An equivalent circuit of a two-winding medium-frequency transformer is
shown in Fig.4.17 [22].
If p (im ) = Np m (im ) + Np lp (ip )is the total flux linkage of the primary wind-
ing and s (im ) = Ns m (im ) Ns ls (is ) the total flux linkage of the secondary
winding, and lp and ls are the leakage flux of the primary and secondary wind-
ings, respectively, the differential self-inductances of the primary and secondary
windings can be calculated as
p
Ldp = . (4.49)
ip
and
s
Lds = . (4.50)
is
For self- and mutual inductances the leakage inductances of the primary and sec-
ondary windings can be calculated as
Ldlp = Ldp Ldmp (4.51)
and
Ldls = Lds Ldms . (4.52)
4.2.6Design Optimization
A frequency window in the range of 120kHz is considered in the optimization.

The differential evolution algorithm (DEA) can be used as the optimization algo-
rithm. DEA was first presented in 1997 as a population-based stochastic global
optimization method [23]. As demonstrated by many studies, it converges fast
and is robust and simple to implement, requiring only a few control parameters.
Fig.4.17Equivalent circuit Ldlp Ldls

Rp ip i s Rs
of a two-winding transformer
with leakage inductances; im
Vp Rm dim Vs
all parameters referred to the Ldmp
dt
primary side [22]
It has been widely employed for optimization of electromagnetic devices and has
achieved many improvements [24, 25].
The procedure of DEA, very similar to that of the genetic algorithm, consists of
three processes, namely mutation, crossover, and selection. It starts by initializing
the population randomly in the design space. The individuals in the population are
then perturbed with others through mutation and crossover operators, and a new
population consisting of the most promising solution can be generated by applying
a selection criterion [23, 24]. Different factors are considered during the optimi-
zation, such as the winding dimensions, hole reserve for natural cooling, maxi-
mum temperature limits, maximum power loss, availability of core material stripe
dimensions (available sizes of alloys 2605S3A and 2605SA1: widths of 2.550
and 5213mm, respectively, and thickness of 20m), parasitic parameters, skin
and proximity effects, and possibility to induce identical voltage in multiple sec-
ondary windings. Power loss, size, and weight of the high-frequency magnetic link
have been considered the main objective functions of optimization process.
In the implementation, the algorithm parameters of DEA are mutation factor of
0.8, crossover factor of 0.8, the maximum number of iteration of 1,000, and the maxi-
mum stall generation of 100 (as the stop criterion) [25]. The optimal parameters of
2.5-kVA magnetic links are summarized in Table4.5. The designs show that magnetic
alloy 2605S3A has a slightly larger volume and weight but much lower specific core
loss than that of alloy 2605SA1 and is suitable for high-frequency magnetic links.
Ansoft Maxwell 3D solves the electromagnetic field problems for a given
model with appropriate materials, boundaries, and source conditions, applying
Maxwells equations over a finite region of space. There are two types of solutions
available in Maxwell 3D, such as magnetic fields and electric fields. Category
magnetic fields cover three solvers, such as magnetostatic to calculate static mag-
netic fields, forces, torques, and inductances caused by DC currents, eddy current
to calculate sinusoidally varying magnetic fields, forces, torques, and impedances
caused by AC currents and oscillating external magnetic fields, and transient mag-
netic to calculate transient magnetic fields caused by time-varying or moving elec-
trical sources and permanent magnets. Electric fields also cover three solvers, such
as electrostatic to calculate static electric fields, forces, torques, and capacitances
caused by voltage distributions and charges, DC conduction to calculate voltage,
electric field, and the current density calculated from the potential, and transient
electric to calculate transient electric fields caused by time-varying voltages,
charge distributions, or current excitations in inhomogeneous materials.
In the Ansoft Maxwell 3D environment, two magnetic cores are mod-
eled using the parameters obtained from optimization and non-sinusoidal
Table4.5Optimal parameters for 2.5-kVA magnetic links [26]

Core material Parameters
WD (cm) HT (cm) B (T) f (kHz) Vol. (cm3) Wt. (kg) Pcore (W/kg)
2605S3A 2.22 2.78 1 6.05 169 1.23 71
2605SA1 2.29 2.61 1 6.25 165 1.18 96
high-frequency characteristics of Metglas alloys 2605S3A and 2605SA1. When

the Maxwell starts a new project, it is automatically added to the project tree.
By selecting Project>Insert Maxwell Design, a new Maxwell design can be
added to the project. The Maxwell desktop consists of a numbers of functional
icons, and windows, toolbars, etc. Figure4.18 shows a part of Maxwell desktop.
There are two geometry models available in Maxwell 3D, i.e., Cartesian (XY)
model and Axisymmetric (RZ) model to model project in the 3D modeler win-
dow (graphics area).
Using the Modeler Materials toolbar, default materials can be selected for the
project. Electromagnetic properties of new materials can be added through View/
Edit Material window as shown in Fig.4.19. Figure4.20 shows the Maxwell desk-
top windows to create coils and assign excitation current. By selecting the menu
item Show Conduction Path, it is possible to visualize the conduction path in 3D
model. The menu item Mesh Operations has the options to select maximum length
of elements and maximum number of elements for refinement. Figure4.21 shows
the Maxwell desktop to select element size and numbers. An analysis setup can be
created by selecting menu item Add Solution Setup, which provides options to set
adaptive and transient setups, such as number of passes, percent error, stop time,
and time step. Figure4.22 shows the Maxwell desktop to create an analysis setup.
After saving the whole project and checking the validity of the model, it is possi-
ble to start the solution process by selecting menu item Analyze All. By selecting
the menu item Mag_B, one can create a field plot. Figure4.23 shows the mag-
netic field distribution of Metglas alloy 2605S3A with square-wave voltage excita-
tion at 6kHz. Figure4.24 shows the magnetic field distribution of Metglas alloy
2605SA1.
Fig.4.18Maxwell desktop with model

Fig.4.19Maxwell desktop windows to select material or add a new material
Fig.4.20Maxwell desktop windows to create excitation coil
4.2.7Transformer Core Development
Based on the optimization results, the optimal core dimensions are chosen as 6.5-
cm inner diameter (ID), 10.5cm outer diameter (OD), and 2.5cm height (HT) for
the design, as shown in Fig.4.25.
The Metglas sheet was glued with Araldite 2011 on the surface of each layer,
providing both the electrical insulation and mechanical bonding. During the wrap-
ping process, equal and opposite forces were applied to make a uniform distribu-
tion of Araldite. Figure4.26 shows a photograph of the wrapping process.
Fig.4.21Maxwell desktop windows to select element size and numbers
Fig.4.22Maxwell desktop with solve setup window
After wrapping, the frame was quickly removed before the Araldite dried up.
The core volume and mass are 133.52cm3 and 0.96kg, respectively. Figure4.27
shows a photograph of the Metglas core with frame; just after wrapping process.
The dried core is shown in Fig.4.28. The core was covered by transformer
tape. To minimize the proximity effect, Litz wires are used for windings with a
Fig.4.23Magnetic field illustration for Metglas alloy 2605S3A-based high-frequency magnetic

link in the Ansoft Maxwell 3D [26]
Fig.4.24Magnetic field illustration for Metglas alloy 2605SA1-based high-frequency magnetic

link in the Ansoft Maxwell 3D [26]
single-layer placement. The electromagnetic performances were analyzed and

compared with the experimental results in Sect.4.3. Figures4.29 and 4.30 show
the photograph of developed transformers with Metglas alloys 2605SA1 and
2605S3A, respectively.
m
2c
6.5 c
m
ID OD
HT = 2.5 cm ID = 6.5 cm
OD = 10.5 cm HT
Fig.4.25Transformer core dimensions
Fig.4.26Transformer core development setup (raping of Metglas sheet)

Fig.4.27A photograph of the transformer core with frame; just after wrapping
Fig.4.28Photograph of the transformer core with Metglas alloy 2605SA1 ribbon of 30m
thickness and 25mm of width
Fig.4.29Photograph of high-frequency magnetic link with Metglas alloy 2605SA1 core
Fig.4.30Photograph of high-frequency magnetic link with Metglas alloy 2605S3A core
4.3Characterization of High-Frequency Magnetic Links
4.3.1Windings DC Resistance Calculation

The electrical resistance, R, of an element is defined as the ratio of the voltage, V,
across the element and the current, I, through that element, i.e.,
V
R= . (4.53)
I
4.3 Characterization of High-Frequency Magnetic Links 137
For given element, R depends on the resistivity, , of the material (temperature

dependant), length, L, and cross-sectional area, A, i.e.,
L
R= . (4.54)
At 20C, the resistivity of copper is 1.678108 m. The average length of

A
each turn is 0.13m. Since the primary winding has 14 turns and each second-
ary winding has 25 turns, the total winding wire lengths are 1.82 and 3.25m,
respectively. About 0.14 and 0.08m extra wires are required for the end con-
nection of primary and secondary windings, respectively. When the strands are
twisted together, there is an increase in the length of the windings. About 1.14
and 1.09% increases are observed in the length of primary and secondary wind-
ings, respectively. Therefore, the overall length (1.82+0.14)1.14=2.24m
and (3.25+0.08)1.09=3.62m are considered for the primary and second-
ary windings, respectively. The diameters of primary and secondary wind-
ing copper wires are the same (0.4mm), and also the cross-sectional areas
(12.56 108m2). Therefore, the resistance of a single strand at 20C can be
calculated by
2.24
RP = 1.678 108
12.56 108
= 0.30 (for single strand of primary winding) (4.55)
= 0.0230 (for primary winding, 13 strands) and
3.62
RS = 1.678 108
12.56 108
= 0.483 (for single strand of secondary winding) (4.56)
= 0.161 (for each secondary winding, 3 strands)
At the room temperature of 23C (during the test), the resistance can be calcu-
lated by
234.5 + t2
Rt2 = Rt1 (4.57)
234.5 + t1
234.5 + 23
Rt2 = Rt1 = 1.012 Rt1 (4.58)
234.5 + 20
Therefore, 0.0233 and 0.162resistances are considered theoretically for the pri-
mary and secondary windings, respectively.
A digital multimeter (DMM) was used to get an idea about the range of resist-
ances. The DMM readings are tabulated in Table4.6.
The Wheatstone bridge is also suitable for measuring a small unknown electrical
resistance by balancing two legs of a bridge circuit. Figure4.31 shows the Wheatstone
bridge circuit diagram, and Fig.4.32 describes the measurement technique.
Table4.6DC resistance measurement using DMM

Primary side P () Secondary sides ()
A B C D E F
0.20 0.50 0.50 0.50 0.50 0.50 0.50
R1 R3
G
5V
R2 RX
Fig.4.31Wheatstone bridge circuit diagram
E
F
R1 R3
D
G P
5V
R2 RX C
A
B
Fig.4.32DC resistance measurement technique
A photograph of the experimental setup is shown in Fig.4.33. When the galva-

nometer deflection is zero (i.e., the bridge circuit is balanced), we have the ratio of
R2 RX
= . (4.59)
R1 R3
R2
Therefore, RX = R3 . (4.60)
R1
The Wheatstone bridge is used to measure the resistance of the coils since the coil
resistance is very small. The lead wire resistance is also considered and subtracted
from the measured values, as tabulated in Table4.7. The measured winding resist-
ances are compared with theoretical values and the variation in percentage is also
tabulated in Table4.7.
4.3.2Voltage Transformation Ratio Calculation
If the number of turns of primary coil (P) is NP, the numbers of turns of secondary
coils (A, B, C, D, E, and F) are NA, NB, NC, ND, NE and NF, and the corresponding
induced voltages in the primary and secondary coils are vP, vA, vB, vC, vD, vE, and
Fig.4.33Photograph of DC resistance measurement experimental setup
Table4.7DC resistance measurement using Wheatstone bridge

Primary side Secondary sides
A B C D E F
Total () 0.043 0.191 0.190 0.205 0.185 0.182 0.182
Leading wires 0.019 0.029 0.029 0.029 0.029 0.029 0.029
Winding () 0.024 0.163 0.161 0.176 0.156 0.153 0.153
Error (%) +3 +0.6 0.6 +8.6 3.7 5.5 5.5
vF, respectively, the voltage transformation ratio can be expressed by (the same
number of turns in all secondaries)
NA NB NC ND NE NF
= = = = = (4.61)
NP NP NP NP NP NP
vA vB vC vD vE vF
= = = = = = . (4.62)
vP vP vP vP vP vP
The theoretical turn ratio can then be calculated as
374
= = 1.781.
210
The coil P is excited by a 50-Hz alternating current, iP, and the corresponding second-
ary coil-induced voltages, vP, vA, vB, vC, vD, vE, and vF, obtained from the respec-
tive coils, P, A, B, C, D, E and F, are listed in Tables4.8 and 4.9. The transformation
ratios of all secondary coils, TRA, TRB, TRC, TRD, TRE, and TRF, are calculated with
respect to the primary coil. The measured ratios compared with the calculated values
Table4.8Transformation P-coil A-coil B-coil C-coil

ratio calculations at 50Hz
iP vP vA TRA vB TRB vC TRC
(AC)
1.00 1.16 2.070 1.784 2.068 1.782 2.066 1.781
1.60 1.25 2.226 1.781 2.225 1.780 2.226 1.781
2.25 1.38 2.460 1.782 2.46 1.782 2.45 1.775
3.29 1.63 2.890 1.773 2.890 1.773 2.887 1.771
Avg. ratio 1.780 1.779 1.777
Error (%) 0.05 0.11 0.22
Table4.9Transformation P-coil D-coil E-coil F-coil

ratio calculations at 50Hz
iP vP vD TRD vE TRE vF TRF
(DF)
1.00 1.16 2.067 1.781 2.068 1.782 2.068 1.782
1.60 1.25 2.229 1.783 2.229 1.783 2.227 1.781
2.25 1.38 2.46 1.782 2.45 1.775 2.46 1.782
3.29 1.63 2.889 1.772 2.886 1.770 2.890 1.773
Avg. ratio 1.779 1.777 1.779
Error (%) 0.11 0.22 0.11
are also listed in Tables4.8 and 4.9. The measured ratios fairly coincide with the cal-
culated values, considering the very small change in voltage with the large change
in excitation current since at this frequency, the winding resistances are quite small.
Utility supply with single-phase auto-transformer can be used to calculate the volt-
age transformation ratios at 50-Hz sinusoidal excitation. A Fluke clip-on wattmeter
is used to measure the voltage and current of the coils. Figure4.34 shows the experi-
mental setup to calculate voltage transformation ratios at 50-Hz sinusoidal excitation.
Voltage transformation ratios are also calculated with a 1-kHz excitation cur-
rent. The calculated transformation ratios are compared with the theoretical values
as listed in Tables4.10 and 4.11. The ratios are highly coincident with the theoreti-
cal values and better than power frequency transformation ratios.
The voltage transformation ratios have also been measured at 10kHz and com-
pared to the calculated value (1.781). The ratios and their percentage variation from
the calculated value are summarized in Tables4.12 and 4.13. The high-frequency
transformer generates 6 almost equal and isolated sources (i.e., all secondary wind-
ings show similar voltage transformation characteristics). Such a similarity of char-
acteristics is obligatory to generate balanced multiple sources for the MMC system.
4.3.3Measurement of Losses
High-frequency performance has been measured and appraised in the laboratory.

The high-frequency (e.g., 112kHz) square-wave signal is generated by an IGBT-
based H-bridge inverter supplied by a controlled 210-V DC voltage source. The
Fig.4.34Photograph of experimental setup to calculate voltage transformation ratios at 50-Hz

sinusoidal excitation

ratio calculations at 1kHz
vP vA TRA vB TRB vC TRC
(AC)
14.20 25.31 1.782 25.30 1.781 25.28 1.780
16.00 28.51 1.781 28.50 1.781 28.48 1.780
16.82 29.97 1.781 29.95 1.780 29.93 1.779
Avg. ratio 1.7813 1.7806 1.7797
Error (%) +0.016 0.022 0.073

vP vD TRD vE TRE vF TRF
(DF)
14.20 25.29 1.781 25.30 1.781 25.30 1.781
16.00 28.48 1.780 28.50 1.781 28.52 1.782
16.82 29.93 1.779 29.95 1.780 29.96 1.781
Avg. ratio 1.7800 1.7806 1.7813
Error (%) 0.056 0.022 +0.016
Tektronix DPO 2024 Digital Phosphor Oscilloscope with a P5200 high-voltage

differential probe and the Tektronix TCPA300 current probe are used to observe
the waveforms. The total loss (core loss plus copper loss) is measured by a Voltech

vP vA TRA vB TRB vC TRC
(AC)
68.2 121 1.774 121 1.774 122 1.788
104 186 1.788 187 1.798 187 1.798
207 367 1.773 366 1.768 365 1.763
Avg. ratio 1.778 1.780 1.783
Variation (%) 0.16 0.05 +0.11

vP vD TRD vE TRE vF TRF
(DF)
68.2 121 1.774 121 1.774 122 1.788
104 186 1.788 187 1.798 186 1.788
207 366 1.768 367 1.773 368 1.777
Avg. ratio 1.776 1.781 1.784
Variation (%) 0.28 0.00 +0.16
PW3000A universal power analyzer. Figure4.35 shows a photograph of the exper-

imental setup for high-frequency non-sinusoidal excitation. Figure4.36 shows the
measured primary and secondary side voltage waveforms. The secondary winding
current waveform is shown in Fig.4.37.
All secondary windings are excited one by one with a square-wave excitation
current at different frequencies ranging from 4 to 12kHz. Table4.14 summarizes
the measured losses of all secondary windings. From the oscilloscope data, the
Fig.4.35A photograph of the experimental setup for characterization of high-frequency mag-

netic links
V B (374 V)
V p (210 V)
10 kHz excitation current
Fig.4.36Measured primary and secondary side voltage waveforms
Voltage Current
Voltage and current of a winding

Fig.4.37Measured voltage and current waveforms
calculated by using the DC resistances (0.024 for the primary and 0.16 for
copper loss of each winding at a different frequency ranging from 4 to 12kHz was
the secondary), since the AC/DC resistance ratios, Kr, in this design are almost
unity due to the use of the Litz wires. Figure4.38 shows that the total losses of all
secondary windings measured at different excitation frequencies ranging from 4 to
12kHz are almost the same. Measurements were also conducted by exciting the
secondary windings one by one with different excitation currents (0.10.6 A) and
a fixed frequency of 10kHz. Table4.15 summarizes the losses of all six second-
ary windings with different values of excitation current. As shown in Fig.4.39, at
10kHz, all secondary windings also have very similar total losses. Such a similar-
ity of characteristics is also obligatory to generate balanced multiple sources for
the MMC system. The proper balancing of all sources means the elimination of
the extra control circuit with the switching algorithm.
The core loss against flux density was measured within the frequency rang-
ing from 6 to 12kHz. The specific core loss of Metglas alloy 2605SA1 measured
was 157W/kg under 10-kHz square-wave excitation of magnitude 0.8T as shown
in Fig.4.40, while the specific core loss of 125W/kg is reported by the material
Table4.14Total loss Coil A excited Coil B excited

calculations with respect to
f (kHz) v (V) P (W) f (kHz) v (V) P (W)
each secondary coil with
0.35-A excitation current 4.015 111.08 30.24 4.015 109.81 29.91
6.045 155.23 46.39 6.045 155.10 46.61
8.068 197.82 62.34 8.068 200.10 62.53
10.035 230.40 74.20 10.035 230.00 74.47
12.095 247.81 81.42 12.095 246.70 80.60
Coil C excited Coil D excited

4.015 110.60 30.45 4.015 109.83 29.91
6.045 156.38 46.67 6.045 156.87 46.67
8.068 199.77 62.75 8.068 200.30 63.08
10.035 229.90 75.14 10.035 228.30 73.50
12.095 248.80 81.70 12.095 246.90 81.06
Coil E excited Coil F excited

4.015 110.03 30.11 4.015 110.56 30.38
6.045 156.30 46.20 6.045 156.26 46.19
8.068 199.14 62.29 8.068 198.25 61.74
10.035 230.18 75.64 10.035 230.80 74.96
12.095 248.40 81.87 12.095 247.40 81.52
Fig.4.38Measured total
losses of Metglas alloy 80
2605SA1 against excitation 70 A
Loss (W)
frequencies at excitation
current of 0.35A; only one 60 B
coil is energized at a time C
50
while the others are open D
circuited 40 E
F
30
4 5 6 7 8 9 10 11 12
Frequency (kHz)
manufacturer under 10-kHz sinusoidal excitation of magnitude 0.8T. In compari-

son with the material manufacturers data, about 2030% extra loss is observed
due to the non-sinusoidal excitation waveform.
4.3.4Measurement of BH Characteristics
The magnetic field intensity and magnetic flux density are calculated by measuring
the P-coil excitation voltage and current and the open-circuit terminal voltage of the
secondary side D-coil. As the coils are uniformly wound on the toroidal core, the
Table4.15Total loss Coil A excited Coil B excited

calculations with respect
i (A) v (V) W (W) i (A) v (V) W (W)
to each secondary coil
at 10-kHz Metglas alloy 0.105 31.01 2.56 0.100 27.81 2.12
2605SA1-based magnetic 0.201 75.05 13.18 0.201 72.30 12.44
link) 0.299 171.19 48.06 0.298 169.74 47.58
0.396 259.9 92.06 0.4 261.50 92.81
0.501 289.4 113.33 0.505 280.91 108.21
0.599 302.3 122.99 0.602 291.30 116.45
Coil C excited Coil D excited

i (A) v (V) W (W) i (A) v (V) W (W)
0.101 29.09 2.28 0.100 27.57 2.13
0.202 75.85 13.34 0.201 74.76 13.10
0.300 173.12 48.98 0.296 168.02 46.89
0.402 261.50 93.01 0.396 258.70 91.42
0.496 286.50 111.45 0.495 287.80 112.05
0.596 299.00 120.97 0.601 303.10 122.57
Coil E excited Coil F excited

i (A) v (V) W (W) i (A) v (V) W (W)
0.101 28.20 2.19 0.101 29.24 2.31
0.201 74.04 12.90 0.201 76.42 13.42
0.302 174.3 49.60 0.298 171.52 48.22
0.399 260.00 92.08 0.398 259.50 91.94
0.505 289.60 113.61 0.499 288.10 112.17
0.601 302.7 122.10 0.602 302.30 122.79
Fig.4.39Measured total
losses of Metglas alloy
2605SA1 against excitation 100
Total loss (W)
currents of 10kHz; only one A

coil is energized at a time B
while the others are open C
50
D
circuited
E
F
0
0.1 0.2 0.3 0.4 0.5 0.6
Excitation current (A)
magnetic field intensity H and magnetic flux density B within the core can be con-
sidered as uniform. By the Amperes law, the field intensity can be calculated as
N1 i(t)
H= . (4.63)
le
where N1 is the number of turns in the primary coil, i(t) the excitation current, and
le the mean length of the core.
Fig.4.40Measured core
12 kHz
losses of Metglas alloy
2 10 kHz
2605SA1 at different 10
Core loss (W/kg)

frequency ranging from 6 to 8 kHz
12kHz; experiments were 6 kHz
carried out at 40C
1
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Flux density (T)
By the Faradays law, the magnetic flux density in the core can be calculated as

1
B= VL dt. (4.64)
N2 Ae
where N2 is number of turns in the pickup coil (D-coil) and Ae the cross-sectional
area of the core, and VL the pickup coil voltage.
Different magnitude of 10-kHz excitation currents (e.g., 13A) are applied to
the primary windings. The BH loops are plotted in Fig.4.41 based on experimen-
tal data. Up to 0.5T, the flux density rises sharply with a constant field intensity
of about 50A/m. At 3A, the maximum magnetic flux density is about 0.8T with
the field intensity of about 600A/m, which satisfies the design requirement. The
BH loops are also measured at different temperatures ranging from 40 to 100C
as depicted in Fig.4.42. The maximum flux density remains approximately con-
stant for this temperature range, as illustrated in Fig.4.43. The FLUKE infrared
temperature probe 80T-IR and FLIR infrared thermal imaging camera i7 were
used to measure the temperature of the high-frequency magnetic links. Figure4.44
shows the thermal image of the high-frequency magnetic link when the probe
was pointed at the surface of the core. Quite lower temperature is observed at the
center of the toroid hole. Figure4.45 shows the thermal image of the high-fre-
quency magnetic link when the probe was pointed at the center of the toroid hole.
Fig.4.41Measured BH
loops of Metglas alloy
2605SA1 at 40C; maximum
magnetic flux density at 1,
2, and 3A is 0.65, 0.74, and
0.80T, respectively
Fig.4.42Measured BH
2605SA1 at different
temperature ranging from 40
to 100C
Fig.4.43Measured
0.8
maximum magnetic flux
Max. flux density (T)
density of Metglas alloy

0.6
2605SA1 at different
temperature ranging from
40 to 120C: maximum 0.4
magnetic flux density at 40,
60, 80, 100, and 100C is 0.2
0.65, 0.65, 0.64, 0.63, and
0.62T, respectively 0
40 50 60 70 80 90 100 110 120
Temperature (Degree centigrade)
Fig.4.44Thermal image of
high-frequency magnetic link
when the probe was pointed
at the surface of the core
Figure4.46 shows the BH loops of Metglas alloy 2605S3A based on experi-

mental data with excitation currents of 3 and 5A. The plotted BH loops have
been compared with the material manufacturers data and found to be highly con-
sistent. The maximum flux density was also measured under different temperature
Fig.4.45Thermal image of
high-frequency magnetic link
when the probe was pointed
at the center of the core
Fig.4.46Measured BH
2605S3A at excitation current
of 3 and 5 A. The link is
excited by 12-kHz square-
wave primary voltages
Fig.4.47Measured
BH loops of Metglas
alloy 2605S3A at various
temperatures. The link is
excited by 12-kHz square-
wave primary voltages
conditions, which is almost stable for the temperature range of 23120C.

Figure 4.47 shows the BH loops of Metglas alloy 2605S3A at various temper-
atures. Figure4.48 plots the BH loops at different frequencies. The BH loops
Fig.4.48Measured BH
2605S3A at different
frequencies
Fig.4.49Measured BH
loops of Metglas alloys
2605SA1 and 2605S3A;
the links are excited by
12-kHz square-wave primary
voltages
of Metglas alloy 2605S3A are compared with that of Metglas alloy 2605SA1, as
shown in Fig.4.49. In comparison with 2605SA1, the 2605S3A shows a slightly
narrower BH loop, i.e., lower core loss.
4.3.5Measurement of DC-Link Voltages
The output of each secondary winding is connected to a fast recovery diode-based

rectifier with a low-pass RC filter circuit. An Agilent Technologies DS06034A oscil-
loscope with a P5200 high-voltage differential probe is used to observe the voltages.
The DC-link voltages were found approximately equal at about 370V, which can
serve satisfactorily as the isolated and balanced DC sources for the proposed MMC
converter. The voltage waveforms of one secondary winding and corresponding rec-
tified DC voltage are shown in Fig.4.50. Without using any special control algo-
rithm regarding the capacitor voltage unbalancing, the prototype converter generates
a satisfactory output-voltage waveform as the transformer can provide balanced
sources for all of the H-bridge inverter cells of the MMC converter. Figure4.51
shows the output phase voltage of the prototype 5-level MMC converter.
Vdc (370 V)
VB (376 V)
Fig.4.50DC-link voltage of an H-bridge inverter cell
50 Hz phase voltage
738 V peak value
Fig.4.51Measured phase voltage of the prototype 5-level MMC converter before the LC line
filter circuit
4.4Summary
The amorphous alloy or nanocrystalline magnetic material-based high-frequency

magnetic link can be a good solution to provide multiple isolated and balanced DC
supplies for the MMC converter to step-up the low voltage of a photovoltaic array
or wind generator to a medium 3-phase AC voltage suitable for power transmis-
sion from remote (e.g., offshore) wind farms to the main grid transmission lines. It
can also overcome the common mode and voltage imbalance problems. It is found
that magnetic alloy 2605S3A, which has a narrower BH loops and lower specific
core loss than those of 2605SA1 at the high-frequency range, can be an excellent
choice for the development of high-frequency magnetic links for grid-connected
power converters.
References 151
References
2. Islam MR, Guo YG, Zhu JG (2011) H-bridge multilevel voltage source converter for direct grid
connection of renewable energy systems. In: Proceedings of IEEE Power and Energy Society
innovative smart grid technology conference, 1316 Nov 2011, Perth, Australia, pp 17
3. Islam MR, Guo YG, Zhu JG (2012) A Transformer-less compact and light wind turbine gener-
ating system for offshore wind farms. In: Proceedings of 2012 IEEE international conference
on power and energy (PECon2012), 25 Dec 2012, Kota Kinabalu, Malaysia, pp 605610
4. Prasai A, Yim JS, Divan D, Bendre A, Sul SK (2008) A new architecture for offshore wind
farms. IEEE Trans Power Electron 23(3):11981204
5. Pereda J, Dixon J (2011) High-frequency link: a solution for using only one DC source in
asymmetric cascaded multilevel inverters. IEEE Trans Ind Electron 58(9):38843892
Phys 113(17):17A324-117A324-3
Electron 29(8):41674182
115(17):17E710-117E710-3
10. Islam MR, Guo YG, Zhu JG (2013) Multiple-input-multiple-output medium frequency-
link based medium voltage converter for direct grid connection of photovoltaic arrays. In:
Proceedings of international conference on electrical machines and systems (ICEMS2013),
2629 Oct 2013, Busan, Korea, pp 202207
11. Agheb E, Bahmani MA, Hoidalen HK, Thiringer (2012) Core loss behavior in high
frequency high power transformers-II: arbitrary excitation. J Renew Sustain Energy
4(3):033113033113-11
12. Burdt R, Curry RC, McDonald KF, Melcher P, Ness R, Huang C (2006) Evaluation of
nanocrystalline materials, amorphous metal alloy, and ferrites for magnetic pulse compres-
sion applications. J Appl Phys 99(8):08D91108D911-3
13. Zhang Z, Sharma P, Makino A (2012) Role of Si in high Bs and low core-loss Fe85.2B10-
XP4Cu0.8SiX nano-crystalline alloy. J Appl Phys 112(10):103902103902-8
14. Makino A, Kubota T, Yubuta K, Inoue A, Urata A, Matsumoto H, Yoshida S (2011) Low core
losses and magnetic properties of Fe85-86Si1-2B8P4Cu1 nanocrystalline alloy with high B
for power applications. J Appl Phys 109(7):07A30207A302-5
15. Metglas Inc. Technical bulletin. Available at: http://metglas.com/products/magnetic_materi-
als/. Accessed on 20 Jan 2014
16. Vacuumschmelze. Compound Viteroperm. Available at http://vacuumschmelze.com.
Accessed on 03 Jul 2013
17. Pleite J, Prieto R, Asensi R, Cobos JA, Olias E (1999) Obtaining a frequency-dependent and
distributed-effects model of magnetic components from actual measurements. IEEE Trans
Magn 35(6):44904502
18. Villar I (2010) Multiphysical characterization of medium-frequency power transformers.
PhD dissertation, EPFL, 2010
19. Sippola M, Sepponen RE (2002) Accurate prediction of high-frequency power-transformer
losses and temperature rise. IEEE Trans Power Electron 17(5):835847
20. Baccigalupi A, Daponte P, Grimaldi D (1994) On a circuit theory approach to evaluate the
stray capacitances of two coupled inductors. IEEE Trans Instrum Measur 43(5):774776
21. Corsonello P, Daponte P, DeGrazia F, Grimaldi D (1992) On two-coupled inductors stray

capacitances evaluation for CAD-oriented modeling. In: Proceedings of the 35th midwest
symposium on circuits and systems, vol 2, 912 Aug 1992, Washington, DC, pp 12521255
22. Lu HY, Zhu JG, Hui SYR (2003) Experimental determination of stray capacitances in high
frequency transformers. IEEE Trans Power Electron 18(5):11051112
23. Storn R, Price K (1997) Differential evolutiona simple and efficient heuristic for global
optimization over continuous spaces. J Global Optim 11(4):341359
24. Baatar N, Zhang DH, Koh CS (2013) An improved differential evolution algorithm adopt-
ing -best mutation strategy for global optimization of electromagnetic devices. IEEE Trans
Magn 49(5):20972100
25. Lei G, Shao KR, Guo YG, Zhu JG, Lavers JD (2008) Sequential optimization method for the
design of electromagnetic device. IEEE Trans Magn 44(11):32173220
26. Islam MR, Lei G, Guo YG, Zhu JG (2014) Optimal design of high-frequency magnetic-links
for power converters used in grid connected renewable energy systems. IEEE Trans Magn, In
Press, doi:10.1109/TMAG.2014.2329939
Chapter 5
FPGA-Based Digital Switching Controller
for Multilevel Converters
AbstractThe available microcontrollers at present can only provide about six

pairs of PWM channels, which are clearly insufficient for multilevel converter
systems. On the other hand, the field programmable gate array (FPGA) may pro-
vide multiple PWM generators according to the converter requirements. Unlike
the digital signal processor (DSP) which runs a sequential program in its micro-
processor, an FPGA may run all the operations in parallel with the clock signal.
The capability of parallel processing of the FPGA affords the opportunity to the
switching controller to update all gate signals simultaneously. Various design
techniques and software environments are available for the modeling of switch-
ing control schemes with the FPGA technology. Most of the techniques require
special software such as HDL coder, System Generator, PSIM and ModelSim,
which increases the development time and cost. In this chapter, the most common
software such as the MATLAB/Simulink- and Xilinx ISE-based alternative design
technique is proposed, which may reduce the developmental time and cost of the
switching controller.
Keywords Multilevel converters Digital switching controller FPGA technology

DSP technology Design Experimental verification
5.1Introduction
Three basic technologies, the application-specific integrated circuit (ASIC), the

digital signal processor (DSP), and the field programmable gate array (FPGA),
have commonly been used for embedded applications. Due to some special fea-
tures such as fast processing speed, capability of parallel processing, and ability
to provide multiple PWM generators according to the converter requirements, the
FPGA technology may be an excellent choice for medium-voltage multilevel con-
verters [1]. Although several design techniques are available for the development
of FPGA-based switching controller, most of them require special software, which

154 5 FPGA-Based Digital Switching Controller for Multilevel Converters
Fig.5.1Photograph of an
ASIC by AMI
increases not only the developmental time but also the cost of the controller. In
this chapter, a fully digital switching controller is developed for a 3-phase 5-level
converter. The SK 30 GH 123 IGBTs are used to develop a prototype multilevel
converter, and an XC3S500E FPGA is used to develop the switching controller.
The most common software such as the MATLAB/Simulink- and Xilinx ISE-
based alternative design technique is used, which may reduce the developmental
time and cost of the switching controller. The simulation results serve as a pre-
liminary validation of the proposed design technique, which will be finally verified
by the experimental results. The developed switching scheme can be used for any
multilevel converter configurations with very little change in the software environ-
ment. Moreover, the proposed design and implementation techniques may be use-
ful for designing any other modern power converters switching controller.
5.1.1ASIC Technology
The ASIC technology is custom-designed for a particular application. ASICs pro-

vide the highest performance, because their design can be optimally configured
with respect to the applications requirements. However, the costs of ASIC-based
controllers are also the highest, mainly due to the nonrecurring engineering (NRE)
costs such as mask manufacturing and the cost of hardware and software develop-
ment. Figure5.1 shows an ASIC by AMI.
5.1.2DSP Technology
Compared to an ASIC, the DSP and FPGA technologies have much lower NRE
costs. The DSP has a fixed hardware configuration, which is the hardware of the
5.1Introduction 155
Fig.5.2Photograph of a
DSP by TI
basic functions of many signal processing algorithms. The DSP technology is a

fast microprocessor, which has much flexibility with different features. The DSP
technology is typically programmed in C/C++ or with assembly language for
performance. This technology is well suited to extremely complex mathematical
tasks, with conditional processing. However, the performance of DSP technology
is limited by the clock rate, and the number of useful operations per clock. For
example, the TMS320C6201 can achieve 400 millions multiplications per second
with two multipliers and 200-MHz clock [2]. Figure5.2 shows a photograph of
DSP by TI.
5.1.3FPGA Technology
An FPGA is a programmable logic device that consists of a large number of sim-

ple logic elements (LEs). The FPGA is programmed using a very-high-speed inte-
grated hardware description language (VHDL), which programs the connections
between the individual LEs to create logical functions (i.e., digital hardware).
Figure 5.3 shows a photograph of Xilinx FPGA. The main limitations of FPGA
technology are the number of LEs and signal propagation delay. Advancements in
FPGA technologies have increased the number of the available LEs and reduced
the propagation delay. Recent FPGAs have more multipliers. For example, Xilinx
Spartan-6 XC6SLX150T has 180 multipliers that can operate at 100MHz, which
gives 18,000 millions multiplications per second [3]. In comparison with the DSP
technology, the advanced FPGA technology gives 10100 times faster processing
speed. The yearly growth of FPGA capabilities compared to the traditional DSP is
depicted in Fig.5.4 [4]. The most important performance criteria of modern con-
trollers are use of high performance control algorithms, easy to modify the control
strategy and parameters, fully integrated single device controllers, low cost as well
Fig.5.3Photograph of an
FPGA by Xilinx
Fig.5.4Multiplications per 10
6
second of texas instruments
DSP and Altera FPGA 10
5
Multiplications/s
4
10
3
10
DSP
2 FPGA
10
1
10
1990 1995 2000 2005 2010
Year
as implementation time, and high reliability and accuracy. In this regard, recent
studies have shown that FPGA could be an appropriate alternative over DSP for
many applications.
5.1.3.1FPGA for Multilevel Converters
The PWM generator requirements of multilevel converters scale quadratically with

the number of levels. For example, a 3-phase voltage source N-level modular mul-
tilevel cascaded (MMC) converter needs 3(N 1) pair PWM generators [5].
The power circuit of a 3-phase 5-level converter system is illustrated in Fig.5.5.
The available microcontrollers at present can only provide about six pairs of PWM
channels, which are clearly insufficient for multilevel converter systems [6, 7]. On
the other hand, the FPGA may provide multiple PWM generators according to
the converter requirements. Unlike the DSP which runs a sequential program in
its microprocessor, an FPGA may run all the operations in parallel with the clock
5.1Introduction 157

Vdc Vdc Vdc
SA4 A B SC2 SC4 C
SA2 SB2 SB4
Fig.5.5Power circuit of a 3-phase 5-level converter
Table5.1A comparison between the DSP and the FPGA technology

Performance indicators Options
DSP FPGA
More than a few MHz sample rate Hard to handle Easy to handle
Programming language Standard C/C++ VHDL
More than 30Mb/s data rate Hard to handle Easy to handle
Conditional operations Easy to implement Hard to implement
Use of floating point Easy to implement Hard to implement
PWM output channel Limited (about six pairs) Can be configured as required
Parallel processing Serial implementation Serial/parallel implementation
signal. The capability of parallel processing of the FPGA provides the opportu-
nity to the switching controller to update all gate signals simultaneously [8]. In
most cases, the processing time is independent of the number of converter levels.
Therefore, the FPGA technology is a natural choice for the control of multilevel
converters. A comparison between the FPGA and the DSP technology is summa-
rized in Table5.1.
5.1.3.2FPGA Developmental Boards and Kits
Xilinx and Altera are the current market leaders in the FPGA technology. These
two vendors have developed different series of FPGA boards and kits. FPGA
developmental boards and kits are specially designed and suitable for the design of
prototype test platforms. Before selecting the board, the software environment also
needs to be considered because different vendors may use different programming
software. The selection of FPGA boards depends on many factors, such as clock
Table5.2FPGA developmental boards and kits

Brand FPGA boards and kits Clock I/O LEs Memory kB Multipliers Price AUD
MHz pins K
Altera Arria II GX dev. kit 125 1,152 124 8,121 576 1,495
Altera DE2-115 D&E 50 528 114 486 266 595
board
Altera DE3 50 736 254 1,836K 768 2,695
dev. system
Stratix III dev. kit 125 744 142 128K 384 2,495
Stratix IV E dev. kit 125 744 531 2,000K 1,024 5,495
Stratix V GX dev. kit 156 696 622 1,152K 512 6,995
EZ1CUSB-altera 50 128 12 29 2 219
dev. board
Cyclone III dev. kit 125 531 119 256K 288 995
Xilinx Spartan-3 starter 50 173 4 27 12 139
board
Spartan-3E 1600 50 250 33 8 36 225
dev. board
Spartan-3E starter 50 232 10 8 20 189
board
Spartan-6 LX150T 100 540 147 1,355 180 995
dev. kit
speed, number of LEs, number of I/O pins, number of multipliers, area of appli-
cations, market availability, and commercial price. A few commercially available
developmental boards and kits are summarized in Table5.2 [9, 10].
5.2FPGA-Based Switching Controller Design Techniques
Various design techniques and software environments are available for the mod-
eling of switching control schemes with the FPGA technology. The most com-
monly used design techniques are (i) modeling the switching circuit and target
system in the MATLAB/Simulink environment and generation of the pro-
gramming file with hardware description language (HDL) coder, as shown
in Fig.5.6 and (ii) modeling the switching circuit and target system in the
MATLAB/Simulink environment and generation of the programming file with
the System Generator, as shown in Fig.5.7. In order to verify the performance
of generated VHDL code in the simulation environment, the ModelSim and
PSIM with ModCoupler represent the possible options. These techniques require
special software such as HDL coder, System Generator, PSIM, and ModelSim,
which increases the developmental time and cost. Figure5.8 shows the interfacing
with target system in simulation environment. In this chapter, the most common
5.2 FPGA-Based Switching Controller Design Techniques 159
Design
MATLAB/Simulink
HDL coder
HDL verifier
Implementation/Verification
VHDL/Verilog code
FPGA
Fig.5.6MATLAB/Simulink and HDL coder-based FPGA programming
Design
MATLAB/Simulink
System generator
Develop system generator
representation
Automatic RTL generation

Bitstream
FPGA
Fig.5.7MATLAB/Simulink- and system generator-based FPGA programming
software such as the MATLAB/Simulink- and Xilinx ISE-based alternative design

technique is proposed which may reduce the developmental time and cost of the
switching controller [1].
In a 5-level converter, two H-bridge cells on each phase are shown in Fig.5.5.
For each H-bridge cell, one reference signal has to be compared with two triangu-
lar function carrier signals (one for the left half-bridge and its inverted signal for
PSIM Mod Coupler Model Sim

Digital control
(PSIM schematic) Communication-link circuit (VHDL)
(PSIM model)
Time step control

Sensors Control
Clock generation algorithms
Gate
signals I/O management
Fig.5.8Interfacing with target system in simulation environment
Reference signal
generator (3-phases)
SA1
>= 0 SB1
SC1
Comp
NOT
SC2
Inv
SA3
>= 0 SB3
Inv SC3
Comp
SA4
SB4
NOT
SC4
Inv
SA5
>= 0 SB5
SC5
Comp
NOT
SC6
Inv
SA7
>= 0 SB7
Inv SC7
Comp
SA8
SB8
NOT
SC8
Inv
Gate pulses
Fig.5.9Switching scheme for a 5-level MMC converter

5.2 FPGA-Based Switching Controller Design Techniques 161
the right half-bridge). The carrier for each H-bridge cell on a particular phase leg
is phase-shifted by 360/(m1) [11, 12]. Figure5.9 shows the switching scheme
for a 5-level converter [13].
On each phase leg, in total four half-bridge cells, only two carrier signals are
required and the other two can be generated just by inverting them. In total, three
reference signals are required, one for each phase leg. The frequency of the refer-
ence signal determines the frequency of the converter output voltage. The com-
parator module compares the carrier signal with the respective reference signal and
generates control pulses including reasonable deadtime as required by switching
devices. An onboard crystal (e.g., 50MHz) is used for a clock source, and a clock
divider reduces the clock frequency. A lookup table (LUT) is used to generate the
reference signals, which makes the control circuit totally digital and integrated. The
basic architecture of the FPGA-based switching controller is illustrated in Fig.5.10.
In total, there are eight switching devices in a phase, requiring eight gate pulses
to drive them. Including the inverted carrier signals, a total of four carriers are able
to generate four gate pulses when comparing them with a reference signal. The
other four gate pulses can be generated by just inverting these four gate pulses
with a consideration of deadtime.
The available libraries of the Xilinx FPGA development software Xilinx ISE
Design Suite 13.2 only have basic logic symbols, which are clearly insufficient for
FPGA board
SA1
00 Com SB1
00 SC1
1200 1200
SA2
2400 2400
NOT SB2
SC2
Reference
SA3
signal Counter00 00 Com SB3
generator
Dead band unit
SC3
Inv 1800
Carrier
Gate signals
(3-phases) SA4
signal 900 900 NOT SB4
Look-up generator SC4
Inv 2700
Table SA5
Slow Clk Com SB5
SC5
Clk divider SA6
NOT SB6
SC6
Comparator
SA7
Com SB7
Crystal Clk SC7
SA8
NOT SB8
SC8
Fig.5.10Switching controller architecture

designing a sophisticated switching controller. According to the system require-

ment, new logic symbols need to be created by writing the VHDL program. In
the Xilinx ISE software environment, it is difficult to model a target system (e.g.,
switching devices and loads). Therefore, the switching controller performance
cannot be verified before implementation, which may be unsafe especially in
power circuits. On the other hand, in the MATLAB/Simulink environment, the
switching controller with target system can be modeled and the performance can
be readily verified. Hence, during the modeling of switching controller in the
Xilinx ISE environment, the MATLAB/Simulink results are considered as a refer-
ence. The Simulink and Xilinx ISE 13.2 Design Suite software-based design tech-
nique is illustrated in Fig.5.11.
In this technique, the switching control scheme with the target system is mod-
eled in the MATLAB/Simulink environment first. After getting satisfactory perfor-
mance from Simulink, the updated model can be used for behavioral modeling of
the switching controller in the Xilinx ISE environment. The behavioral simulation
results can be compared with the Simulink results. After getting satisfactory simu-
lation results, the design can be implemented with verification of timing simula-
tion. Before connecting with the target system, the gate signals could be measured
and compared with the Simulink results as well as theoretical values.
Fig.5.11Simulink and Design & Simulation

Xilinx ISE 13.2 Design
Suite software-based design MATLAB/Simulink
technique
Results
Xilinx ISE Design Suite
Design & Simulation

(Compare results with
Simulink results)
RTL generation
Timing simulation
Bitstream
FPGA
5.3 Modeling and Schematic Symbol Creation Using VHDL 163
5.3Modeling and Schematic Symbol Creation Using

VHDL
5.3.1Carrier Signal Generation Unit

The MATLAB/Simulink model used to generate the carrier signals is depicted
in Fig.5.12. The second signal is phase-shifted by 90. The carrier signals from
Simulink are shown in Fig.5.13.
The triangular carrier waves as shown in Fig.5.13 can be generated by an up
down counter as depicted in Fig.5.14. To obtain the correct PWM frequency, a
clock divider is used to reduce the crystal frequency before it is used to clock the
1
sin asin -K- 1
theta
-K- 1
0
sin asin -K- 2
90 -K- 2
Fig.5.12MATLAB/Simulink model for two carrier signals, where = t
1
Amplitude (V)
0.5
0
-0.5
-1
0 0.5 1 1.5
Time (s) -3
x 10
Fig.5.13Two carrier signals; one for the top H-bridge and 90 shifted carrier for the bottom
H-bridge
PWM period PWM period

H H
H-1 H-1 H-1 H-1
H-2 H-2 H-2 H-2
t
3 3 3 3
2 2 2 2
1 1 1 1
0 0 0
Up counter Down counter Up counter Down counter
Fig.5.14Updown counter-based carrier signal generation, where H is the highest count, and
t the step size
counter unit. The frequency of the carrier signal should be many times higher than
the reference signal. In this work, 1.5-kHz and 50-Hz frequencies are considered
for the carrier and reference signals, respectively (i.e., carrier is set 30 times fast
as the reference signal). For the 9-bit updown counter, the total number of steps
in a PWM period is 229=1,024. If switching frequency 1.5kHz is considered,
then the required counter clock frequency is 1,0241.5103Hz.
Therefore, the clock divider value, k1, for a system clock frequency of 50MHz
is given by
50 106 Hz
k1 = = 32.552. (5.1)
1024 1.5 103 Hz
In this work, the value of k1 has been considered to be 32, and with this value,
the switching frequency is calculated as 1.525kHz. The updown counter and
comparator unit combination generates a pulse center aligned within the PWM
period. The center-aligned PWM has advantages over the edge-aligned PWM in
that the outputs that control the inverter are not all switching on at the same time,
at the beginning of every period, as they would do with edge-aligned PWM. This
can help reduce noise on the converter power lines, thus increasing conversion
efficiency.
The VHDL-code-based behavioral model of the updown counter is designed
and synthesized using the Project Navigator tool of the Xilinx ISE Design Suite
13.2 software. The syntax of the behavioral model is verified using the behavioral
check syntax tool. After having analyzed the performance of the behavioral model
using the ISim simulator, a schematic symbol is created as shown in Fig.5.15a.
Similarly, the 90 phase-shifted counter is created as depicted in Fig.5.15b. An
inverter or 180 phase shifter is synthesized to generate the inverted carrier sig-
nals from the output of the T_Cou_V_3 and B_Cou_V_3 counters. The schematic
symbol of the inverter is illustrated in Fig.5.16.
Fig.5.15Schematic symbols of updown counter: a without phase shift and b with 90 phase
shift
Fig.5.16Schematic
symbols of inverter, 180
phase shifter
5.3.2Reference Signal Generation Unit
In order to generate the 50-Hz 3-phase voltage output waveform, three ref-
erence signals with 120 phase-shifted from each other are needed. The
MATLAB/Simulink model for the generation of these reference signals is illus-
trated in Fig.5.17, and the generated reference signals are shown in Fig.5.18.
The microcontrollers and DSPs are commonly used to generate reference signals.
This additional external component increases the system size, weight, and cost and
requires proper synchronization. In this instance, logic-based reference signal gen-
eration (programming the FPGA logic cells to generate reference signals) using
direct digital synthesis (DDS) has attracted significant attention recently [1416].
The DDS is an LUT-based technique to generate analog signal using stored digital
samples. An LUT is actually a set of memory locations which contain the sampled
values of a desired analog signal.
2*pi
m
Freq
1
sin Ust
12:34
[0 -2*pi/3 2*pi/3]
Phase -K-
deg. ->rad
Fig.5.17Reference signals generation model
1
Phase A Phase B Phase C
Reference signals
0.5
0
50 Hz
-0.5
-1
0 0.005 0.01 0.015 0.02
Time (s)
Fig.5.18Three-phase 50-Hz reference signals (in MATLAB/Simulink environment)

The frequency of the sine wave can be increased by skipping entries in the sine
lookup table as the angular resolution depends on the number of entries, Ln for
a cycle. The angular resolution can be calculated from
360
= (i + 1) (i) = . (5.2)
Ln
where i varies from 0 to Ln1.

If Aout is the amplitude of the output signal, then the sample value can be calcu-
lated from [1]

360
SR (i + 1) = Aout sin i+ . (5.3)
Ln
where is the phase angle. In order to keep the output span positive between 0
and Aout, the sample value stored in the LUT can be expressed as [1]

Aout Aout 360
SLUT (i + 1) = + sin i+ (5.4)
2 2 Ln
The clock divider may determine the frequency of the reference signals. For exam-
ple, if there are 256 entries in the LUT, to generate a 50-Hz reference signal from
256 entries using the 50-MHz clock, the clock divider value k2 is given by
50 106 Hz
k2 = = 3,906.25. (5.5)
256 50 Hz
The sample values of the reference signals can be calculated from (5.4). The ref-
erence signals for phases A, B, and C can be generated by considering the values
of as 0, 120, and 120, respectively. If the peak value, Aout, is 255, and Ln
250 Phase A Phase B Phase C

255
200
Reference value
128
150
100
50
0
0
1 50 100 150 200 250
Entries
Fig.5.19Three-phase reference signals (in ISim Simulator environment)

Fig.5.20Schematic symbol
of reference signal block
256, the decimal sample values of three sine reference signals can be presented as
shown in Fig.5.19. In order to store the sample values, the decimal number needs
to be converted to hexadecimal number.
A VHDL-code-based program is used to create the LUT, which contains the
sine reference. Different carrier-based modulation techniques used in the tradi-
tional converters can also be used in the MMC converter system (e.g., SPWM and
THPWM) [17]. Each technique has its unique advantages and disadvantages. In
order to analyze the converter performance against different types of reference sig-
nals, sine and third-harmonic injected sine reference signals are stored in the LUT.
Only one reference signal from each type is stored in the LUT. Based on the stored
reference signal, three sine waves (each 120 out of phase from the others) are cre-
ated by using three separate indexes into the LUT. Only one type reference is used
at a time, which can be controlled by an onboard switch without modifying the
program. After verifying the syntax of the behavioral model and analyzing the per-
formance through the ISim Simulator environment, a schematic symbol is created
as shown in Fig.5.20.
5.3.3Comparator Unit
One comparator unit is used for each of the half-bridge cells. The comparator units
compare the carrier signal with a reference signal. Each comparator unit gener-
ates one switching signal for the top switching device of the half-bridge cell. The
inverted form of this switching signal drives the bottom switching device. For the
left half-bridge cell, one is asserted when the reference signal value is greater than
or equal to the carrier signal value, and the other is asserted when the reference
signal value is less than the carrier signal value. For the right half-bridge cell, one
is asserted when the inverted carrier signal value is greater than or equal to the
reference signal value, and the other is asserted when the inverted carrier signal
value is less than the reference signal value. The MATLAB/Simulink model of the
comparator unit for a 5-level converter is illustrated in Fig.5.21. The four carrier
signals and the three reference signals as required by the 3-phase 5-level converter
are depicted in Fig.5.22.
The overall switching system is implemented in the MATLAB/Simulink envi-
ronment to verify the performance roughly. Each H-bridge cell requires four gate
1
TT1
Ust >=
NOT TT2
Com
0 Inv
TT3
>=
-1
Com1 NOT TT4
1
1 Inv1
2 theta
wt
TT5
2 >=
NOT TT6
Com2
2 Inv2
Carriers
Generator TT7
-1
>=
Com3 NOT TT8

3
Inv3
Fig.5.21Simulink block diagram of comparator module
0.5
Amplitude
-0.5
-1
0 0.002 0.004 0.006 0.008 0.01
Time (s)
Fig.5.22Carrier and reference signals in Simulink environment: thick waves represent sine ref-
erence signals and dotted triangular lines represent the inverted form of carrier signals as pre-
sented by triangular solid lines
pulses as presented in separate figures. In a 3-phase 5-level converter, there are

totally six H-bridge cells and two on each phase. In total, 24 gate pulses are illus-
trated in Figs.5.23, 5.24, 5.25, 5.26, 5.27, and 5.28. In the Simulink environment,
the deadband is not considered since the switching devices are ideally operated.
The concept as used in Simulink is implemented in the Xilinx ISE Design Suite
13.2 software environment using VHDL coding. Since practical power semicon-
ductor devices do not switch-on or switch-off immediately, i.e., certain time is
(a) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
(b) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
(c) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
(d) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
Time (s)
Fig.5.23Gate pulses for the top H-bridge cell in phase A: a left top, b left bottom, c right top
and d right bottom
(a) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
(b) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
(c) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
(d) 1
0.5
0
0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014
Time (s)
Fig.5.24Gate pulses for the bottom H-bridge cell in phase A: a left top, b left bottom, c right
top and d right bottom
(a) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
(b) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
(c) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
(d) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
Time (s)
Fig.5.25Gate pulses for the top H-bridge cell in phase B: a left top, b left bottom, c right top
and d right bottom
(a) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
(b) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
(c) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
(d) 1
0.5
0
0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02
Time (s)
Fig.5.26Gate pulses for the bottom H-bridge cell in phase B: a left top, b left bottom, c right
(a) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
(b) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
(c) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
(d) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
Time (s)
Fig.5.27Gate pulses for the top H-bridge cell in phase C: a left top, b left bottom, c right top
and d right bottom
(a) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
(b) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
(c) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
(d) 1
0.5
0
0.03 0.031 0.032 0.033 0.034 0.035 0.036 0.037
Time (s)
Fig.5.28Gate pulses for the bottom H-bridge cell in phase C: a left top, b left bottom, c right
required to switch them on or off with the switching technique as illustrated in

Fig. 5.29, there could be a period of time when both the switching devices of a
half-bridge cell are on at the same time, causing a direct short circuit of the ter-
minals of the respective source. In order to avoid this crucial problem, a deadband
period is inserted between the turning off of one switching device and the turn-
ing on of the other switching device on a half-bridge leg. Thus, both switching
devices remain off before one turns on, which ensures the safe operation of both
switching devices. The switching technique with deadband is shown in Fig.5.30,
where one PWM output is asserted when the count value is greater than or equal
Fig.5.29Switching Top PWM

technique without deadband
Bottom
PWM
TPWM TPWM TPWM
Carrier Reference
Time
Fig.5.30Switching Top PWM

technique with deadband
Bottom
PWM
TPWM TPWM TPWM
Carrier Deadband Reference
Time
Fig.5.31Schematic symbol
of the comparator and PWM
pulse generator unit, which
also has an output terminal
called DBA_L to access the
deadband wave against PWM
waves at pwm_L_T and
pwm_L_B
to the reference value plus half of the deadband value and the other PWM output
is asserted when the carrier value is less than the reference value minus half of the
deadband value.
The designed comparator as well as the PWM pulse generator is simulated by
the ISim Simulator, and a schematic symbol is created from the VHDL program-
based behavioral model. The schematic symbol for the comparator and PWM
pulse generator unit is depicted in Fig.5.31. The overall algorithm to synthesize
a single fully digital integrated circuit (IC) to control a 5-level converter is illus-
trated in Fig.5.32 [9].
Reference
signals
Crystal clk
Up-down
Clock 900 shifted
divider up-down
counter
counter
Inverter Inverter
Comparator Comparator Comparator Comparator

NOT
NOT
NOT
NOT
Deadband
SC7
SC8
SB7
SA7
SB8
SA8
SC1
SC2
SC3
SC4
SC5
SC6
SB1
SA1
SB2
SA2
SB3
SA3
SB4
SA4
SB5
SA5
SB6
SA6
Gate signals
Fig.5.32Basic block diagram of FPGA-based 5-level converter control algorithm
5.4 Model of Complete Switching Controller 173
5.4Model of Complete Switching Controller
In the Project Navigator environment, a schematic-based complete switching

controller for the 3-phase 5-level converter is developed using schematic sym-
bols created by VHDL programs. The complete switching controller is depicted
in Fig.5.33, where totally there are 12 comparators in a three-column arrange-
ment. Each column is related to a phase of the converter and each comparator unit
corresponds to a half-bridge cell of the converter. The top and bottom counters
are named T_Cou_V_3 and B_Cou_V_3, respectively. The comparator circuits
corresponding to the left and right half-bridge cells of an H-bridge inverter cell
are named L_Com_V_3 and R_Com_V_3, respectively. The deadtime is inserted
within each comparator unit as a variable, which can be adjusted according to
the requirements of the semiconductor devices. All units, such as carrier gen-
erators, reference generators and comparators, are connected in parallel with the
same clock source, which ensures the parallel operation of FPGA. The complete
switching controller is simulated using the behavioral simulation process. Then,
the model is synthesized and followed by the implementation process. Before
synthesizing the model, all I/Os are assigned to available I/O ports of the FPGA
board. During the implementation process, the model verification is carried out.
In the verification process, timing simulation is carried out to check the real-time
performance.
Fig.5.33Basic block diagram of FPGA-based 3-phase 5-level converter control circuit

5.5Behavioral Simulation of Switching Controller
A test bench program file is created using VHDL code to verify the control cir-
cuit performance. The behavioral model of the complete switching control-
ler is simulated by the ISim Simulator with the created test bench file. The gate
pulse waveforms, deadtime wave forms, and other timing signals such as clock,
load, enabling and reset are displayed on the waveform viewer. The gate pulses
as well as deadtime waveforms are illustrated in Figs.5.34, 5.35, and 5.36, each
Fig.5.34Gate pulses and deadband signals for switches in phase A, where the narrow pulses
represent the deadband (adjustable in software environment)
Fig.5.35Gate pulses and deadband signals for switches in phase B, where the narrow pulses
5.5 Behavioral Simulation of Switching Controller 175
Fig.5.36Gate pulses and deadband signals for switches in phase C, where the narrow pulses
corresponding to one phase of the converter circuit. Through the waveform viewer,
it is possible to analyze (by comparing with the MATLAB/Simulink results) the
generated gate pulses. The deadtime is not considered in the MATLAB/Simulink
system, but the pulse pattern verification can be done with this comparison.
5.6Model Implementation
After analyzing the performance of the behavioral model of the complete switching
controller with behavioral simulation, the model is ready for implementation. The
schematic model is first synthesized using the Synthesized-XST tool. All input and
output ports are assigned according to the available pin configuration of the Xilinx
FPGA chip using the User Constraints-Floorplan Area/IO/Logic (PlanAhead) tool.
All output signals (24 gate signals and twelve deadband signals) are assigned
to the first row connectors of the Hirose 100-pin FX2 connector on the Xilinx
Spartan-3E development board used for testing. Most of all the connectors in the
second row are connected to the ground. There are eight small LEDs (LD0LD7)
parallel connected with eight pins of the top row. Therefore, these signals should
be assigned to the pins whose output requires monitoring. The board has four slide
switches labeled as SW0SW3. When a switch is in the UP or ON position, it con-
nects the respective FPGA pin to 3.3V (i.e., a logic high) and when in the DOWN
or OFF position, it connects the FPGA pin to the ground (i.e., a logic low). The
input control signals, such as the load, enabling, and reference type, are assigned
to these switches. The board also has four momentary push button switches.
Pressing a push button connects the associated FPGA pin to 3.3V (i.e., a logic
high). The I/O pin assignment pattern is summarized in Table5.3.
Table5.3I/O pin Signal name Direction Pin number

configuration
SineType [0] Input L13
SineType [1] Input L14
Ena Input N17
Load Input H18
Clk Input C9
Reset Input K17
SA1 Output C14
SA2 Output B14
DA1 Output A16
SA3 Output E12
SA4 Output E11
DA2 Output B16
SA5 Output A14
SA6 Output B13
DA3 Output E13
SA7 Output F11
SA8 Output C11
DA4 Output C4
SB1 Output A13
SB2 Output F12
DB1 Output B11
SB3 Output D11
SB4 Output E9
DB2 Output A11
SB5 Output E7
SB6 Output B6
DB3 Output A8
SB7 Output F9
SB8 Output E8
DB4 Output G9
SC1 Output A6
SC2 Output C5
DC1 Output C3
SC3 Output F8
SC4 Output C7
DC2 Output B4
SC5 Output D5
SC6 Output A4
DC3 Output D14
SC7 Output D7
SC8 Output F7
DC4 Output D10
5.6 Model Implementation 177
There are two ways to configure the FPGA pins in the Xilinx ISE Design Suite
13.2 software environment. The first method is to enter the pin number in the site
field in the I/O Port Properties tab when the I/O signal is selected. The I/O Port
Properties-based I/O pins assignment is illustrated in Fig.5.37.
Dragging into the Package view is an alternative way for the I/O pins assign-
ment. In this method, a signal needs to be selected, dragged into the Package
Fig.5.37Assigning I/O pins via I/O port properties
Fig.5.38Assigning I/O pins by dragging into package view

Fig.5.39I/O ports viewer
view and then dropped on the desired pin location. The Package view is shown
in Fig.5.38. The I/O port viewer displays the ports with their configuration such
as input/output and pin number as depicted in Fig.5.39. Once the PlanAhead is
closed by selecting the Exit tab, a ucf file is added to the project, which contains
the constraints. In order to translate the model to something that can physically
be mapped into the FPGA, i.e., xc3s500e-4fg320, the Implement Design tab is
used.
Actually under the Implement Design tab, there are three main functions such
as Translate, Map, and Place and Route. The Translate process merges all of the
input netlists and design constraint information and creates a Xilinx native generic
database (NGD) file. The Map process creates a native circuit description (NCD)
file, which actually implements the logic functions in a device. The Place and
Route process creates another NCD file to place and route the design by using the
NCD file created during the Map process.
5.7Design Verification
The design Summary viewer reports all the design issues such as overall summary
(as shown in Fig.5.40), timing constraints, pinout report (as shown in Fig.5.41),
errors and warnings and other information.
5.7 Design Verification 179
Fig.5.40Design summary/report viewer
Fig.5.41Design summary/pinout report

Fig.5.42RTL schematic viewer
It is essential to check the pinout report carefully (e.g., whether or not all sig-
nals have been assigned to the correct FPGA pins). The View RTL Schematic
process views the schematic of the behavioral model as depicted in Fig.5.42.
Moreover, the View Technology Schematic process is used to view the physical
architecture of the I/O configuration as illustrated in Fig.5.43.
The timing (post-place and route) simulation uses the block and routing delay
information to give a more accurate assessment of the behavior of the circuit. The
timing simulation is carried out to ensure the actual device operation in the simula-
tion environment.
5.8 FPGA Programming 181
Fig.5.43View technology
schematic
5.8FPGA Programming
After analyzing the design, a configuration bitstream is created. The process tab
Generate Programming File generates a bit file that can be used to program the
FPGA. Before programming the FPGA, the configuration mode jumper settings
must be set for JTAG programming. After connecting the FPGA board with the
PC through a standard USB cable on the USB-based download/debug port of
the test board, the power switch is turned on. Figure5.44 shows a photograph of
FPGA board with PC connection through USB cable. A green light-emitting diode
(LED) lights up after a successful connection. The FPGA is programmed using
the iMPACT tool under the Configure Target Device process. There is no need to
include some configuration files for SPI ports or BPI PROM to this device. The
configuration files assignment options for other ports of xcf04s and xc2c64 should
be bypassed. During this process, the device xc3s500e is selected to program the
FPGA. The iMPACT process viewer is shown in Fig.5.45. There are two indi-
cations of successful programming: a message Program Succeeded shows up in
iMPACT process view and at the same time a yellow LED (underneath the J30
jumper) on the board lights up.
Fig.5.44Photograph of FPGA board with PC connection through USB cable
Fig.5.45iMPACT process view
5.9Experimental Testing and Verifications
All FPGA I/Os that interface to the Hirose connector on the test board are in Bank
0 of the FPGA. The I/O Bank 0 supply is 3.3V by default. Therefore, the PWM
gate pulses from the FPGA are also 3.3V, which is insufficient for the IGBT drivers
used. Driver ICs TC4427A are used to increase the voltage level to the desired value
5.9 Experimental Testing and Verifications 183
(TC4427A can support up to 18V). Through Hirose 100-pin FX2 connector and rib-
bon cables, the generated 3.3-V gate pulses are applied to the interface circuit. The
experimental setup of the switching controller for the 5-level 3-phase multilevel con-
verter system is illustrated in Fig.5.46. The gate pulses from the driver circuit can be
used to switch the switching devices. The gate pulses for a 3-phase 5-level converter
are illustrated in Figs.5.47, 5.48, 5.49, 5.50, 5.51, and 5.52.
An Agilent Technologies DS06034A oscilloscope was used to observe the
signal waveforms. This oscilloscope has four input channels, and the gate pulses
for each H-bridge cell (i.e., four) are shown in a figure. The measured switching
Fig.5.46Photograph of the FPGA-based switching controller section
Left top
Left bottom
Deadband
Right top
1.525 kHz
Right bottom
Top H-bridge cell, phase A
Fig.5.47Measured gate pulses for the top H-bridge cell in phase A: 1 left top, 2 left bottom, 3
right top, and 4 right bottom
Left top
Left bottom
Deadband
1.525 kHz
Right top
Right bottom
Bottom H-bridge cell, phase A
Fig.5.48Measured gate pulses for the bottom H-bridge cell in phase A: 1 left top, 2 left bottom,
3 right top, and 4 right bottom
Left
top
Left bottom Deadband

Right top
1.525 kHz
Right bottom
Top H-bridge cell, phase B
Fig.5.49Measured gate pulses for the top H-bridge cell in phase B: 1 left top, 2 left bottom, 3
frequency of the switching signals is 1.525kHz, which precisely matches the the-
oretical result. Pulse patterns are also verified with theoretical results as well as
MATLAB/Simulink and Xilinx ISE simulation results, and they are found highly
consistent. Gate pulses are applied to the switching devices of the converter, and
5.9 Experimental Testing and Verifications 185
Left
top
Left bottom
Deadband
1.525 kHz Right top
Right bottom
Bottom H-bridge cell, phase B
Fig.5.50Measured gate pulses for the bottom H-bridge cell in phase B: 1 left top, 2 left bot-
tom, 3 right top, and 4 right bottom
Left top
Left bottom
Deadband
Right top
1.525 kHz Right bottom
Top H-bridge cell, phase C
Fig.5.51Measured gate pulses for the top H-bridge cell in phase C: 1 left top, 2 left bottom, 3
the output phase voltage is measured using oscilloscope as shown in Fig.5.53, and
it is found satisfactory. The phase voltage waveforms were also found highly con-
sistent with MATLAB simulation results. Figure5.54 shows simulated phase volt-
age waveform of a 5-level converter.
Left top
Left bottom
Deadband
1.525 kHz
Right top
Right bottom
Bottom H-bridge cell, phase C
Fig.5.52Measured gate pulses for the bottom H-bridge cell in phase C: 1 left top, 2 left bottom,
3 right top, and 4 right bottom
Fig.5.53Measured phase voltage of a prototype 5-level converter
1000
Phase voltage
500
Voltage (V)
-500 50 Hz
-1000
0 5 10 15 20 25 30 35 40
Time (ms)
Fig.5.54Simulated phase voltage of a prototype 5-level converter

5.10Summary 187
5.10Summary
In order to develop switching circuits for multilevel converters, the FPGA may
be the natural choice because of its high-speed operation and capacity to han-
dle multiple PWM signals. An integrated fully digital switching controller has
been developed for a 3-phase 5-level converter system using an FPGA devel-
opment board, which has some user-friendly onboard switches and buttons for
I/O. Through these switches and buttons, it is possible to control the operation
of the switching controllers such as reset, enabling, load, type of reference sig-
nal, and many more. During the behavioral modeling of the switching control-
ler, the MATLAB/Simulink model and its responses have been considered as the
reference. Finally, the switching circuit has been implemented for experimental
verification. The experimental results are found highly consistent with theoreti-
cal and MATLAB/Simulink results. This switching scheme and design technique
can be used for any other converter with minor modification to the software
environment.
References
1. Islam MR, Guo Y, Zhu J (2014) A high-frequency link multilevel cascaded medium-voltage
converter for direct grid integration of renewable energy systems. IEEE Trans Power Electron
29(8):41674182
2. Texas Instruments, fixed-point digital signal processor, Application Note, SPRS051H (2004)
3. Xilinx, Spartan-6 family overview, product specification, DS160 (2011)
4. Ekas P (2007) FPGAs rapidly replacing high-performance DSP capability. DSP-FPGA.COM
MAGAZINE. [Online] Available at: http://dsp-fpga.com/article-id/?2051. Accessed on 10
Feb 2013
5. Islam MR, Guo Y, Zhu J (2014) A review of offshore wind turbine nacelle: technical chal-
lenges, and research and developmental trends. Renew Sustain Energy Rev 33:161176
6. Islam MR, Guo Y, Zhu J (2014) A multilevel medium-voltage inverter for step-up-trans-
7. Islam MR, Guo Y, Zhu J (2012) A novel medium-voltage converter system for compact and
light wind turbine generators. In Proceedings of Australasian universities power engineering
conference (AUPEC2012), Bali, Indonesia, pp 16, 2629 Sept 2012
8. Islam MR, Guo Y, Zhu J (2013) Multiple-input-multiple-output medium frequency-link
based medium voltage converter for direct grid connection of photovoltaic arrays. In:
Proceedings of International Conference on Electrical Machines and Systems (ICEMS2013),
Busan, Korea, pp 202207, 2629 Oct 2013
9. ALTERA, All Development Kits. [Online] Available at: http://www.altera.com/
products/devkits/kit-dev_platforms.jsp. Accessed on 13 July 2013
10. XILINX, Boards and Kits. [Online] Available at: http://www.xilinx.com/products/boards-
and-kits/index.htm. Accessed on 01 Apr 2013
11. Rodriguez J, Lai JS, Peng FZ (2002) Multilevel inverters: a survey of topologies, controls
and applications. IEEE Trans Ind Electron 49(4):724738
12. Islam MR, Guo Y, Jafari M, Malekjamshidi Z, Zhu J (2014) A 43-level 33kV 3-phase mod-
ular multilevel cascaded converter for direct grid integration of renewable generation sys-
tems. In: IEEE innovative smart grid technologies conference (ISGT) Asia, Kuala Lumpur,
Malaysia, 2023 May 2014
13. Islam MR, Guo Y, Zhu J (2014) Multilevel converters for step-up transformer-less direct grid
integration of renewable generation units with medium voltage smart microgrids. Large scale
renewable power generation: advances in technologies for generation, transmission and stor-
age. Springer, Berlin, Heidelberg, pp 127149
14. Suchenek M, Starecki T (2012) Programmable pulse generator based on programmable logic
and direct digital synthesis. Rev Scientific Instruments 83(12):124704-1124704-4
15. Li Z, Zhou W, Zhou H, Zhang X, Zhao J (2013) The optimization of super-high resolution
frequency measurement techniques based on phase quantization regularities between any fre-
quencies. Rev Scientific Instruments 84(2):025106-1025106-5
16. Tierney J, Rader CM, Golg B (1971) A digital frequency synthesizer. IEEE Trans Audio
Electroacoust AU-19:4857
17. Islam MR, Guo Y, Zhu J (2011) H-bridge multilevel voltage source converter for direct
grid connection of renewable energy systems. In: IEEE power and energy society innova-
tive smart grid technology conference (ISGT 2011 Asia), Perth Australia, pp 17, 1316 Nov
2011
Chapter 6
Experimental Validation of 1-kV
Modular Multilevel Cascaded Converter
with High-Frequency Magnetic Link
AbstractIn this chapter, the high-frequency magnetic-link modular multilevel

cascaded (MMC) converter is experimentally validated for grid integration of
renewable power generation systems. The high-frequency magnetic link generates
multiple isolated and balanced DC supplies for all the H-bridge inverter cells of
the MMC converter. To verify the feasibility of the converter, a scaled down 1-kV
laboratory prototype test platform with a 5-level MMC converter is developed in
this chapter. The design and implementation of the prototyping, test platform, and
experimental results are analyzed and discussed. The methods/techniques for com-
ponent selection, converter fabrication, and experimentation are also applicable to
other converter applications.
KeywordsModular multilevel cascaded (MMC)Power convertersPrototype

Test platform Component selection Experimental validation
6.1Introduction
Recently, the development of the medium-voltage converter, especially for off-

shore and remote area renewable power generation applications, has attracted sig-
nificant attention [1, 2]. In 2008, a multi-coil modular permanent magnet generator
(isolated stator coils) was proposed to eliminate the step-up transformers of wind
turbine generators [3]. This was validated by a 230-V test rig. This multiwinding
generator generates multiple DC supplies for the modular multilevel cascaded
(MMC) converter, which requires a special winding arrangement (still at a heavy
weight and large volume due to the power frequency operation) and complicated
control strategies. A multicoil generator-based wind turbine system is shown in
Fig. 6.1. An improved control strategy was proposed and verified on almost the
same generator converter system [4, 5]. In 2010, another approach to eliminate the
step-up transformers of wind turbine generators was proposed [6]. A few 6-phase
generators are placed in the wind turbine nacelle to generate multiple DC supplies

190 6 Experimental Validation of 1-kV Modular Multilevel
1 1 1
Module Module Module
2 2 2

G
MV grid
k k k
Multi-coil generator MV converter
Fig.6.1Multi-coil generator-based wind turbine power generation systems [3, 7]
6-phase PMSG 6-phase PMSG 6-phase PMSG
G G G
MV converter
Converter Converter Converter Converter Converter Converter

module module module module module module
MV grid
C1 C2
Converter module
Fig.6.2Multiple generator-based wind turbine system: single-phase layout [6, 7]
for the MMC converter. All the generators are driven by the same wind turbine,
and each stator winding generates an isolated source for all the H-bridge inverter
cells of the MMC converter. The MMC converter generates medium-voltage AC
output, which can be connected to the medium-voltage network directly. A multi-
ple generator-based wind turbine system is shown in Fig.6.2.
6.1Introduction 191
380690/380690 V
Medium voltage grid (e.g., 636 kV)

1 1 1
380690 V
Inverter
G Rectifier (HF) 2 2 2

Rectifier H-bridge k k
k
(HF) inverter
Medium
Module frequency-link Modular multilevel cascaded inverter
Fig.6.3Wind turbine generator system for direct medium-voltage grid integration [1]
However, these approaches require special modular generators or multiple tra-

ditional generators to generate isolated multiple DC supplies for all the H-bridge
inverter cells of the MMC converter and introduce electrical isolation problems
between the generator and grid [8]. On the other hand, compared with the con-
ventional transformers operated at the power frequency, the high-frequency mag-
netic links have much smaller and lighter magnetic cores and windings, and thus
much lower costs. They have, therefore, already been widely used in the low-
voltage applications [911]. Operated at 1.2-kHz, the weight and size of a 3-MW,
transformer can be less than 8% of an equivalent 50-Hz unit [12]. As a result, the
advanced magnetic material-based common high-frequency magnetic link may be
the natural choice to generate multiple isolated and balanced DC supplies for all
the H-bridge inverter cells of the MMC converter. To verify the feasibility of the
new concept of generating balanced multiple sources from a single source with
a high-frequency magnetic link, a comprehensive electromagnetic analysis was
reported in [13, 14].
In this chapter, a high-frequency magnetic-link MMC medium-voltage con-
verter is experimentally validated for step-up-transformer-less direct grid integra-
tion of renewable sources. The common magnetic link generates multiple isolated
and balanced DC supplies for all the H-bridge inverter cells of the MMC converter
from a single or multiple renewable sources. The grid electrical isolation and volt-
age imbalance problems are solved through the common high-frequency magnetic
link. Figure6.3 shows the basic block diagram of the proposed high-frequency
magnetic-link MMC converter for wind turbine generator systems [1].
The available variable voltage as well as frequency AC power converts to DC
power through a controller rectifier circuit, which also controls the rotor speed of
the wind turbine generator. The DC power then converts to constant magnitude
and frequency high-frequency AC power with a high-frequency H-bridge inverter.

The single-input and multiple-output high-frequency magnetic links generate mul-
tiple isolated balanced sources for all the H-bridge inverter cells. Each module
contains a dedicated rectifier circuit, which converts high-frequency AC power to
DC power. The DC power works as the DC-link voltage of the respective H-bridge
inverter cell. To ensure fixed grid voltage, a constant output voltage of the inverter
is maintained. The inverter output is supplied to a primary winding of a multi-
winding high-frequency magnetic link. Each secondary winding is connected with
the H-bridge inverter through a bridge rectifier. The number of primary windings
depends on the number of renewable sources, and the number of secondary wind-
ings depends on the number of levels of the MMC converter. The grid electrical
isolation and voltage imbalance problems are solved through the high-frequency
magnetic link. Figure6.4 shows the detailed power converter circuit for wind tur-
bine generators with a high-frequency magnetic-link MMC converter.
A few modules are cascaded on each phase, and each module consists of a
bridge rectifier and an H-bridge inverter. The power density, switching loss, con-
trol complexity, and semiconductor cost are the significant factors, affecting the
design of high-frequency magnetic-link-based DC/DC power conversion system.
Different circuit topologies were proposed in recent years to improve the perfor-
mance of the DC/DC power converters [1517]. The dual active bridge (DAB)
converter topology has received a lot of attention due to the lower switching
losses [1820]. In the DAB converter topology, a total of four additional switches
are needed for each DC/DC conversion stage, which not only increases the sys-
tem cost but also increases control complexity as opposed to topologies, which
use a simple diode bridge rectifier at the output. The semiconductor cost and con-
trol complexity become critical for medium-voltage converter systems, where
several parallel DC/DC converters are necessary. For example, a 19-level system
requires 27 parallel DC/DC converters [7]. On the other hand, very short recov-
ery time (25ns) high-performance fast recovery power diodes are now available
on the market [21]. Thus, the fast recovery diode rectifier-based DC/DC converter
topology is considered for the proposed system. The proposed converter can be
used with most of the commercial wind turbine generators: permanent magnet
synchronous generator (PMSG), squirrel cage induction generator (SCIG), and
wound rotor synchronous generator (WRSG). In WRSG-based systems, to achieve
variable speed operation, the systems may use an extra excitation circuit, which
feeds the excitation winding of WRSG. Therefore, the controlled rectifier can be
replaced by an uncontrolled rectifier. The uncontrolled rectifier may also be used
in the PMSG-based wind turbine generator system, where the converter may be
equipped with a step-up chopper circuit. The step-up chopper adapts the rectifier
voltage to the DC-link voltage of the inverter. Controlling the inductor current in
the step-up chopper can control the generator torque and speed [22].
The high-frequency magnetic-link MMC medium-voltage converter-based
grid integration system will have the following advantages: (i) no requirement
for special or multiple generators for the wind turbine generator systems, (ii) a
wide range of MPPT operation for PV systems, (iii) an inherent DC-link voltage
6.1Introduction 193
Phase-A
H-bridge

Vdc
+
Generator
Cd
A
_
Phase-B
B
Rectifier Inverter (high C
frequency)
High-frequency link
Rectifier (high H-bridge

frequency)
Module
Phase-C
Fig.6.4Detailed power conversion circuit of high-frequency magnetic-link MMC converter for

step-up-transformer-less direct grid connection of wind turbine generators
balance due to the common magnetic link, (iv) direct gird connection without
using the step-up transformer, (v) an overall compact and lightweight system, and
(vi) an inherent minimization of the grid isolation problem through the high-fre-
quency magnetic link.
6.2Development of High-Frequency Magnetic-Link MMC

Converter
To verify the feasibility of the high-frequency magnetic-link MMC converter, a

scaled down 1.73-kVA laboratory prototype system with a modular 5-level cas-
caded converter is developed and reported in this chapter, which converts a 210V
DC (rectified generator voltage) into 3-phase 1kV rms 50Hz AC. Metglas amor-
phous alloy 2605SA1-based 10-kHz magnetic link is also developed to generate
the isolated and balanced six DC supplies for the 3-phase 5-level MMC converter.
Figure6.5 shows power conversion circuit of high-frequency magnetic-link-based
3-phase 5-level MMC converter for grid connection of wind turbine generators.
6.2.1Development of a 3-Phase 5-Level MMC Converter
A scaled down 1-kV 3-phase 5-level MMC is developed, where each phase leg
consists of two H-bridge inverter cells. Table6.1 summarizes the converter
specifications. The DC-link voltage of each H-bridge inverter cell is calcu-
lated at 367.70V. Considering the market availability, cost and suitability for
High-frequency
magnetic-link
Module A-1
Vdc
A
+
Cdc
Generator
B
Cd C
Phase-C
Phase-B
_
AC grid
Rectifier Inverter (high

Cdc
frequency)
Module A-2
Phase-A
Fig.6.5Power conversion circuit of high-frequency magnetic-link-based 3-phase 5-level MMC

converter for grid connection of wind turbine generators
6.2 Development of High-Frequency Magnetic-Link MMC Converter 195
Table6.1Converter Technical data Value

specifications in the
Line voltage (kV) 1
experiment
Line current (A) 1
Carrier frequency (kHz) 1.525
Output frequency (Hz) 50
Number of levels 5
DC-link voltage (V) 367.70
Number of phases 3
Carrier shifting technique Phase shifted
Modulation scheme SPWM and
THPWM
Fig.6.6A Photograph of
Semikron IGBT modules
SK30GH123
prototyping, the Semikron compact insulated gate bipolar transistor (IGBT)

modules SK30GH123 have been collected for the development of the 3-phase
5-level MMC converter. The module consists of four N-channel 1.2-kV IGBTs
in H-bridge connection. This module is specially designed for the power inverter,
which recommends a switching speed of up to 20kHz. Figure6.6 shows a photo-
graph of Semikron IGBT module SK30GH123. The Semikron driver SKHI 20opA
is a dual driver for the half-bridge inverter cell, which may drive the IGBT with
commutation voltage up to 1.2kV. The driver has some special features. It offers
short-circuit protection with soft turnoff, error memory, output signal with external
or automatic reset, and it supports DC-bus voltage up to 800V.
The driver circuits require 15-V DC supplies. The Semikron dual IGBT driver-
isolated power supply SKHI PS1 has been considered. It has the capability to sup-
ply energy for up to seven IGBTs with additional transformers. The IGBT module
and the driver and protection circuit (with power supply and filtering capacitors of
the respective rectifier circuit) are mounted on the top of the heat sink. Figure6.7
shows a photograph of the laboratory prototype 3-phase 5-level 1-kV MMC
r
r te
ve
er
ert
n
co
z inv
l
ve
kH
e
10
til
ul
M
Distr
ibuti
tive boar on
r o tec es d
P evic
d
FPG
A bo
ard
Drive
r circ
uit
Fig.6.7A photograph of 3-phase 5-level 1-kV MMC converter
converter. A field programmable gate array (FPGA)-based control circuit is used to

control the magnitude and phase angle, and to ensure the power quality and stabil-
ity of the system. The phase-shifted carrier and sinusoidal reference-based pulse
width modulation scheme (SPWM) and phase-shifted carrier and third harmonic
injected sinusoidal reference-based pulse width modulation scheme (THPWM) are
used in the experiment. The desired output voltage level can be generated by cas-
cading more modules on each phase. The high-level converter implies elimination
of the step-up transformer and a lower total harmonic distortion (THD) with lower
switching frequency, thereby eliminating the output filters and reducing the run-
ning cost. The high-level number attainability also allows a lower level DC-link
voltage requirement for each H-bridge module that eliminates boosters.
6.2.2Design of Driver Circuit
The Semikron driver SKHI 20op requires 15-V gate pulses. The generated gate
pulses by Xilinx FPGA XC3S500E are 3.3V, and these require stepping-up
before feeding to the driver SKHI 20op. Figure6.8 shows a photograph of Xilinx
Spartan-3E starter kit. An extra driver circuit is required to interface the FPGA
and driver SKHI 20op. The microchip dual high-speed power metal oxide field
effect transistor (MOSFET) driver TC4427A is used to design the extra driver cir-
cuit. The IC has some special features: high output current (e.g., 1.5A), gener-
ated gate pulse voltage of up to 18V, short delay time (e.g., 40ns), low output
impedance (e.g., about 7), and low supply current (e.g., about 4mA for logic 1
input and 400A for logic 0 input). Xilinx Spartan-3E board has 100-pin header
Spartan-3E starter kit
ribbon cable and header
socket
to interface with peripherals. Therefore, 100-pin socket with proper ribbon cable
is required for the connection of FPGA board with driver circuit. Figure6.9 shows
the ribbon cable connection with 100-pin socket. Figure6.10 shows the detailed
circuit diagram of the extra driver circuit. A photograph of the extra driver circuit
is shown in Fig.6.11. The 3-phase 5-level MMC converter consists of 24 IGBTs,
and the switching controller evolves 24 gate signals. Each microchip dual high-
speed power MOSFET driver TC4427A can drive two gate signals. Therefore, in
total, 12 drivers are used. Each driver SKHI 20op handles two gate signals: one
for the top IGBT of a half-bridge leg and the other for the bottom IGBT. The
driver also requires DC supply with proper ground connection. A separate header
NC NC NC NC
24 30
TC4427A
TC4427A
A1RT A1LT
10 k 10 k
23 29 A1LT
A1RB A1LB
10 k 10 k A1LB
0.047 F 0.047 F A1RT
NC NC NC NC
22 28 A1RB
TC4427A
TC4427A
A2RT A2LT A2LT
10 k 10 k
21 27 A2LB
A2RB A2LB A2RT
10 k 10 k
0.047 F 0.047 F A2RB

NC NC NC NC
30 B1LT
20 26
TC4427A
TC4427A
B1RT B1LT B1LB
10 k 10 k
B1RT
19 25
B1RB B1LB B1RB
10 k 10 k
B2LT
0.047 F 0.047 F
NC NC NC NC B2LB
7 18 12
TC4427A
TC4427A
B2RT
7 6 B2RT B2LT
10 k 10 k
2 B2RB
1 17 11
Header B2RB B2LB
C1LT
10 k 10 k
C1LB
0.047 F 0.047 F
NC NC NC C1RT
NC
16 10
TC4427A
TC4427A
C1RB
C1RT C1LT
10 k 10 k C2LT
15 9
C2LB
C1RB C1LB
10 k 10 k
C2RT
0.047 F 0.047 F
C2RB
NC NC NC NC
14 8
Terminals
TC4427A
TC4427A
C2RT C2LT
10 k 10 k
13 7
C2RB C2LB
10 k 10 k
0.047 F 0.047 F
Ground + 15 V
Fig.6.10Detailed circuit diagram of driver circuit to step-up the gate signal to 15V
and socket set, each set for a half-bridge leg with ribbon cable are used to handle
the gate signals and DC supplies of the driver circuits. A special switching control
signal distribution board is designed for proper management. The gate signals, DC
supply, and ground connection of the distribution board are depicted in Fig.6.12.
A photograph of the switching and control signal distribution board is shown in
Fig.6.13.
Fig.6.11A photograph of extra driver circuit to step-up the gate signal to 15V
6.2.3Development of High-Frequency Rectifiers with Fast

Recovery Diodes
In order to minimize the switching loss, control complexity, and semiconductor

cost, the diode bridge rectifier is considered to convert high-frequency alternating
quantity to DC quantity with fast recovery diodes. Each module is associated with
a separate rectifier. In total, six isolated bridge fast recovery rectifiers are required,
as the system consists of six H-bridge inverter cells. Figure6.14 shows the circuit
diagram of the high-frequency rectifiers. The available average voltage is about
367V. Table6.2 summarizes the technical information of the module rectifier. A
photograph of the fast recovery rectifier section is shown in Fig.6.15.
The input-voltage waveform of this rectifier is similar to the alternating train
of pulses. The rectified output waveform can be compared with the DC/DC chop-
per circuit output. The IXYS DSEE15-12CC superfast recovery (25ns) dual diode
module is considered for the development of the module rectifier of the medium-
voltage converter. If we assume the capacitor discharge rate will remain constant
at the DC level, the peak-to-peak ripple voltage, Vr(pp) can be approximated by
a waveform as shown in Fig.6.16, which has a peak-to-peak value of Vo and a
time period of Tr and is centered around the DC level. If Idc is the output rectifier
average current, fs is the source voltage frequency, Vdc is the output rectifier DC
output voltage, Doff is the diode off time or capacitor discharge time factor, and r
is the ripple factor, then the peak-to-peak and rms output voltage can be deduced
as
Idc Doff Tr Idc Doff Idc Doff
Vr(pp) = = = (6.1)
C Cfr 2Cfs
Fig.6.12Layout of the distribution board
and
Vr(pp) Idc Doff
Vr(rms) = = . (6.2)
2 3 4 3Cfs
Hence, the ripple factor can be deduced as [1]
Vr(rms) Idc Doff
r =
Vdc
= (6.3)
4 3Cfs Vdc
Fig.6.13A photograph of the switching and control signal distribution board
C dc C dc C dc
C dc C dc C dc
Fig.6.14High-frequency rectifier circuits with fast recovery diodes
Table6.2Specification of Technical data Value

module rectifier (V)
Unit voltage (rms) 375
Average output voltage 367
PIV of the diodes 375
From (6.3), the minimum value of the capacitor can be deduced as

Idc Doff
C= F. (6.4)
4 3r fs Vdc
Fig.6.15A photograph of the fast recovery rectifier section
Fig.6.16Capacitor voltage V
Rectified voltage (without capacitor)
of the fast recovery rectifier
[1] Vm
Tr t
Trc
Ts
Vo
V dc Vr(p-p)
Rectified voltage (with capacitor)
If Idc=1 A, fs=10kHz, Vdc=367V, Doff=2%, and the ripple factor is 1%,

the minimum value of capacitor, C, for each module rectifier can be calculated as
1 0.02
C= = 78.66 nF.
4 3 0.01 10 103 367
The 3188EE152T400APA1 aluminum electrolytic capacitor is used for the mod-
ule rectifier circuit.The maximum power handled by a power diode and the tem-
perature of the diode junction are related since the power dissipated by the device
causes an increase in temperature at the junction of the device. The junction
temperature, TJ, case temperature, TC, and ambient (air) temperature, TA, are
related by the device heat-handling capacity and can be presented in terms of ther-
mal-electric analogy as
JA = (JC + CS + SA ) (6.5)
and
TJ = PD JA + TA , (6.6)
where JA is the total thermal resistance (junction to ambient), JC is the transistor
thermal resistance (junction to case), CS is the insulator thermal resistance (case
to heat sink), SA is the heat-sink thermal resistance (heat sink to ambient), and PD
is the power dissipation. Using (6.5) and (6.6), the heat-sink thermal resistance can
be deduced as [1]
TJ TA
JC + CS + SA =
PD
TJ TA
SA = (JC + CS ). (6.7)
PD
From the data sheet of DSEE15-12CC, we obtain JC =1.60C/W,
CS=0.5C/W, TJ=170C, and PD is about 2W. Therefore, the heat-sink ther-
mal resistance can be calculated as
170 40
SA = (1.6 + 0.5) = 62.9 C/W.
2
Fischer Elektronik FK 245 MI 247 O clip-on heat sinks without soldering lug are
considered for this project. The heat sink is made of copper material with a ther-
mal resistance of 20.2C/W. For proper contact with semiconductor devices, the
Electrolube HTC10S non-silicone heat transfer compound is used on the contact
surface area of the heat sinks.
6.2.4Development of High-Frequency Inverter
A full-bridge high-frequency inverter is developed by using the Semikron

SK30GH123 compact insulated gate bipolar transistor (IGBT) module to gen-
erate a high-frequency square wave primary excitation voltage of the high-fre-
quency magnetic link. The Semikron SKHI 20opA with SKHI PS1 is used as an
isolated driver for the SK30GH123 module. Figure6.17 shows a photograph of
the high-frequency inverter. The Texas Instruments high-speed PWM controller
UC3825BN is used to generate the switching signals. Figure6.18 shows a pho-
tograph of high-speed PWM controller UC3825BN. A simplified version of the
switching signal generator circuit using high-speed PWM controller UC3825BN
is shown in Fig.6.19. The Texas Instruments high-speed PWM controller
Filtering
capacitors IGBT driver
circuit
Switching
signals
SK 30 GH 123 IGBT based H-
bridge inverter
(mounted on the top of the
heatsink)
Heatsink
Fig.6.17A photograph of high-frequency inverter; generating high-frequency square wave pri-

mary excitation voltage of the high-frequency magnetic link [23]
Texas Instruments high-speed
PWM controller UC3825BN
UC3825BN is a high-performance PWM controller with low start up current,

accurate oscillator frequency, leading edge blanking, latched fault logic, full-cycle
soft start, restart delay after fault, and many other features. The developed dedi-
cated switching circuit is shown in Figs.6.20 and 6.21.
The oscillator of the UC3825BN is a sawtooth that utilizes two pins; one for
the timing resistor RT and the other for the timing capacitor CT. The resistor pro-
grams the charging current to the timing capacitor via an internal current mirror
VCC
200
0.1F
15 13
VC
0.1F 16 V VCC
RF
A 11 Out A
2 NI
UC3825BN
B 14
Out B
1 INV
1nF 9
3 E/A
4.3k
Ramp 7
3.3k
NC 4 Clk CT 6
RT CT
R PGnd 12
5 Gnd
T
SS
10 8
0.1F
Fig.6.19Simplified circuit of the switching signal generator using Texas Instruments high-
speed PWM controller UC3825BN [23]
Fig.6.20A photograph of control circuit using Texas Instruments high-speed PWM controller
UC3825BN
Fig.6.21Basic internal
circuit of the Texas
Instruments high-speed
PWM controller UC3825BN
(oscillator section only)
Fig.6.22Alternative two
output topology
with high accuracy. The maximum signal high time is determined by the rising
capacitor voltage, whereas dead time is determined by the timing capacitor dis-
charge. Figure6.21 shows the internal circuit diagram of the Texas Instruments
high-speed PWM controller UC3825BN (oscillator section only).
Based on the desired maximum duty cycle, Dmax the timing resistor can be cal-
culated as
3V
RT =
10 mA (1 Dmax )
3V (6.8)
= = 3 103 or 3 k.
10 mA (1 0.90)
A B
Thigh_A
Thigh
Fig.6.23Measured switching signals; on-time calculation
Two alternative signals are generated through a T flip-flop, and output signal fre-
quency is half that of oscillator frequency, as shown in Fig.6.22. Therefore, dou-
bled oscillator frequency (compared to the output) is required for designing the
timing capacitor.
Based on the calculated value of the timing resistor and desired maximum duty
cycle, the timing capacitor can be calculated as
1.6 Dmax
CT =
RT f
(6.9)
1.6 0.90
= = 2.4 108 (F)
3 103 20 103
The complete switching signal generator circuit with UC3825BN IC is fabricated

and tested in the laboratory. The gate pulses were measured and compared with the
simulation results. The measured pulses were found to be highly consistent with the
simulation results. The signal high time (switch on) and dead time (switch off) are
measured using cursors of the oscilloscope as shown in Figs.6.23 and 6.24, and the
duty cycle is also calculated. The percentage of duty cycle, D%, can be calculated as
Thigh
D% = 100
Thigh + Tdead
(6.10)
44.8
= 100 = 90.30 %
44.8 + 4.8
The expected time period, T, and frequency, fH, of the medium frequency inverter
can also be calculated as
T = Thigh_A + Tdead_A + Thigh_B + Tdead_B
(6.11)
= 44.8 s + 4.8 s + 44.8 s + 4.8 s = 99.20 s
1 1
fH = = = 10.08 kHz.
T 99.2 106
Thigh_B
Tdead
Fig.6.24Measured switching signals; dead-time calculation
6.2.5Development of High-Frequency Magnetic Link
With the advent of new power semiconductor devices, different high magnetic sat-
uration and low power loss soft magnetic materials are conceived to reduce the
weight and volume of conventional power transformers. The grain-oriented silicon
sheet steels, which are commonly used as the core material of power frequency
transformers, are not suitable for high-frequency applications because of the heavy
eddy current loss [24]. The soft ferrites have been widely used in medium- and
high-frequency inductors and transformers due to the low price and general avail-
ability. Because of the low saturation flux density [24] (only 0.30.5T), which
would make the transformer bulky, they are not suitable for large power applica-
tions. On the other hand, the amorphous alloy and nanocrystalline materials have
excellent magnetic characteristics for high-frequency applications, such as high
permeability, high saturation flux density, and relatively low core losses. Two
commercially available amorphous and nanocrystalline materials are Metglas
and Finemet, respectively, both manufactured by Hitachi Metals, Japan. Although
Finemet has lower specific core loss than Metglas, its saturation flux density
(about 1T) is much lower than that of Metglas, which is 1.56T. Until now, many
kinds of soft magnetic alloys with high magnetic flux density combined with low
core loss have been developed [25, 26]. Taking into account the flux density, spe-
cific core loss, cost, and availability, we chose Metglas 2605SA1 stripe of 30m
thickness and 25mm width as the core material. The other parameters of 2605SA1
include mass density of 7.18g/cm3, saturation flux density of 1.56T, and spe-
cific core loss of 180W/kg, at 10kHz sinusoidal excitation of 1T. Figure6.25
shows a photograph of the Metglas alloy 2605SA1 sheet-based multi-output high-
frequency magnetic link. For small skin depth and proximity effect, the number
of layers as well as the conductor diameter should be kept as small as possible.
Moreover, the insulated strands should be twisted or braided together to equalize
the flux linkages throughout the conductors. To achieve this so as to reduce the
winding loss, a Litz wire with small number of layers should be always used in a
high-frequency magnetic link. Figure6.26 shows a photograph of Litz wire.
6.3 Experimental Testing and Performance Analysis 209
Fig.6.25Photograph of the Metglas amorphous alloy 2605SA1-based high-frequency magnetic

link [2]
Fig.6.26A photograph of Litz wire
6.3Experimental Testing and Performance Analysis
To verify the feasibility of the high-frequency magnetic-link MMC converter-based

power conversion system, a scaled down laboratory prototype 1.73-kVA test plat-
form is developed with a 5-level MMC converter, which converts 210V DC (recti-
fied generator voltage) into 3-phase 1kV rms 50-Hz AC. The Semikron compact
IGBT module SK30GH123 with Semikron-isolated driver SKHI 20opA is used to
develop the MMC converter. The 10-kHz frequency magnetic link with single pri-
mary and six secondary windings is developed to generate the balanced isolated six
DC supplies for a 3-phase 5-level MMC converter. Figure6.27 shows a photograph
Fig.6.27A photograph of the experimental setup of high-frequency magnetic-link-based MMC

converter (1kV, 1.73 kVA)
Fig.6.28Measured gate pulses of the high-frequency inverter
of the experimental test platform. The gate pulses for the high-frequency inverter
were measured, as shown in Fig.6.28, and compared with the simulation results.
The measured pulses were found to be highly consistent with the simula-
tion results. The primary and secondary voltages of the high-frequency mag-
netic link were measured and the voltage transformation ratio was calculated, as
listed in Table6.3, and found to be highly consistent with the theoretical values.
Figure 6.29 shows the voltage waveforms of the primary and secondary sides of
the high-frequency magnetic link. Tektronix DPO 2024 digital phosphor oscillo-
scope with the P5200 high-voltage differential probe and Tektronix TCPA300 cur-
rent probe were used to observe the voltage and current waveforms.
Table6.3Voltage Windings A B C D E F
transformation ratios against
Ratios 1.781 1.780 1.779 1.780 1.780 1.781
primary winding [27]
Variation 0.00 0.05 0.11 0.05 0.05 0.00
(%)
Fig.6.29Measured primary and secondary voltage waveforms of high-frequency magnetic link
Fig.6.30A photograph of Tektronix digital phosphor oscilloscope DPO 2024
Figure 6.30 shows a photograph of Tektronix digital phosphor oscilloscope

DPO 2024, which is a 200-MHz oscilloscope with four channels. The current
probe TCPA300 can measure AC and DC currents. The probe converts the sensed
Fig.6.31A photograph of Tektronix current probes
AC or DC current into a proportional voltage signal that can be measured directly

with an oscilloscope. The TCPA300 current probe has a few notable features, such
as capability of DC and AC current measurements up to 750-A peak, high sensi-
tivity, and one-button auto-balancing (no adjustment needed to match the probe).
Figure6.31 shows a photograph of Tektronix current probe TCPA300. To use the
full dynamic range of the probe/amplifier combination, the oscilloscope must be
capable of displaying a vertical scale factor of 1mV/div to 1V/div.
Tektronix high-voltage differential probe P5200 can be used with any oscillo-
scope to safely make measurements of floating circuits. The probe converts float-
ing signals to low-voltage ground-referenced signals that can be displayed safely
and easily on any ground-referenced oscilloscope. The P5200 high-voltage dif-
ferential probe requires an oscilloscope with grounded inputs. Figure6.32 shows
a photograph of Tektronix high-voltage differential probe P5200. Tektronix high-
voltage differential probes have a few remarkable features and benefits, such as
high bandwidths up to 100MHz, high-voltage-handling capacity up to 2,200V
common rms and up to 5,600V differential, and over range indicator.
The outputs of each secondary voltage and the DC-link voltages (after the filter cir-
cuit) were found to be approximately equal, which can serve satisfactorily as the iso-
lated and balanced DC sources for the MMC converter. The total losses (core loss plus
copper loss) of all secondary windings of the Metglas amorphous alloy 2605SA1-based
high-frequency magnetic link measured at different excitation frequencies ranging from
6 to 12kHz are almost the same. Such a similarity of characteristics is obligatory to
generate multiple balanced DC sources for all the H-bridge inverter cells of the MMC
converter. A Voltech PW3000A universal power analyzer was used to measure the total
Fig.6.32A photograph of Tektronix high-voltage differential probe P5200
Fig.6.33A photograph of Tektronix power analyzer PA4000
loss. In March 2013, Voltech transferred all its power analyzer technology to Tektronix.
Recently, Tektronix developed power analyzer PA4000, which delivers consistently
accurate measurements, even with challenging power waveforms. Figure6.33 shows a
photograph of Tektronic power analyzer PA4000. The PA4000 supports input voltage
and current up to 1,000V rms and 30 A rms, respectively, with a measurement band-
width up to 1MHz. From the oscilloscope data, the copper loss of each winding at a
different frequency ranging from 6 to 12kHz was calculated by using the DC resist-
ances (0.024 for the primary and 0.16 for the secondary), since the AC/DC resist-
ance ratios, Kr, in this design is almost unity due to the use of Litz wires.
Fig.6.34A photograph of Fluke infrared temperature probe 80T-IR
High-frequency magnetic-link core temperature was observed and maintained

in the range of 40120C during the experimentation, which ensures stable maxi-
mum flux density. The Fluke infrared temperature probe 80T-IR and Flir infrared
thermal imaging camera i7 were used to measure the temperature of the high-fre-
quency magnetic link. The Fluke 80T-IR is a non-contact temperature measure-
ment accessory for use with a test instrument capable of measuring DC voltage in
the millivolt range such as digital multimeter. Figure6.34 shows a photograph of
Fluke infrared temperature probe 80T-IR. The probe has a temperature-handling
capacity up to 260C, with a basic accuracy of 3%. Temperature is measured by
pointing the probe at the surface to be measured and by reading the temperature on
the digital multimeter display.
Infrared radiation is emitted by high-frequency magnetic link above a tem-
perature of 273C. Although the human eye cannot detect infrared radiation, a
thermal imaging camera can, and takes picture of high-frequency magnetic link to
show the amount of heat that is emitting. The Flir thermal imaging camera i7 pro-
duces non-contact temperature measurements as thermal images. The Flir i7 has a
few remarkable features, such as high accuracy and sensitivity of 2% and 0.1C,
focus-free lens, different measurement modes (spot, area, and isotherm), and tem-
perature range up to 250C. Figure6.35 shows a photograph of Flir thermal imag-
ing camera i7. With the reporting software included in the camera package, it is
possible to create reports, analyze, and document the image easily.
The switching scheme is implemented with XC3S500E FPGA. The PWM
gate pulses from FPGA are 3.3V, which is insufficient for the IGBT drivers used.
Driver IC TC4427A is used to increase the voltage level to the desired value
(TC4427A can support up to 18V). The gate pulses for an H-bridge module of
a 5-level modular cascaded converter are illustrated in Fig.6.36. The measured
Fig.6.35A photograph of Flir thermal imaging camera i7
Left top
Left bottom
Right top Deadband
Right bottom
15 V, 1.525 kHz
Fig.6.36Measured gate pulses of four IGBTs in the top module in phase A
switching frequency of the switching signals is 1.525kHz, which precisely

matches the theoretical result. Pulse patterns are also verified with theoretical
results as well as MATLAB/Simulink and Xilinx ISE simulation results and they
are found highly consistent. Gate pulses are applied to the switching devices of the
Fig.6.37Measured phase voltage of the prototype system (phase voltage before LC filter
circuit)
Fig.6.38Zoomed phase voltage of the prototype system (phase voltage before LC filter circuit;
with 1.525-kHz PWM frequency
converter and the output phase voltage and line current are measured, and they are
found to be satisfactory. Figure6.37 depicts the output phase voltage of the proto-
type converter. Each level of the output voltage contains a number of PWM pulses.
Figure6.38 plots the zoomed (zoom factor=8) output phase voltage of the pro-
totype system. The line voltages of the prototype converter were measured; they
were found to be highly consistent with the simulation results.
The SPWM and THPWM schemes were implemented practically and perfor-
mances were verified by observing the line voltage waveforms, and they were
found to be highly consistent with the theoretical and simulation results. The LC
filter circuit is designed with the 3-mH MTE RL 00401 reactor and 6-F RS
Fig.6.39A photograph of MTE reactor to design LC filter
Fig.6.40A photograph of RS AC capacitor
MR-P-MC-S-NF capacitors. Figure6.39 shows a photograph of MTE reactor. The

MTE reactors provide stable inductance, even at 150% of their current rating;
these reactors still have 100% of their nominal inductance. The RS-metallized
polypropylene motor run AC capacitors can be used to design LC line filter cir-
cuit. The RS MR-P-MC-S-NF is a 440-V AC screw mount 35-mm-diameter poly-
propylene film capacitor of 6F with a tolerance of 10%. A few capacitors can
be connected in series to obtain the required voltage rating. Figure6.40 shows a
Fig.6.41A photograph of LC line filter
Fig.6.42Measured line voltage with SPWM of the prototype system before filter
photograph of RS capacitor MR-P-MC-S-NF. Figure6.41 shows a photograph

of LC line filter circuit of high-frequency magnetic-link MMC converter-based
power conversion system.
In each phase, three capacitors are connected in series, which provides equiva-
lent 2-F capacitance with a rated voltage at 1.32kV. The line voltage before the
filter circuit with SPWM scheme is illustrated in Fig.6.42, and this shows that the
SPWM scheme gives a higher THD, which is about 27%. The performances are
also observed with the filter capacitors connected in Y and . The line voltages
after filter are shown in Fig.6.43, when the filter capacitors were connected in Y.
Significant improvement was observed when filter capacitors were connected in
Fig.6.43Measured line voltage with SPWM of the prototype system after the filter when the
filter capacitors are connected in Y
Fig.6.44Measured line voltage with SPWM of the prototype system after the filter when the
filter capacitors are connected in
as the THD was reduced from 4.5% to about 3.2%. After filter circuit, the line
voltages are depicted in Fig.6.44, when the filter capacitors are connected in .
The THPWM scheme provides much better results than the SPWM, which was
found to be highly consistent with the simulation results. The line voltage wave-
form before the filter circuit with THPWM is illustrated in Fig.6.45. Figure6.46
plots the zoomed (zoom factor=6) output line voltage with THPWM. The fre-
quency spectrum of the measured line voltage (before the filter circuit) is shown
in Fig.6.47, which contains about 19% THD. The line voltages were also meas-
ured after the filter circuit, and they were found to be consistent with the simulation
Fig.6.45Measured line voltage (before filter circuit) of the prototype 1-kV modular multilevel
cascaded converter with THPWM
Fig.6.46Zoomed measured line voltage (before filter circuit) of the prototype 1-kV modular
multilevel cascaded converter with THPWM
results. The measured line voltages after the filter circuit are shown in Fig.6.48.
Figure6.49 plots zoomed 3-phase line voltage after the filter circuit with THPWM.
As measured, after the filter circuit, the output voltage waveform contains about
2.75% THD. The frequency spectrum of the line voltage after the filter circuit is
shown in Fig.6.50. The line currents were measured and compared with theoretical
as well as simulation results, and they were found to be almost the same.
If Pc_inv is the conduction loss and Psw_inv is the switching loss, the losses in the
inverter section of the proposed converter can be described as
Ploss_inv = Pc_inv + Psw_inv (6.12)
0.06
Normalized amplitude 19.3% THD

0.04
0.02
0
0 10 20 30 40 50
Frequency (kHz)
Fig.6.47Frequency spectrums of line voltages (before the filter circuit)
Fig.6.48Measured line voltage (after filter circuit) of the prototype 1-kV modular multilevel
cascaded converter with THPWM
The switching losses in a two-level and multilevel inverters were approximated by

[29]
Psw_inv_2 = (AIr + BI2r )fc_2 (6.13)
and
Psw_inv_M = (AIr + BI2r )fc_M . (6.14)
where fc_M and fc_2 are the carrier frequencies of multilevel and 2-level inverters,
respectively. If A and B are assumed to be the same for both inverter systems, the
switching losses in a 2-level inverter would be fc_2/fc_M times higher than that of a
multilevel inverter.
Fig.6.49Zoomed measured line voltage (after filter circuit) of the prototype 1-kV modular
multilevel cascaded converter with THPWM
0.015
0.01 2.75% THD
0.005
0
0 10 20 30 40 50
Frequency (kHz)
Fig.6.50Frequency spectrum of line voltages (after the filter circuit)
The conduction losses in an IGBT and in an antiparallel diode were represented

by [28]

1 1 ma 3 ma
Pc_sw = Ir Vt + pf + Ir2 RCE + pf . (6.15)
2 4 8 3
and

1 1 ma 3 ma
Pc_D = Ir Vf pf + Ir2 RAK pf . (6.16)
2 4 8 3
where ma is the amplitude modulation index, pf the power factor of the current, Ir
the device current, and Vt and Vf are the voltage drops at zero current condition,
and RCE and RAK the forward resistances of IGBT and diode, respectively, which
can be collected from the manufacturers data sheets. The total conduction loss in
a 3-phase 2-level inverter can be approximated as
Pc_inv_2 = 6(Pc_sw + Pc_D ). (6.17)
The inverter section of the medium/high-frequency-link MMC converter consists
of a series of H-bridge inverter cells in a cascaded connection. Therefore, the total
conduction losses of an m-level inverter can be approximated as
Pc_inv_M = 6(m 1)(Pc_sw + Pc_D ). (6.18)
The device commutation voltage of m-level inverter is (m1) times lower than
that of a device in the 2-level inverter. The on-state saturation voltage of an IGBT
and forward voltage of a diode are highly dependent on device voltage ratings. For
these reasons, although the multilevel inverter uses a large number of devices, the
total conduction loss is similar to that of a 2-level inverter with the same power
conditions. Due to lower switching losses, the total loss of a multilevel inverter is
much lower than that of a 2-level inverter with the same power conditions.
The efficiency of each section and for the whole system of the high-frequency
magnetic-link MMC converter can be calculated from
Pout
= . (6.19)
Pin
and percentage of efficiency can be calculated from
Pout
% = 100. (6.20)
Pin
As calculated, the multilevel inverter section of the high-frequency magnetic-link
MMC converter gives about 80% efficiency at 20% rated load. The efficiency
reaches 90% at about 50% rated load. Almost 95% efficiency was measured with
the fully rated power condition. The overall efficiency of the whole system was
also calculated. The overall efficiency of the whole system was much lower than
that of the multilevel inverter section due to the significant power losses in the
10-kHz inverter, high-frequency magnetic link, and fast recovery rectifiers. About
58% efficiency was evaluated with the 20% rated power. The overall efficiency
increases to 70% at about 50% rated power. The full load overall efficiency of
the proposed system was measured at 76%. Compared with the traditional 2-level
converter (inverter section only), about 15% lower efficiency was evaluated.
Although the high-frequency magnetic-link MMC system gives 15% lower effi-
ciency than that of a 2-level converter, it is still similar to a 2-level converter-based
traditional system, because the traditional system uses two main additional com-
ponents, i.e., line filter, and the power frequency step-up transformer. The step-
up transformers and harmonic neutralizing filters produce about 50% of the total
losses and occupy up to 40% of the system volume [29]. With the high-frequency
magnetic-link MMC converter, the elimination of the heavy and large step-up
transformer and line filter will enable large cost savings in terms of the installation,
running, and maintenance of renewable power generation systems. No matter what

type of power transmission system, the medium-voltage converter is equally appli-
cable in the high-voltage AC (HVAC) or high-voltage DC (HVDC) system.
6.4Summary
The multiple secondary windings high-frequency magnetic link can be a good

solution to provide multiple isolated and balanced DC supplies for the modular
multilevel cascaded converter and to ensure electrical isolation between grid and
renewable generation systems. The modern FPGA-based switching controller
may solve the control complexity of MMC converters. In this chapter, the high-
frequency magnetic-link MMC converter-based medium-voltage system has been
validated by a scaled down 3-phase 1-kV system with a 5-level MMC converter
topology. The same concept can be used to model the 11-kV system with 19-level
MMC converter topology. The switching scheme and design technique can be
used for any other converter with minor modifications in the software environ-
ment. Although the high-frequency magnetic-link MMC converter-based system
does not improve the efficiency significantly, the elimination of the heavy and
large-size step-up transformer and line filter will enable large cost savings in the
installation, running, and maintenance of renewable generation systems. These
design and implementation techniques would have great potential for the devel-
opment of new medium-voltage converters for renewable generation systems and
smart micro-grid applications.
References
Electron 29(8):41674182
3. Ng CH, Parker MA, Ran L, Tavner PJ, Bumby JR, Spooner E (2008) A multilevel modu-
lar converter for a large, light weight wind turbine generator. IEEE Trans Power Electron
23(3):10621074
4. Yuan X, Li Y, Chai J, Ma M (2009) A modular direct-drive permanent magnet wind generator
system eliminating the grid-side transformer. In: Proceedings of the 13th European confer-
ence on power electronics and applications, Barcelona, Spain, 810 Sept 2009, pp 17
5. Yuan X, Chai J, Li Y (2012) A transformer-less high-power converter for large permanent
magnet wind generator systems. IEEE Trans Sustain Energy 3(3):318329
6. Deng F, Chen Z (2010) A new structure based on cascaded multilevel converter for variable
speed wind turbine In: Proceedings of the 36th annual conference on IEEE industrial elec-
tronics society, Glendale, AZ, USA, 710 Nov 2010, pp 31673172
References 225
8. Islam MR, Guo YG, Zhu JG (2014) A novel medium-voltage converter system for compact
and light wind turbine generators In: Proceedings of Australasian universities power engi-
neering conference (AUPEC2012), Bali, Indonesia, 2629 Sept 2012, pp 16
9. Jia L, Mazumder SK (2012) A loss-mitigating scheme for DC/pulsating-DC converter of a
high-frequency-link system. IEEE Trans Ind Electron 59(12):45374544
10. Huang R, Mazumder SK (2009) A soft-switching scheme for an isolated DC/DC converter
with pulsating DC output for a three-phase high-frequency-link PWM converter. IEEE Trans
Power Electron 24(10):22762288
11. Mazumder SK, Rathore AK (2011) Primary-side-converter-assisted soft-switching scheme
for an AC/AC converter in a cycloconverter-type high-frequency-link inverter. IEEE Trans
Ind Electron 58(9):41614166
12. Prasai A, Yim JS, Divan D, Bendre A, Sul SK (2008) A new architecture for offshore wind
farms. IEEE Trans Power Electron 23(3):11981204
13. Islam MR, Guo YG, Zhu JG (2013) A medium frequency transformer with multiple second-
ary windings for medium voltage converter based wind turbine power generating systems. J
Appl Phys 113(17):17A32417A324-3
14. Islam MR, Guo YG, Lin ZW, Zhu JG (2014) An amorphous alloy core medium frequency mag-
netic-link for medium voltage photovoltaic inverters. J Appl Phys 115(17):17E710-117E710-3
15. Xu D, Chen M, Luo M (2004) Transformer secondary leakage inductance based ZVS dual
bridge DC/DC converter. IEEE Trans Power Electron 19(6):14081416
16. Mezaroba M, Martins DC, Barbi I (2007) A ZVS PWM half-bridge voltage source inverter
with active clamping. IEEE Trans Ind Electron 54(5):26652672
17. Chen G, Lee YS, Hui SYR, Xu D, Wang Y (2000) Actively clamped bidirectional flyback
converter. IEEE Trans Ind Electron 47(4):770779
18. Kheraluwala MN, Gascoigne RW, Divan DM, Baumann ED (1992) Performance characteri-
zation of a high-power dual active bridge. IEEE Trans Ind Appl 28(6):12941301
19. Krismer F, Kolar JW (2012) Closed form solution for minimum conduction loss modulation
of DAB converters. IEEE Trans Power Electron 27(1):174188
20. Krismer F, Kolar JW (2012) Efficiency-optimized high-current dual active bridge converter
for automotive applications. IEEE Tran Ind Electron 59(7):27452760
21. Islam MR, Guo YG, Zhu JG (2011) H-bridge multilevel voltage source converter for direct
grid connection of renewable energy systems. In Proceedings of IEEE power & energy soci-
ety innovative smart grid technology conference, Perth, Australia, 1316 Nov 2011, pp 17
22. Islam MR, Guo YG, Zhu JG (2013) Power converters for wind turbines: current and future devel-
opment. In: Mendez-Vilas A (ed) Materials and processes for energy: communicating current
research and technological developments, Energy book series-2013 edition. Spain, pp 559571
23. Islam MR, Guo YG, Zhu JG, Lu H, Jin JX (2013) Medium frequency-link power conversion
for high power density renewable energy systems. In: Proceedings of IEEE 2013 interna-
tional conference on applied superconductivity and electromagnetic devices (ASEMD2013),
2527 Oct 2013, Beijing, China, pp 102106
24. Agheb E, Bahmani MA, Hoidalen HK, Thiringer H (2012) Core loss behavior in high frequency
high power transformers-II: arbitrary excitation. J Renew Sustain Energy 4(3):033113033113-11
25. Zhang Z, Sharma P, Makino A (2012) Role of Si in high Bs and low core-loss Fe85.2B10-
XP4Cu0.8SiX nano-crystalline alloy. J Appl Phys 112(10):103902103902-8
26. Makino A, Kubota T, Yubuta K, Inoue A, Urata A, Matsumoto H, Yoshida S (2011) Low core
losses and magnetic properties of Fe85-86Si1-2B8P4Cu1 nanocrystalline alloy with high B
for power applications. J Appl Phys 109(7):07A30207A302-5
27. Islam MR, Guo YG, Zhu JG (2012) A medium-frequency transformer with multiple secondary
windings for grid connection through H-bridge voltage source converters. In: Proceedings of inter-
national conference electrical machines and systems, Sapporo, Japan, 2124 Oct 2012, pp 16
28. Tolbert LM, Peng FZ, Habetler TG (1999) Multilevel converters for large electric drives.
IEEE Trans Ind Appl 35(1):3644
29. Peng FZ, Lai JS, McKeever JW, Coevering JV (1996) A multilevel voltagesource inverter
with separated DC sources for static var generation. IEEE Trans Ind Appl 32(5):11301138
Chapter 7
Design and Analysis of 11- and 33-kV
Modular Multilevel Cascaded Converters
Abstract The high number of levels of modular multilevel cascaded (MMC) con-
verters enables the direct connection of the renewable generation units to the
medium-voltage grid and improvement of the output power quality. The component
number and control complexity increase linearly with the increase in the number of
levels. On the other hand, the distortion in generated output voltage and semicon-
ductor cost of the converter decrease dramatically with the increase in the converter
number of levels. As the number of levels increase, it is possible to use lower switch-
ing frequencies, even the fundamental switching frequency, which can significantly
reduce the switching losses. Therefore, the optimal selection of the number of con-
verter levels is important for the best performance/cost ratio of the medium-voltage
converter systems and this is the central content of this chapter. In this chapter, an
11-kV system and a 33-kV system are designed and analyzed taking into account the
specified system performance, control complexity, and cost and market availability
of the power semiconductors. It is found that the 19-level and 43-level converters are
the optimal choice for the 11- and 33-kV systems, respectively. Besides the design
and analysis of medium-voltage converters, the traditional low-voltage converters
with power frequency step-up transformers are also discussed.
Keywords Medium voltage Modular multilevel cascaded Number of levels

Design and optimizationConventional converters/invertersRenewable gen-
erations Step-up transformers
7.1Introduction
Different power electronic converters have been developed using the conventional
topologies to fulfill the requirements of renewable generation systems. However, it
is hard to connect the traditional converters to the grids directly, as the distortion
in generated output voltages is high and a single switch cannot stand the grid volt-
age level [1, 2]. Therefore, the traditional low-voltage converters/inverters have
typically used step-up transformer to step up the voltage to the grid voltage level.

228 7 Design and Analysis of 11- and 33-kV Modular
7.1.1Traditional Converters for Wind Power

Generation Systems
Many power semiconductor vendors such as Semikron, ASEA brown boveri (ABB),
IXYS, and Mitsubishi Electric produce devices specially designed for the diode rec-
tifier-based converter and back-to-back converter (modular form) for wind turbine
generator systems. All of the devices are in a single pack, which reduces the cost and
complexity of the power conditioning system. Semikron developed modules IGDD6-
4-426-D3816-E1F12-BL-FA and SKS 660F B6U+E1C+B6CI 250 V06 for the
diode rectifier-based power conditioning systems [3, 4]. This type of power converter
is normally used in a wound rotor synchronous generator (WRSG) or a permanent
magnet synchronous generator (PMSG)-based wind power generation system instead
of an induction generator. According to the internal circuit configuration, module
IGDD6-4-426-D3816-E1F12-BL-FA is suitable for WRSG-based wind turbine gen-
erator systems and module SKS 660F B6U+E1C+B6CI 250 V06 is suitable for
PMSG-based wind turbine generator systems. In a WRSG-based system, to achieve
variable-speed operation, the system uses an extra excitation circuit, which feeds the
excitation winding of WRSG. Figure7.1 shows the Semikrons SEMIKUBE con-
verter module IGDD6-4-426-D3816-E1F12-BL-FA. The internal electrical circuit of
the Semikrons SEMIKUBE converter module IGDD6-4-426-D3816-E1F12-BL-FA
is shown in Fig.7.2. The PMSG-based wind turbine generator systems are equipped
with a step-up chopper circuit. The step-up chopper adapts the rectifier voltage to
the DC-link voltage of the inverter. Controlling the inductor current in the step-up
chopper can control the generator torque and speed. The diode rectifier with step-up
chopper-based power conditioning system is illustrated in Fig.7.3. In this converter
system, the grid-side inverter controls the active and reactive power delivered to the
grid. Mitsubishi Electric developed the IGBT module CM100MXA-24S with this
converter topology which can be used for wind turbine generator systems [5]. The
module is recommended for the collector current of100A and the collectoremitter
Semikrons SEMIKUBE
converter module IGDD6-4-
426-D3816-E1F12-BL-FA
7.1Introduction 229
Rectifier Inverter
Fig.7.2Electrical circuit of Semikrons SEMIKUBE converter module IGDD6-4-426-D3816-

E1F12-BL-FA
Rectifier Inverter
L
Q C
Fig.7.3Power conditioner circuit for PMSG-based wind turbine generator systems
Fig.7.4A photograph of MITSUBISHI ELECTRIC IGBT module CM100MXA-24S
voltage of 1,200V. Figure7.4 shows a photograph of Mitsubishi Electric IGBT mod-

ule CM100MXA-24S. By adding the DC-link capacitor and chopper inductor, the
module can be used for both the WRSG- and PMSG-based wind turbine generator
systems. Although it is a cost-effective choice, the power quality and unidirectional
power flow capability of this converter system are critical to the selection.
When the diode rectifier and chopper circuit is replaced by controlled rectifier,
the controlled rectifier may have the bidirectional power flow capability, which
is not possible in the diode rectifier-based power conditioning system. Moreover,
the controlled rectifier strongly reduces the input current harmonics and harmonic
power losses. The grid-side converter enables the control of the active and reactive
power flows to the grid and keeps the DC-link voltage constant, improving the
output power quality by reducing the total harmonic distortion (THD). The gen-
erator-side converter works as a driver, controlling the magnetization demand and
the desired rotor speed of the generator. The decoupling capacitor between grid-
side converter and generator-side converter provides independent control capabil-
ity of the two converters. Due to some special features, this converter topology
has received great attention recently. Many power semiconductor manufacturers,
such as Semikron, ABB, Hitachi, Siemens, IXYS, and Mitsubishi Electric, pro-
duce components in module forms, suitable for this converter, which makes the
converter compact and lightweight. The back-to-back converter can be used for
PMSG and squirrel cage induction generator (SCIG)-based wind power genera-
tion systems. Siemens employs back-to-back converter for power conditioning of
SCIG-based wind turbine generator systems. Semikrons SEMISTACK RE mod-
ule SKS B2 140 GDD 69/12 U-A11 MA PB is a back-to-back converter typically
used for power conditioning of synchronous and doubly fed generator-based wind
power systems, as well as in central solar PV inverters [6]. Figure7.5 shows a
photograph of SEMISTACK RE. The modules maximum ratings are the input/
Semikron SEMISTAK RE
converter module SKS B2
140 GDD 69/12 U-A11 MA
PB
7.1Introduction 231
Inverter Inverter
Fig.7.6Electrical power circuit of Semikron SEMISTAK RE converter module SKS B2 140

GDD 69/12 U-A11 MA PB
output current of 1,400 A, input/output voltage of 760V, and switching frequency

of 5kHz, respectively. Up to four SEMITACK RE converters can be connected in
parallel to support applications of capacity up to 6MVA. SEMISTACK RE uses
the latest generation of Semikrons SKiiP 4 intelligent power module family which
integrates the power components, driver, and heat sink in a single case. Figure7.6
shows the internal electrical circuit of SEMISTACK RE. In addition to bidirec-
tional power flow capability, back-to-back converters have some other remarkable
features. For example, the DC-link voltage can be boosted to a level higher than
the amplitude of the grid line-to-line voltage in order to achieve full control of the
grid current and the capacitor between the inverter and rectifier makes it possible
to decouple the control of the two inverters, allowing the compensation of asym-
metry on both the generator side and the grid side. On the other hand, the back-
to-back converter has some critical disadvantages. The heavy and large DC-link
capacitor increases the costs and reduces the overall lifetime of the system. Since
the back-to-back converter consists of two inverters, the switching losses might be
even more pronounced, and the combined control of the controlled rectifier and
inverter is quite complicated.
7.1.2Traditional Converters for Solar PV Power Plants
Semikrons SEMIKUBE modules IGD-8-326-E1F12-BH-FA and IGD-8-426-

E1F12-BH-FA are three-phase inverters typically used for solar PV applications.
Figure7.7 shows a photograph of the Semikrons SEMIKUBE solution family, cov-
ering maximum power of 110, 220, 400, and 900kW (from the left), for output volt-
ages up to maximum 460V. The SEMIKUBE is UL508c recognized and fulfills the
EN 50178 standard. The IGBTs of SEMIKUBE may allow a maximum switching
frequency of 15kHz. Figure7.8 shows the power circuit of SEMIKUBE.
The ABB central inverters are specially designed for the medium-scale PV
power plants. The PVS800 version is a 3-phase inverter with a power capacity in
the range of 1001,000kW [7]. Figure7.9 shows a photograph of ABB central
PV inverter PVS800. The transformer steps up the inverter output voltage from
Fig.7.7A photograph of Semikrons SEMIKUBE converter modules designed for 110-, 220-,
400-, and 900-kW (from the left) solar PV power converters
Inverter
Fig.7.8Electrical equivalent circuit of Semikrons SEMIKUBE converter modules
Fig.7.9ABB central solar PV inverter PVS800

7.1Introduction 233
Fig.7.10Housing of ABB central solar PV inverter PVS800 with step-up transformer
300400V AC to the distribution grid voltage level (e.g., 636kV). Figure7.10

shows the assembly of PVS800 inverter with step-up transformer. Siemens also
developed the SINVERT PVS inverter for the medium-scale PV power plants. The
AC output voltage and power capacity of the PVS version inverters are in the range
of 288370V and 500630kW, respectively [8]. Figure7.11 shows a photograph
Fig.7.11Siemens PV inverter SINVERT PVS; the left two cabinets are the DC cabinets and the
other two the AC cabinets
Fig.7.12Function units of Siemens PV inverter SINVERT PVS
of Siemens central PV inverter SINVERT PVS. The cabinet holds touch panel
which consists of indicators and key-operated switch. Figure7.12 shows the func-
tion units of Siemens PV inverter SINVERT PVS, where the markings of 112
indicates the modules for 1,000V option, modules for PV array grounding option,
modules for options, DC connectors, DC terminal compartment of the PV array
and LV HRC fuses, inverter module (power unit), connection to AC cabinet, com-
munication area, AC filter, cooling ventilators, AC contactor, and circuit breaker
for isolating the AC system and overvoltage protection, respectively.
7.1.3Power Transformers in Traditional Converter-Based

Renewable Generation Systems
Installation of megawatt-scale renewable power generation unit, such as wind

farms and solar PV power plants, is becoming common in recent years. The volt-
age levels of the wind generator and PV inverter are usually in the range of 380
690V and 280370V, respectively [911]. To integrate the scattered wind turbine
generators and solar PV arrays into a medium-voltage grid before the voltage step-
up for long-distance transmission, a step-up transformer is typically used to step
up the voltage. In a wind turbine generation system, this transformer is usually
installed inside the nacelle together with other equipment, such as the generator
7.1Introduction 235
and power converter at a height of about 80 meters, whereas in PV power plants,

these transformers are typically installed together with low-voltage PV inverter.
Several vendors have been developing special transformers for renewable
power generation systems aiming to reduce the size and weight with the consider-
ation of environmental concern. ABB has been delivering worldwide vacuum cast
coil transformers and liquid-filled transformers for wind farm and solar PV power
plant applications, and has already installed more than 5,000 units for wind farm
applications. The cast coil dry-type transformers are nonflammable and moisture
proof, featuring a discharge-free solid isolation system. The volume and weight of
a 2.5-MVA vacuum cast coil transformer are about 8.5m3 and 6,200kg, and the
no-load and full-load losses are 5.8 and 25kW, respectively [1, 12]. Figure7.13
shows a photograph of ABB cast coil dry-type three-phase transformer. However,
the dry-type transformers can be sensitive to water, microcracks, temperature vari-
ations, and pollution which can block cooling ducts. ABB cast coil dry-type trans-
formers are typically installed in different locations of wind turbines, such as in
nacelle and tower. Figure7.14 shows the possible locations to install the cast coil
dry-type transformers.
In addition to the ABBs existing high-efficiency transformers, ABB has also intro-
duced new liquid-filled amorphous metal (AM) transformers which can offer even
more energy savings and are suitable for renewable power generation systems. Since
Fig.7.13ABB cast coil dry-type transformer

Fig.7.14Housing of ABB
cast coil dry-type transformer
wind farms and solar PV power plants are intermittent power sources, it is of great
importance to minimize transformer no-load losses. A liquid-filled 2-MVA trans-
former is about 5.7m3 in volume and 4,530kg in weight with 870kg of liquid as
the coolant and insulator [1, 13]. The no-load and full-load losses are approximately
3.2 and 21kW, respectively. Figure7.15 shows a photograph of an ABB liquid-filled
transformer. In offshore wind farms, this transformer is usually installed inside the
nacelle of the wind turbine. Ground-mounted liquid-filled transformers are typically
used for solar PV plants and onshore wind farm applications. Figure7.16 shows a
photograph of a ground-mounted liquid-filled transformer for onshore wind farm.
The SLIM transformer developed by Pauwels is compact in size and has low
no-load losses, typically half of those of the dry-type transformers. For example, a
33/0.69-kV, 2.6-MVA SLIM transformer has a no-load loss of 2.6kW and a full-
load loss of 22.5kW [1, 14, 15]. Figure7.17 shows a photograph of a Pauwels
SLIM transformer. This transformer is typically lifted through the tower door.
7.1Introduction 237
Fig.7.15ABB liquid-filled transformer
Figure 7.18 shows the transformer lifting technique. The conventional liquid-
immersed transformer uses cellulose and mineral oil, whereas SLIM uses a high-
temperature aramid insulation material called NOMEX and a silicone liquid. The
use of about 900kg or more silicone fluid as the coolant and insulator in these
transformers ensures a high degree of fire safety (the flash points of mineral oil
dielectric and silicone fluid are about 150 and 360C, respectively). These insu-
lation materials are usually chosen on the basis of reliability, performance,
safety, and environment concerns. Especially, the liquid degradation has received
a lot of attention. For example, according to European requirements, more than
65% of the liquids must be degraded in 28days. This concern can be overcome
by using fully biodegradable liquids. M&I Materials Ltd. developed the syn-
thetic ester MIDEL 7131, which is the preferred dielectric fluid for transformers
in environmentally sensitive locations and already recognized as readily biode-
gradable. Pauwels developed a new transformer by optimally using the synthetic
ester MIDEL 7131 in combination with the high-temperature NOMEX insulation
system to obtain the compactness and reliability of the SLIM transformer. This
resulted biodegradable liquid-based SLIM transformer is called Bio-SLIM. The
volume and weight of a 20/0.69-kV, 2.3-MVA Bio-SLIM transformer are about
Fig.7.16Housing of liquid-filled transformer in onshore wind farm
Fig.7.17Pauwels SLIM
transformer
7.1Introduction 239
Fig.7.18A Pauwels transformer being lifted through the door of a wind turbine tower during
installation
Fig.7.19Pauwels Bio-
SLIM transformer for
offshore applications
4m3 and 5,040kg, respectively [1, 14, 15]. The Bio-SLIM transformer is physi-
cally similar to SLIM, but painted blue to distinguish it from SLIM and to link to
marine and water applications. Figure7.19 shows a photograph of the Bio-SLIM
transformer. Table7.1 summarizes the data of various transformers.
Table7.1Transformer data
Transformer Rating Size Weight No-load loss Full-load Liquid
type (MVA) (m3) (kg) (kW) loss (kW) (kg)
Cast coil 2.50 8.50 6,200 5.80 25.00
Liquid filled 2.00 5.70 4,530 3.20 21.00 870
SLIM 2.30 4.00 5,040 2.60 22.50 900
Although industries have been trying to develop environmentally friendly

compact and lightweight transformers for renewable generation systems, still
the transformers are heavy and large. This heavy and large-size step-up trans-
former significantly increases the weight and volume of the nacelle as well as the
mechanical stress of the tower. These penalties are critical in offshore applica-
tions, where the cost of installation and regular maintenance is extremely high. For
example, an offshore wind farm requires about 20% higher operating and mainte-
nance cost compared with an onshore farm. Moreover, environmental concerns are
still unsolved and can never be compromized.
7.1.4Medium-Voltage Converters for Direct

Grid Integration
With the traditional low-voltage converters, the conventional renewable generation

systems possess the power frequency (i.e., 50 or 60Hz) step-up transformers, fil-
ters, and boosters, resulting in not only large size, weight, and loss but also high
cost and complexity of the system, both capital and operational [16]. Today, the
industrial trend is to move away from these heavy and large-size passive compo-
nents to power electronic systems that use more and more semiconductor elements
controlled by a digital circuit. In such a way, smart operation can be ensured. In
comparison with the conventional two-level converters, multilevel converters pre-
sent lower switching losses, lower voltage stress on switching devices, and better
harmonic performance. These remarkable features enable the connection of renew-
able energy systems directly to the grid without using large, heavy, and costly
power transformers and also minimizing the input and output filter requirements.
Because of some special features (e.g., the number of components scales linearly
modular in constriction, hence enabling high-level attainability), the MMC con-
verter topology can be considered as the best possible candidate for medium-volt-
age applications [17, 18]. The high number of levels means that medium-voltage
attainability is possible to connect the renewable generation units to the medium-
voltage grid directly and it is also possible to improve the output power quality.
The component number and control complexity increase linearly with the increase
7.1Introduction 241
in number of levels. Figure7.20 plots the relationship between component number

and complexity of different multilevel converters.
On the other hand, the distortion in generated output voltage and semiconduc-
tor cost of the converter decrease dramatically with the increase in the converter
number of levels [9, 10, 19]. Figure7.21 plots the THD (%) and semiconductor
cost of different number of MMC converter levels for an 11-kV system.
Due to the unavailability of IGBT ratings, the 13- and 17-level converters used
the IGBTs that are used in the 11- and 15-level converters, respectively. Hence,
the semiconductor cost curve is up and down in nature. Moreover, lower switch-
ing frequency, even fundamental switching frequency, can be used with the high-
level number converters, which can significantly reduce the switching losses of the
converter [9]. Therefore, the optimal selection in the number of converter levels
is important for the best performance/cost ratio of the medium-voltage converter
systems and this is the central content of this chapter. In this chapter, two systems
of 11 and 33kV, respectively, are designed and analyzed taking into account the
specified system performance, control complexity, and cost and market availability
of the power semiconductors. It is found that the 19-level and 43-level converters
are the optimal choice for the 11- and 33-kV systems, respectively. The design and
analysis of the 11-kV MMC converter and 33-kV MMC converter systems are pre-
sented in detail in the following sections.
Fig.7.20Number of IGBTs/
Number of IGBTs/Complexity
120
control complexity versus Number of IGBTs
converter level numbers 100 Control complexity
80
60
40
20
5 10 15 20
Number of levels
Fig.7.21THD (%) and 12

semiconductor cost versus
THD (%)/Cost (10000)
Semiconductor cost
converter level numbers 10 (AUD)
THD (%)
8
8 10 12 14 16 18 20
Number of levels
7.2Design and Analysis of 11-kV Converter Systems
The total nominal DC-link voltage for an 11-kV 2-level converter is 16,176V.
Each H-bridge inverter cell commutation voltage of a 5-level converter is about
4,044V. The highest voltage rating of a commercially available IGBT is 6.5kV,
which is recommended for a maximum voltage of 3,600V. Therefore, the 5-level
or lower-level converter cannot be used to design the 11-kV converter. Each
H-bridge inverter cell commutation voltage of a 7-level topology-based 11-kV
converter is 2,696V which may be supported by the 6.5-kV IGBT. Accordingly,
at least 7-level topology is required to design an 11-kV converter. The output
power quality of a 21-level converter is good enough to feed into the 11-kV AC
grid directly. The cheap 1.7-kV IGBT can be used to design the 21-level converter.
Therefore, 7-level to 21-level MMC topologies are considered for an 11-kV con-
verter system. If Vcom is the commutation voltage of respective commutation cells,
the device voltage utilization factor (DVUF) can be calculated from
Vcom
DVUF = . (7.1)
Vcom@100FIT
Table7.2 summarizes the commutation voltage and DVUF of different level MMC
converters. A higher DVUF is essential for cost-effective design, since the semicon-
ductor cost is a significant figure in medium-voltage converter applications. From
Table7.2, it can be seen that only a few converters have high DVUFs. Considering
the availability of the power semiconductor devices, one can find only a few converter
topologies of 9, 11, 15, 19, and 21 levels may give good DVUFs. Although the 7-level
circuit topology can be used to develop an 11-kV converter, the available commercial
IGBTs are not well suited. Therefore, devices of higher ratings are required, which
would yield a very poor DVUF of about 75%. A low DVUF means the use of unnec-
essarily high-cost semiconductors. In general, to use the active switching devices cost-
effectively, a converter must have a DVUF of 90% or above. As listed in the table,
only two converter systems give more than 90% DVUF. In order to ensure a cost-
effective design, the converters of 9, 11, 15, 19, and 21 levels for 11-kV systems were
considered for further analysis. The lowest number of levels of an MMC converter is
Table7.2DVUF with different level number; 11-kV system

No. of levels Vcom (V) Rated device voltage (kV) Vcom@100FIT (V) DVUF
(%)
7 2,696 6.50 3,600 75
9 2,022 4.50 2,250 90
11 1,618 3.30 1,800 90
13 1,348 3.30 1,800 75
15 1,156 2.50 1,200 96
17 1,011 2.50 1,200 84
19 898 1.70 900 99
21 809 1.70 900 90
7.2 Design and Analysis of 11-kV Converter Systems 243
3, and such a converter consists of only one H-bridge inverter cell on each phase leg.
Each additional cascaded H-bridge inverter cell contributes two voltage levels to the
output voltage waveform. Therefore, the numbers of levels are only of odd values.
7.2.1Design and Analysis of 9-Level 11-kV Converter
The circuit diagram of a 9-level MMC converter is shown in Fig.7.22. There are
48 active switching devices in the three-phase circuit, and each phase contains
16 devices. Each phase leg consists of 4 H-bridge inverter cells in cascaded con-
nection. The line voltage of a 9-level MMC converter is illustrated in Fig.7.23.
The line voltage THDs of an 11-kV 9-level converter are calculated and found as
H-bridge inverter cell H-bridge inverter cell H-bridge inverter cell
a
b
c

SB6
Vdc SA10 SA12 Vdc SB10 SB12 Vdc SC12

SC10
Vdc Vdc Vdc

Fig.7.22Three-phase 9-level MMC converter

9
8
7
10 Vab 2022 V 6
5
4
Voltage (kV) 3
2
1
0
-10
10 15 20 25 30 35 40
Time (ms)
Fig.7.23Line voltage of 9-level MMC converter
0.05
0.04
9.60% THDs
0.03
0.02
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
Fig.7.24Line voltage frequency spectrum of 9-level MMC converter
10 Vab Vbc Vca

Voltage (kV)
-10
10 15 20 25 30 35 40
Time (ms)
Fig.7.25Three line voltages of 9-level MMC converter
9.60%. The frequency spectrum of line voltage is depicted in Figs.7.24 and 7.25
shows the three line voltages of three-phase converter system.
Table7.3Commercially IGBT modules Rated current (A) Brand

available 4.5-kV-rated IGBTs
CM400HB-90H 400 Mitsubishi Electric
CM600HB-90H 600 Mitsubishi Electric
5SNA 0650J450300 650 ABB
5SNA 1200G450350 1,200 ABB
FZ800R45KE3_B5 800 Infineon
A total of 44 arithmetic and logic operations (ALOs) are required in the

MATLAB/Simulink environment to control the 3-phase 9-level MMC converter
with the phase-shifted carrier-based modulation scheme. The switching controller
should have at least 22 dedicated PWM channels. Different vendors, such as ABB,
Infineon, and Mitsubishi Electric, commercially developed 4.5-kV-rated IGBTs.
Table 7.3 tabulates the possible 4.5-kV IGBT modules for design of the 9-level
11-kV converter. Figure7.26 shows a photograph of Mitsubishi Electric 4.5-kV
IGBT module CM400HB-90H, whose current rating is 400 A. A 4.5-kV-rated 48
IGBT costs about AUD 86,400.00.
The conduction losses of a 9-level MMC converter can be approximated by
Pc_9 = 8 6(Pc_sw + Pc_D ). (7.2)
Compared with a 2-level converter, the loss from (7.2) appears to be 8 times
higher, but the total conduction loss of a 9-level converter is similar to that of
the 2-level converter because the device commutation voltage is 8 times lower
than that of the switching devices in the 2-level converter. The conduction loss
highly depends on the collectoremitter saturation voltage, VCE(sat), and the diode
Fig.7.26Photograph of Mitsubishi Electric 4.5-kV 400-A IGBT CM400HB-90H

saturation voltage (V)

Collector-emitter
4
1
0 1 2 3 4 5 6
Rated voltage (kV)
Fig.7.27Collectoremitter saturation voltages of 600-A-rated IGBTs [10]
forward voltage, which in turn depend on the voltage rating of switching devices.
Figure 7.27 plots the collectoremitter saturation voltages of different voltage
rated Mitsubishi Electric IGBTs [20].
The DC supply voltage rating of an 11-kV 11-level MMC converter is 1,618V.

The available 3.3-kV IGBT may be used to design the 11-kV 11-level converter,
because this IGBT is recommended for 1,800V maximum applications. With
these IGBTs, 90% DVUF can be ensured, which is important to conform to the
high performance/cost of the power converter. Figure7.28 plots the line voltage
waveform of an 11-level 11-kV MMC converter. There are 20 steps in the peak-
to-peak line voltage waveform, and each step contributes 1,618V to the line volt-
age of 32,360V peak to peak. The frequency spectrum of the line voltage of an
11
10 Vab 1618 V
Voltage (kV)
1
0
-10
10 15 20 25 30 35 40
Time (ms)
Fig.7.28Line voltages of 11-level MMC converter

0.05
0.04
8.20% THDs
0.03
0.02
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
10 Vab Vbc Vca

Voltage (kV)
-10
10 15 20 25 30 35 40
Time (ms)
11-level MMC converter is depicted in Fig.7.29. Figure7.30 shows the three line
voltages of three-phase converter system.
In total, 60 IGBTs are in the 3-phase 11-level MMC converter circuit. Each
IGBT price is about AUD 1,369.31, such that the cost of 60 IGBTs is about
AUD 82,159.00. A few vendors, such as Mitsubishi Electric, ABB, and Infineon,
commercially developed 3.3-kV IGBT modules. Table7.4 summarizes the pos-
sible 3.3-kV IGBT modules to design the 11-level 11-kV converter. Figure7.31
shows the photograph of Infineon 3.3-kV IGBT module FZ800R33KL2C. Since
the switching control circuits consist of 55 ALOs, at least 30 PWM channels are
required to control the 3-phase 11-level converter. The single DSP cannot handle
the PWM pulses as required by the 3-phase 11-level converter. The parallel oper-
ation of several DSPs makes the control circuit hard to implement. The modern
FPGA may be an appropriate option to design the control circuit for multilevel
converters.

CM400HG-90H 400 Mitsubishi
Electric
5SNA 1000N330300 1,000 ABB
FZ400R33KL2C_B5 400 Infineon
FZ800R33KL2C 800 Infineon
FZ800R33KF2C 800 Infineon
Fig.7.31Infineon
3.3-kV IGBT module
FZ800R33KL2C
The 15-level converter topology is also well suited with the available commer-
cial semiconductors. Each phase leg consists of seven H-bridge inverter cells,
28 IGBTs in a phase, and 84 in total in the 3-phase circuit. The DC supply volt-
age rating of an 11-kV 15-level converter is 1,155V. The commercially avail-
able 2.5-kV IGBT may be used to design the 11-kV 15-level converter, because
this IGBT is recommended for 1,200V maximum applications. Figure7.32 plots
15
Vab 1155 V
10
Voltage (kV)
1
0
-10
10 15 20 25 30 35 40
Time (ms)

0.05
0.04
6.00% THDs
0.03
0.02
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
10 Vab Vbc Vca

Voltage (kV)
-10
10 15 20 25 30 35 40
Time (ms)
the line voltage waveform of a 15-level MMC converter. There are 28 steps in
the peak-to-peak line voltage waveform, and each step contributes 1,155V to
the line voltage. The frequency spectrum of the line voltage of a 15-level MMC
converter is depicted in Fig.7.33. Figure7.34 shows the three line voltages of
three-phase 15-level 11-kV converter. Quite a few vendors, such as ABB and
Mitsubishi Electric, commercially developed 2.5-kV IGBT modules. Table7.5

5SNA 1500E250300 1,500 ABB
CM400DY-50H 400 Mitsubishi
Electric
5SNA 1000E250100 1,200 ABB
CM800HA-50H 800 Mitsubishi
Electric
CM1200HA-50H 1,200 Mitsubishi
Electric
Fig.7.35ABB 2.5-kV
IGBT module 5SNA
1200E250100
summarizes possible 2.5-kV IGBT modules for designing the 15-level 11-kV con-
verter. Figure7.35 shows the photograph of ABB 2.5-kV IGBT module 5SNA
1200E250100, whose rated current is 1,200 A.
Figure7.36 shows the circuit diagram of a 19-level converter. Each phase leg of a
19-level converter consists of 9 H-bridge inverter cells, i.e., 36 IGBTs in a phase
and 108 in total in the 3-phase circuit. The DC-link voltage rating of an 11-kV
19-level converter is about 899V. Figure7.37 shows voltage waveform across an
IGBT in 19-level 11-kV converter.
The available 1.7-kV IGBT may be used to design the 11-kV 19-level con-
verter. This IGBT is cheap and matured in terms of technology and is recom-
mended for 900V maximum applications. With this converter topology, the
highest achievable DVUF is almost 100%. Although a large number of IGBTs
are required for a 19-level converter, the total semiconductor cost is about AUD
36,670.00, which is only about 42% of a 9-level converter with the same power
rating. Each H-bridge inverter cell in cascaded multilevel converter generates
output voltage as 2-level converter. Figure7.38 shows the voltage waveform
of an H-bridge inverter cell in a 19-level 11-kV converter. There are 18 and 36
steps in the peak-to-peak phase and line voltage waveforms, and each step con-
tributes 899V to the phase or line voltage. Figure7.39 shows the phase volt-
age waveform of a 19-level 11-kV converter. Figure7.40 plots the line voltage
waveform of a 19-level MMC converter. Due to the small step size, the output
power quality is suitable for filter-less grid connection. The frequency spec-
trum of the line voltage of a 19-level MMC converter is depicted in Fig.7.41.
Figure 7.42 shows the three line voltages of the three-phase 19-level 11-kV
a
b
c

SB6
Vdc Vdc Vdc

Fig.7.36Three-phase 19-level MMC converter [1]

900
800
700
600
Voltage (V)
500
400
300
200
100
0
0 5 10 15 20 25 30 35 40
Time (ms)
Fig.7.37Voltage waveform across an IGBT in 19-level MMC converter

800
600
Voltage (V) 400
200
0
-200
-400
-600
-800
15 20 25 30 35 40 45 50
Time (ms)
Fig.7.38Voltage waveform of an H-bridge inverter cell in 19-level MMC converter
8
6
4
Voltage (kV)
2
0
-2
-4
-6
-8
10 15 20 25 30 35 40 45 50
Time (ms)
Fig.7.39Phase voltage of 19-level MMC converter
19
Vab 899 V
10
Voltage (kV)
1
0
-10
10 15 20 25 30 35 40
Time (ms)

0.05
0.04
4.30% THDs
0.03
0.02
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
10 Vab Vbc Vca

Voltage (kV)
-10
10 15 20 25 30 35 40
Time (ms)
Fig.7.42Three line voltages of three-phase 19-level MMC converter
300
200
Current (A)
100
-100
-200
-300
0 10 20 30 40 50 60
Time (ms)
Fig.7.43Current waveform in an IGBT of 19-level MMC converter


available 1.7-kV-rated IGBT
CM500HA-34A 500 Mitsubishi
modules
Electric
SKM 600GA176D 400 Semikron
SKM 800GA176D 600 Semikron
5SNE 0800M170100 800 ABB
FZ400R17KE4 400 Infineon
FZ1200R17HE4 1,200 Infineon
CM600DY-34H 600 Mitsubishi
Electric
Fig.7.44Semikron 1.7 IGBT module SKM 600GA176D
converter. The current waveform of an IGBT in the 19-level 11-kV converter is

shown in Fig.7.43. There are a few vendors, such as Semikron, ABB, Mitsubishi
Electric, and Infineon, who commercially developed 1.7-kV-rated IGBT mod-
ules. Table7.6 summarizes the possible 1.7-kV IGBT modules for designing the
19-level 11-kV converter. Figure7.44 shows the photograph of Semikron 1.7
IGBT module SKM 600GA176D, whose current rating is 400 A.
The DC-link voltage rating of an 11-kV 21-level converter is about 809V. The
1.2-kV IGBT or lower-rated devices cannot be used and at least 1.7-kV IGBT
is required to design the 11-kV 21-level converter. With this converter topology,
21
Vab 809 V
10
Voltage (kV)
1
0
-10
10 15 20 25 30 35 40
Time (ms)
the highest achievable DVUF is 90%, which is much lower than that of the
19-level converter. In total, 120 IGBTs are used in a 21-level converter, where 10
H-bridge inverter cells are cascaded on each phase leg. The total semiconductor
cost is about AUD 40,744.00, which requires about 11% more cost than that of
the 19-level converter. Figure7.45 shows the line voltage waveform of a 21-level
converter, and its frequency spectrum is illustrated in Fig.7.46. The THDs are cal-
culated and found to be 4.25%, which means only 1.1% improvement compared
with the 19-level converter. The control scheme consists of 110 ALOs to drive 120
active switching devices. The control complexity of an 11-kV 21-level converter is
about 11% more than that of a 19-level converter. Figure7.47 shows the three line
voltages of three-phase 21-level 11-kV converter.
0.05
0.04
4.25% THDs
0.03
0.02
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)

10
Vab Vbc Vca
Voltage (kV)
-10
10 15 20 25 30 35 40
Time (ms)
Fig.7.47Three line voltages of 21-level MMC converter
7.2.6Selection of Number of Levels for 11-kV

Converter Systems
The number of ALOs for the switching section and cost of semiconductors
are calculated as tabulated in Table7.7. The number of ALOs is used to com-
pare the complexity of the converters. The THDs are calculated through the
MATLAB/Simulink environment. The price data quoted for the semiconductor
devices were collected from the Galco Industrial Electronics and Farnell catalogs,
and quotations from different vendors where devices were chosen from the same
family so that it was possible to meet the requirements.
Using (3-29), the normalized index values are calculated to obtain the overall
performance/cost of the converters. Table7.8 tabulates the normalized index val-
ues of Table7.7. Based on Table7.8, the performance indicators are plotted as
shown in Fig.7.48. For the 11-kV converter, the total index value is the lowest
Table7.7Converter No. of levels 9 11 15 19 21

comparison for an 11-kV
IGBTs 48 60 84 108 120
system [9, 10]
THDs (%) 9.60 8.20 6.00 4.30 4.25
Cost (AUD) 86,400 82,159 47,066 36,670 40,744
ALOs 44 55 77 99 110
Table7.8Performance No. of levels 9 11 15 19 21

indexes of 11-kV converters
THD 1.00 0.74 0.33 0.01 0.00
[9, 10]
IGBTs cost 1.00 0.92 0.21 0.00 0.09
Complexity 0.00 0.17 0.50 0.83 1.00
Total index 2.00 1.83 1.04 0.84 1.09
1.5
THD Cost Complexity
1
Index value
0.5
0
10 12 14 16 18 19 20
Number of levels
Fig.7.48Graphical representation converter performance parameters [9, 10]
at level 19, because there is no significant output power quality improvement and
semiconductor cost reduction for inverters with more than 19 levels. Moreover,
the component number and control complexity increase linearly with the increase
in the number of levels. Therefore, the 19-level topology is the optimum for the
11-kV converter systems. The output power quality of a 19-level converter is good
enough to feed the converter output directly into the medium-voltage grid directly
(i.e., without using step-up transformer or line filter circuit). Figure7.49 shows the
total indexes of different 11-kV converters. Lower total index value means lower
converter cost and complexity and better quality of output power.
2.00
1.83
Total index
9-level 11 kV converter
1.09
1.04
0.84
Fig.7.49Total performance indexes of different 11-kV MMC converters

7.3Design and Analysis of 33-kV Converter Systems
The total nominal DC-link voltage for a 33-kV 2-level converter is 48,538V. Each
H-bridge inverter cell communication voltage of a 13-level converter is about
4,044V. The highest voltage rating of a commercially available IGBT is 6.5kV,
which is recommended for a maximum voltage of 3,600V. Therefore, the con-
verter should have 13 or more levels for direct connection to a 33-kV grid. Each
H-bridge inverter cell communication voltage of a 15-level topology-based 33-kV
converter is 3,467V which may be supported by the 6.5-kV IGBT. Accordingly, at
least 15-level topology is required to design a 33-kV converter. While the output
power quality of 55-level converter is good enough to feed into the 33-kV AC grid
directly, the cheap 1.7-kV IGBT can be used to construct the 55-level converter.
Therefore, MMC topologies 15 level to 55 level are considered for a 33-kV con-
verter system. The DVUFs are tabulated for converters of different level numbers
in Table7.9.
A high DVUF is essential for a cost-effective design, since the semiconductor
cost is the significant figure in MV inverter applications. From Table7.9, it can
Table7.9DVUFs of 33-kV MMC converters of different levels

No. of levels Vcom (V) Rated device voltage (V) Vcom@100FIT (V) DVUF
(%)
15 3,467 6,500 3,600 96
17 3,034 6,500 3,600 85
19 2,697 6,500 3,600 75
21 2,427 6,500 3,600 68
23 2,206 4,500 2,250 98
25 2,022 4,500 2,250 90
27 1,867 4,500 2,250 83
29 1,734 3,300 1,800 96
31 1,618 3,300 1,800 90
33 1,517 3,300 1,800 84
35 1,428 3,300 1,800 79
37 1,348 3,300 1,800 75
39 1,277 3,300 1,800 71
41 1,214 3,300 1,800 65
43 1,156 2,500 1,200 96
45 1,103 2,500 1,200 92
47 1,055 2,500 1,200 88
49 1,011 2,500 1,200 84
51 971 2,500 1,200 81
53 934 2,500 1,200 78
55 899 1,700 900 100
be seen that only some converters have high DVUFs. In order to ensure a cost-
effective design, the converters with level numbers of 15, 23, 29, 43, and 55 for a
33-kV system were considered for further analysis in the following sections. The
converters with DVUFs of more than 95% are assumed to satisfactorily utilize the
active switching devices.
The 15-level converter topology may be considered for designing a 33-kV con-
verter with the available commercial semiconductors. Each phase leg consists
of seven H-bridge inverter cells with 28 IGBTs in a phase and 84 in total in the
3-phase circuit. The DC supply voltage for a 33-kV 15-level converter is 3,467V.
The available 6.5-kV IGBT may be used for constructing the 33-kV 15-level con-
verter, because this IGBT is recommended for application of 3,600V in maxi-
mum. Figure7.50 plots the phase voltage waveform of a 33-kV 15-level MMC
converter. There are 28 steps in the peak-to-peak line voltage waveform, and each
step contributes 3,467V to the line voltage. Figure7.51 shows the line voltage
waveforms of the 15-level 33-kV converter. The frequency spectrum of the line
voltage of a 15-level MMC converter is depicted in Fig.7.52.
ABB, Infineon, and Mitsubishi Electric commercially developed 6.5-kV IGBT
modules with various power ratings. Table7.10 summarizes some possible 6.5-kV
IGBT modules for constructing medium-voltage converters. ABB and Mitsubishi
Electric are the market leaders to supply 6.5-kV IGBT module, who developed
6.5-kV IGBT modules with a current rating of a few hundred amperes. Figure7.53
shows the ABB 6.5-kV IGBT module 5SNA 0400J650100.
15
Phase 14
20 13
voltage 12
11
10 10
Voltage (kV)
3467 V 9
0
8
7
6
-10 5
4
3
-20 2
1
10 15 20 25 30 35 40 45 50
Time (ms)
Fig.7.50Phase voltage of 33-kV 15-level MMC converter

50
Vab Vbc Vca
Voltage (kV)
0
-50
10 15 20 25 30 35 40
Time (ms)
Fig.7.51Three line voltages of three-phase 33-kV 15-level MMC converter
0.03
0.02
6.40% THDs
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
Fig.7.52Line voltage frequency spectrum of 33-kV 15-level MMC converter

Electric
Electric
5SNA 0400J650100 400 ABB
5SNA 0750G650300 750 ABB
Fig.7.53Photograph of ABB 6.5-kV IGBT module 5SNA 0400J650100
The power circuit of a 23-level 3-phase MMC converter is shown in Fig.7.54.

The H-bridge inverter DC-link voltage of a 33-kV 23-level converter is 2,206V.
The available 4.5-kV IGBT may be used to construct the 33-kV 23-level converter
because this IGBT is recommended for application of 2,250V in maximum. An
about 98% DVUF can be obtained with the 4.5-kV IGBTs. In total, 11 H-bridge
inverter cells are in each phase leg. In order to drive such a converter, 11 phase-
shifted carriers are required and adjacent carriers are shifted by (180/11) degrees.
Figure 7.55 shows the phase-shifted carrier-based switching control scheme for
the 23-level MMC converter. The scheme involves about 121 ALOs. The phase
voltage waveform of the 23-level 33-kV converter is shown in Fig.7.56. The line
voltage waveforms of the 23-level 33-kV converter are depicted in Fig.7.57. The
line peak-to-peak voltage consists of 44 voltage levels, and each level contributes
2,206V to the peak-to-peak line voltage. The line voltage contains 4.54% THDs,
which moderately satisfy the grid requirements. The frequency spectrum of the
23-level 33-kV converter is illustrated in Fig.7.58.
The cost of 132 IGBTs is about AUD 237,600.00, which is about 8.8% lower
than that of a similar 15-level converter. Compared with a 15-level 33-kV con-
verter, about 29% improvement of THDs is obtained.
a
b
c

SB6
Cell A11 Cell 11 Cell C11
Vdc Vdc Vdc

Fig.7.54Circuit diagram of a 33-kV 23-level MMC converter
The H-bridge inverter DC-link voltage for a 33-kV 29-level converter is 1,734V.
The available 3.3-kV IGBT may be used to construct the 33-kV 29-level converter,
because this IGBT is recommended for 1,800V maximum applications. About
96% DVUF can be obtained with the 3.3-kV IGBTs. The devices dv/dt is about
50% lower than that of the devices in the 15-level converter, under the same oper-
ating conditions. About 50% lower dv/dt means significant improvement in the
device reliability. In total, there are 14 H-bridge inverter cells in each phase leg. In
order to drive such a converter, 14 phase-shifted carriers are required and adjacent
carriers are shifted by (180/14) degrees. The phase voltage and line voltage wave-
forms of the 29-level 33-kV converter are depicted in Figs.7.59 and 7.60.
Gate pulses
Reference >= 0 PS1
Comp NOT PS2
Inv
0o shifted >= 0 PS3
Inv
Comp NOT PS4
o
180 shifting Inv
>= 0 PS5
o
(180/11)1 Comp NOT PS6
shifted
Inv
Inv
o
>= 0 PS7
180 shifting
Comp NOT PS8
Inv
>= 0 PS41
(180/11)10o Comp NOT PS42
shifted
Inv
Inv
o
>= 0 PS43
180 shifting
Comp NOT PS44
Inv
Fig.7.55Phase-shifted carrier-based switching scheme for 23-level MMC converter
Phase
20
voltage
10
Voltage (kV)
2206 V
0
-10
-20
10 15 20 25 30 35 40 45 50
Time (ms)

50
Vab Vbc Vca
Voltage (kV)
0
-50
10 15 20 25 30 35 40
Time (ms)
0.03
0.02 4.54% THDs
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
Phase
20 voltage
10
Voltage (kV)
1734 V
0
-10
-20
10 15 20 25 30 35 40 45 50
Time (ms)

50
Vab Vbc Vca
Voltage (kV)
-50
10 15 20 25 30 35 40
Time (ms)
0.03
0.02 4.12% THDs
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
The line peak-to-peak voltage consists of 56 voltage levels, and each level con-
tributes 1,734V to the peak-to-peak line voltage. The line voltage contains 4.12%
THDs, which satisfies the grid requirements and may enable line filter-less grid con-
nection. Figure7.61 plots the frequency spectrum of the 29-level 33-kV converter.
Compared with the 23-level converter, about 9% additional harmonic reduction can
be achieved with the 29-level converter. The switching scheme involves about 154
ALOs, which is 27% more complex than that of a 23-level converter control scheme.
The cost of 168 IGBTs is about AUD 229,992.00, which is about 12% lower than
that of a similar 15-level converter using IGBTs of one level higher voltage rating.
The H-bridge inverter DC-link voltage for a 33-kV 43-level converter is 1,156V.
The available cheap and matured 2.5-kV IGBT can be used for constructing the
33-kV 43-level converter, because this IGBT is recommended for application of
1,200V in maximum. About 96% DVUF can be obtained with the 2.5-kV IGBTs.
In total, there are 21 H-bridge inverter cells in each phase leg and 252 active
switching devices are required for the 3-phase 43-level converter. Figure7.62
shows the circuit diagram of 43-level converter. In order to drive 252 IGBTs, 21
phase-shifted carriers are required and adjacent carriers are shifted by (180/21)
degrees. Figure7.63 shows the block diagram of the switching control scheme for
the 3-phase 43-level converter.
A total of 231 ALOs are involved with the switching scheme. Although three
times more switching devices are used in the 43-level converter, the semiconduc-
tor cost is about 41% lower than that of the 15-level converter using IGBTs of one
level higher voltage rating, due to the cheap cost of low-voltage-rated devices. The
line peak-to-peak voltage consists of 84 voltage levels, and each level contributes
a
b
c

SB6
Cell A21 Cell 21 Cell C21
Vdc Vdc Vdc

Fig.7.62Circuit diagram of a 43-level MMC converter

Gate pulses
Reference >= 0 PS1
Comp NOT PS2
Inv
0o shifted >= 0 PS3
Inv
Comp NOT PS4
o
180 shifting Inv
>= 0 PS5
o
(180/21)1 Comp NOT PS6
shifted
Inv
Inv
o
>= 0 PS7
180 shifting
Comp NOT PS8
Inv
>= 0 PS81
(180/21)20o Comp NOT PS82
shifted
Inv
Inv
o
>= 0 PS83
180 shifting
Comp NOT PS84
Inv
Fig.7.63Switching control scheme of 43-level MMC converter
1,156V to the peak-to-peak line voltage. Figure7.64 shows voltage waveform

across an IGBT in a 43-level 33-kV converter. Each H-bridge inverter cell pro-
duces two voltage levels as in a 2-level converter. Figure7.65 shows the output
voltage waveform of an H-bridge inverter cell in the 43-level 33-kV converter.
Figure7.66 shows the phase voltage of 43-level 33-kV converter. Due to the small
step size, the line voltage waveforms are found to be very consistent with the ref-
erence sine waveforms. Figure7.67 plots the line voltage waveform. The output
power quality of a 43-level converter is good enough to feed directly into the grid.
The line voltage of THDs is about 3.61%, which strongly satisfies the 5% limit
by IEEE1547 and IEC61727 standards. The frequency spectrum of line voltage
is depicted in Fig.7.68. Compared with the 15-level converter, the 43-level con-
verter provides 44% better quality output power. As the number of active switch-
ing devices has increased three times, the complexity of the control scheme of a
43-level converter has also increased three times more than that of a 15-level con-
verter. Current through an IGBT is shown in Fig.7.69.
1200
1000
800
Voltage (V)
600
400
200
0
10 15 20 25 30 35 40 45 50
Time (ms)
Fig.7.64Voltage across an IGBT in 43-level 33-kV MMC converter
1000
500
Voltage (V)
-500
-1000
15 20 25 30 35 40 45 50
Time (ms)
Fig.7.65Output voltage of an H-bridge inverter cell in 43-level 33-kV MMC converter
Phase
20 voltage
10
Voltage (kV)
1156 V
0
-10
-20
10 15 20 25 30 35 40 45 50
Time (ms)

50
Vab Vbc Vca
Voltage (kV)
0
-50
10 15 20 25 30 35 40
Time (ms)
0.03
0.02
3.61% THDs
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
400
300
200
Current (A)
100
0
-100
-200
-300
-400
0 10 20 30 40 50 60
Time (ms)
Fig.7.69IGBT current of 43-level MMC converter

Phase
20 voltage
899 V
10
Voltage (kV)
-10
-20
10 15 20 25 30 35 40 45 50
Time (ms)
The H-bridge inverter DC-link voltage for a 33-kV 55-level converter is 899V.
The available cheap and mature 1.7-kV IGBT can be used to design the 33-kV
55-level converter, because this IGBT is recommended for application of 900V
in maximum. An almost 100% DVUF can be obtained with the 1.7-kV IGBTs. In
total, there are 27 H-bridge inverter cells in each phase leg and 324 active switch-
ing devices are required for the 3-phase 55-level converter. The 1.7-kV IGBT is
cheap, and the total semiconductor cost of the 55-level converter is about AUD
110,030.00, which is 22% lower than that of the 43-level converter using IGBTs
of one level higher voltage rating. The IGBTs price of 2.5kV or lower is in linear
50
Vab Vbc Vca
Voltage (kV)
-50
10 15 20 25 30 35 40
Time (ms)

range. Therefore, the reduction to the total price is small. The line peak-to-peak
voltage consists of 108 voltage levels, and each level contributes 899V to the
peak-to-peak line voltage. Due to the small step size, the line voltage waveforms
are found to be very consistent with the reference sine waveforms. Figures7.70
and 7.71 plot the phase and line voltage waveforms, respectively. The line volt-
age frequency spectrum is shown in Fig.7.72. The 55-level 33-kV voltage wave-
form contains about 3.47% THDs, which is only 3.8% lower than that of the
43-level converter. However, the switching scheme of the 55-level converter con-
sists of 297 ALOs. Compared with the 43-level converter, the control scheme of
the 55-level converter requires 66 more ALOs; that is, the control scheme is 28%
more complex.
0.03
0.02
3.47% THDs
0.01
0
0 0.05 0.1 0.15 0.2 0.25
Frequency (MHz)
1.5
THD Cost Complexity
1
Index value
0.5
0
20 30 40 43 50
Number of levels
Fig.7.73Graphical representation converter parameters [10, 21]

7.3.6Selection of Optimal Number of Levels for 33-kV

Converter
The H-bridge inverter cell communication voltage of a 13-level converter is

about 4,044V. The highest voltage rating of a commercially available IGBT is
6.5kV, which is recommended for application of a maximum voltage of 3,600V.
Therefore, a 13-level or lower-level converter cannot be used to construct the
33-kV converter. Each H-bridge inverter cell communication voltage of a 15-level
topology-based 33-kV converter is 3,467V which may be supported by the 6.5-kV
IGBT. Owing to this, at least 15-level topology is required to design a 33-kV con-
verter. The output power quality of a 55-level inverter is good enough to directly
feed into the 33-kV AC grid. The cheap 1.7-kV IGBT can be used to design the
55-level inverter. There is no significant performance improvement or cost reduc-
tion with converters of more than 55 levels. Moreover, the control complexity
increases with the number of converter levels. For this reason, 15-level to 55-level
MMC converter topologies are considered for a 33-kV inverter system.
The ALOs for switching section, THDs of output power, and cost of semicon-
ductors are calculated as tabulated in Table7.11. The number of ALOs is used
to compare the complexity of the converters. The THDs are calculated in the
MATLAB/Simulink environment. Using (3-29), the normalized index values are
calculated. Table7.12 summarizes the normalized index values of Table7.11.
Based on Table7.12, the performance graphs are plotted in Fig.7.73. For the
33-kV converter, the total index value is the lowest at the 43 level, because there
is no significant output power quality improvement and semiconductor cost reduc-
tion for converters with more than 43 levels. In addition, the component number
and control complexity increase linearly with the increase in the number of levels.
Therefore, 43-level topology is considered as the optimal for 33-kV converter sys-
tems. Figure7.74 shows total indexes of different 33-kV MMC converters.
Table7.11Converter No. of 15 23 29 43 55
comparison for a 33-kV levels
system [10, 21]
IGBTs 84 132 168 252 324
THD (%) 6.40 4.54 4.12 3.61 3.47
Cost (AU$) 258,552 237,600 229,992 141,200 110,030
ALOs 77 121 154 231 297
Table7.12Total indexes No. of levels 15 23 29 43 55

of different 33-kV MMC
Performance 1.00 0.36 0.22 0.04 0.00
converters [10, 21]
IGBTs cost 1.00 0.85 0.80 0.20 0.00
Complexity 0.00 0.20 0.35 0.70 1.00
Total index 2.00 1.41 1.37 0.94 1.00
7.4Summary 273
2.00
Total index
15-level 33 kV converter 1.41 1.37
1.00
0.94
Fig.7.74Total indexes of different 33-kV converters
7.4Summary
In this chapter, the designs of 11-kV and 33-kV converter systems have been ana-
lyzed with a focus on selecting the optimal number of levels for a medium-volt-
age converter. All possible MMC converter topologies have been considered for
each converter system. In order to ensure a cost-effective design, the DVUFs were
calculated and only the selected converters with high DVUFs were considered.
During the design process, the availability of semiconductor devices was consid-
ered in the first instance. After checking the availability of devices, the converter
systems were designed taking into account three main factors: the specified con-
verter output power quality, complexity of the switching controller, and cost of the
semiconductors. The investigation has shown that the 19-level MMC converter is
the optimal choice for an 11-kV converter system and the 43-level topology is the
optimal choice for a 33-kV inverter system.
References
2. Islam MR, Guo YG, Zhu JG (2013) Power converters for wind turbines: current and future
development. In: Mendez-Vilas A (eds) Materials and processes for energy: communicating
current research and technological developments. Energy Book Series-2013 edn. Spain, pp
559571
3. Solutions-SEMISTACK Renewable Energy (2011) Three phase inverter. Catalogue
2011/2012, Semikron International, May 2011, pp 9495
4. Solutions-SEMIKUBE (2011) Three phase rectifier and inverter. Catalogue 2011/2012,
Semikron International, May 2011, pp 8990
5. Mitsubishi Electric, Mitsubishi IGBT Modules CM100MXA-24S, [Online]. Available at: htt
p://www.ineltron.de/english/mitsubishi-data/IGBT/CM100MXA-24S.pdf
6. Solutions-SKiiPRACK (2011) Four-quadrant converter. Catalogue 2011/2012, Semikron
International, May 2011, pp 8889
7. See http://www.abb.com/ for information about ABB central inverter
8. See http://www.siemens.com/ for information about SINVERT PVS
Electron 29(8):41674182
11. Islam MR, Guo YG, Zhu JG (2012) A Transformer-less compact and light wind turbine gen-
erating system for offshore wind farms. In: Proceedings of 2012 IEEE international con-
ference on power and energy (PECon2012), 25 Dec 2012, Kota Kinabalu, Malaysia, pp
605610
12. ABB (2013) Distribution transformer. Available online: http://www.coronabd.com, [access
18.03.2013]
13. ABB (2013) Liquid-filled transformer. Available online: http://www05.abb.com, [access
18.03.2013]
14. Pauwels SLIM Transformer (2013) SLIM transformer inside the worlds highest wind tur-
bine. Available online: http://www.pauwels.com, [access 18.02.2013]
15. Gerdes G, Tiedemann A, Zeelenberg S. Case study: a survey for the analysis of the expe-
riences and lessons learned by developers of offshore wind farms. Available online:
http://www.offshore-power.net, [access 10.03.2013]
Phys 113(17):17A324-117A324-3
18. Islam MR, Guo YG, Zhu JG, Dorrell D (2011) Design and comparison of 11kV multilevel
voltage source converters for local grid based renewable energy systems. In Proceedings of
37th annual conference of the IEEE Industrial Electronics Society (IECON 2011), 710 Nov
2011, Melbourne, Australia, pp 35963601
19. Islam MR, Guo YG, Jafari M, Malekjamshidi, Z, Zhu JG, (2011) A 43-level 33kV 3-phase
modular multilevel cascaded converter for direct grid integration of renewable generation
systems. In IEEE innovative smart grid technologies conference (ISGT) Asia, 2023 May
2014, Kuala Lumpur, Malaysia
20. Mitsubishi Electric, Powerex Power Semiconductor Solutions, IGBT/HVIGBT, [Online]
Available at: http://www.pwrx.com
grid integration of renewable generation units with medium voltage smart microgrids. In:
Large scale renewable power generation: advances in technologies for generation, transmis-
sion and storage. Springer, Berlin, pp 127149
Chapter 8
Conclusions and Future Works
Abstract This chapter concludes the book. In summary, this book has focused on
the optimal design of medium-voltage power converters for step-up-transformer-
less direct grid integration of renewable power generation systems and develop-
ment of a scaled down 1kV laboratory prototype test platform to describe the
implementation process of medium-voltage converter topologies. The design and
prototype development concept can be used to model the 1133kV system taking
into account some additional safety and reliability issues. Further, a few directions
have been recommended for future research and development.
8.1Introduction
Distribution of wind and solar photovoltaic (PV) energy throughout the world
should be first priority in solving our energy and environment crisis. Scientists
all around the world are now seeking energy solution from different renewable
resources and till today only wind and solar PV energy sources are found to be
suitable for future large-scale generation. By generating electric power from our
abundance renewable, such as wind and solar energy sources, we can solve a big
portion of energy deficiency. Moreover, the energy sector of the world is growing
curiously. It is facing an accelerating compound crisis of the globally established
fossil fuels. Immediate different breakthroughs for wind and solar PV power gen-
eration technologies are necessity to mitigate our fast growing energy demand. It
is expected that more than 80% future wind and solar PV power generation sys-
tems will be connected with the grids by 2030. Therefore, it is essential for scien-
tists and researchers to find the most effective converter technologies for the grid
integration of wind and solar PV power generation systems. Conventional grid
integration technique utilizes the power-frequency step-up-transformer, filter, and
booster. These heavy and large size power-frequency step-up-transformers signifi-
cantly increase the weight and volume of the renewable power generation systems.
Because of the heavy weight and large size of the power-frequency transformer,
the wind turbine generator and PV inverter system can be expensive and complex

276 8 Conclusions and Future Works
in terms of installation and maintenance especially in offshore and remote area

applications.
8.2Conclusions
In order to select the most suitable power converter topology, an extensive litera-
ture review was carried out on the existing renewable power generation technolo-
gies covering generation, conversion, transmission, and distribution issues. The
recent commercial medium and large-scale renewable power generation systems
were mainly considered. As these require large areas of land, they are usually
installed in offshore or remote areas, far from cities and industrial area where is
electricity is consumed. The available power converter topologies developed in the
recent decades were investigated for their offshore and remote area applications.
The recently introduced smart micro-grid concept for efficient power transmission
and distribution was also considered in the literature review.
Although several types of multilevel converter topologies have been devel-
oped in the last few decades, most of them are not suitable for medium-voltage
high-power offshore and remote applications. In order to find suitable multilevel
converter topology, an extensive analysis was carried out on different multilevel
converter topologies taking into account the specified converter performance, con-
trol complexity, cost, and market availability of the power semiconductors.
Because of some special features (e.g., number of components scale linearly
modular in construction hence enabling a high-level number attainability), the
MMC converter topology can be viewed as the best possible candidate for medium
voltage applications. However, the MMC converter requires multiple isolated
and balanced DC sources. A high-frequency magnetic link operated at a few kHz
to MHz was therefore developed to generate multiple isolated and balanced DC
sources for all the H-bridge inverter cells of the MMC converter from a single
source and a comprehensive electromagnetic analysis was conducted to verify the
feasibility of this new technology. Compared with the power-frequency transform-
ers, the high-frequency magnetic link has much smaller and lighter magnetic cores
and windings, thus much lower costs.
The multilevel converter requires a number of switching and control PWM
signals, which cannot be generated by the available DSP because the available
DSP can at present only provide about six pairs of PWM channels. Unlike the
DSP which runs a sequential program in its microprocessor, an FPGA may run
all the operations in parallel with the clock signal. In this instance, the FPGA is
the natural choice for medium voltage multilevel converters. Various design tech-
niques and software environments are available for the modeling of switching
control schemes with FPGA technology. Most of the techniques require special
software such as HDL coder, System Generator, PSIM, and ModelSim, which
increase the development time and cost. The most common software, such as the
8.2Conclusions 277
MATLAB/Simulink and Xilinx ISE-based design technique, was used to reduce

the developmental time and cost of the switching controller.
A high-frequency magnetic-link MMC medium-voltage converter could be the
possible option for step-up-transformer-less direct grid integration of renewable
sources. To verify the feasibility of the proposed system, a scaled down 1.73kVA
laboratory prototype system with modular 5-level cascaded converter was devel-
oped and reported, which converts 210V DC (rectified generator voltage) into
3-phase 1kVrms 50Hz AC. A Metglas amorphous alloy 2605SA1-based 10kHz
magnetic link was used to generate the isolated and balanced 6 DC supplies for the
3-phase 5-level MMC converter.
The high number of levels of the converter means that medium-voltage attain-
ability is possible so as to connect the renewable generation units to the medium-
voltage grid directly. The proposed MMC converter can also provide better output
power quality. Since the component number and control complexity increase lin-
early with the number of levels, the optimal selection of the number of converter
levels is important for the best performance/cost ratio of the medium-voltage con-
verter systems. An 11kV and a 33kV system were designed and analyzed for
optimal selection of the converter number of levels taking into account the speci-
fied system performance, control complexity, cost, and market availability of the
power semiconductors.
8.3Future Works
Although the reduced scaled prototype design, development, and experimentation

were described in this book in details, further work is required for the commer-
cial success of the high-frequency magnetic-link MMC medium-voltage converter.
The authors will be continuing their research for successful implementation of a
real-scale 1133kV converter for step-up-transformer-less direct grid connection
of wind and solar PV power generation systems. The future research findings will
be published and incorporated into a new edition of this book, with the following
issues taken into account:
In order to improve the efficiency of the high-frequency magnetic-link section,
advanced soft switching techniques can be implemented with controlled recti-
fiers rather than diode-based rectifiers;
It would be an interesting topic to have a detailed study and analysis of the elec-
tromagnetic design of the multi-inputmulti-output high-frequency magnetic
link to improve the efficiency of the system;
While only a few traditional modulation schemes have been considered in
this research book, it would also be interesting to study the effects of different
advanced modulation schemes for the MMC converter;
Since, the grid synchronization has not been considered in this book, a suitable
grid synchronization system could be described in future;
278 8 Conclusions and Future Works
Although the high-frequency magnetic link can minimize the voltage imbalance
and common mode issues, it would also be an interesting article to develop a
special modulation scheme with voltage imbalance and common mode minimi-
zation capabilities; and
In order to verify the feasibility and ensure the safety, reliability, and stability of
the real-scale high-frequency magnetic-link MMC medium-voltage converter, a
full voltage (e.g., 1133kV) test platform should be developed.

(Md. Rabiul Islam MMC Design

Uploaded by

Copyright:

Available Formats

(Md. Rabiul Islam MMC Design

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Md. Rabiul Islam MMC Design

Uploaded by

Copyright:

Available Formats

Green Energy and Technology

Power Converters for

ISSN 1865-3529 ISSN 1865-3537 (electronic)

Library of Congress Control Number: 2014947125

Springer Heidelberg New York Dordrecht London

Springer-Verlag Berlin Heidelberg 2014

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

To verify the feasibility of the high-frequency MMC converter, a scaled down

2 Power Converters for Small- to Large-Scale Photovoltaic

3 Power Converter Topologies for Grid-Integrated

4 Design and Characterization of High-Frequency Magnetic

5 FPGA-Based Digital Switching Controller for Multilevel Converters. . . 153

6 Experimental Validation of 1-kV Modular Multilevel

7 Design and Analysis of 11- and 33-kV Modular

8 Conclusions and Future Works. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

Fig. 1.1 A photograph of Anholt offshore wind power plant was

Fig. 2.15 Calculated PV curves at different irradiance ranging

Fig. 2.40 Centralized PV inverter topology. . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Fig. 3.24 Frequency spectrum of output line-to-line voltage with

Fig. 4.4 Modular medium-voltage converter for wind

Fig. 4.32 DC resistance measurement technique. . . . . . . . . . . . . . . . . . . . . . . 138

Fig. 5.3 Photograph of an FPGA by Xilinx. . . . . . . . . . . . . . . . . . . . . . . . . . 156

Fig. 6.3 Wind turbine generator system for direct medium-voltage

Fig. 6.30 A photograph of Tektronix digital phosphor

Fig. 7.7 A photograph of Semikrons SEMIKUBE converter

Table 1.1 Land area covered by offshore wind farms. . . . . . . . . . . . . . . . . . 2

Table 4.9 Transformation ratio calculations at 50 Hz (DF) . . . . . . . . . . . . . . 140

A Cross-sectional area (m2)

T Time period (s)

Ldl Differential leakage inductance (H)

Vt, Vf Voltage drop at zero current of IGBT and diode (V)

SDFM Single doubly-fed machine

Md. Rabiul Islam (M14, IEEE) received the

Youguang Guo (S02-M05-SM06, IEEE) re-

Jianguo Zhu (S93-M96-SM03, IEEE) re-

Wind farms PV power plants Power converter topologies Multilevel converters

AbstractThis chapter presents a succinct preface focusing on various aspects,

KeywordsEnergy and environment challengesRenewable energyOffshore

1.1Most Dominating Renewable Power Sources

Springer-Verlag Berlin Heidelberg 2014 1

1.1.1Historical Growth of Wind Power Generation Capacity

Table1.1Land area covered by offshore wind farms

1.1.2Technical Challenges and Possible Solutions

To integrate scattered wind turbine generators into a medium-voltage grid (e.g.,

Table1.2Global offshore Years Annual addition (MW) Cumulative

Table1.3Global wind Years Annual addition (MW) Cumulative

Fig.1.3Wind turbine with step-up transformer

Fig.1.5A photograph of offshore wind farm under maintenance [25]

1.1.3Historical Growth of Solar PV Power Generation

Table1.4Global solar PV Years Annual addition (MW) Cumulative

Fig.1.6A photograph of Beneixama 20MW PV power plant [27]

1.1.4Technical Challenges of Solar PV Power Generation

Fig.1.7A photograph of step-up transformer installation of Tritec 5.2MW solar PV power

Neuendorf AG into the medium-voltage grid. Tritec installed SolarMax trans-

1.1.5Possible Solution to Technical Challenges

1.2Major Objectives of the Book

The major objectives of this book are as follows:

2. to investigate the different multilevel converter topologies, taking into account

Fig. 1.1 A photograph of Anholt offshore wind power plant was

Fig. 2.15 Calculated PV curves at different irradiance ranging

Fig. 6.3 Wind turbine generator system for direct medium-voltage

Fig. 6.30 A photograph of Tektronix digital phosphor

Fig. 7.7 A photograph of Semikrons SEMIKUBE converter

1.1Most Dominating Renewable Power Sources

1.1.1Historical Growth of Wind Power Generation Capacity

1.1.2Technical Challenges and Possible Solutions

1.1.3Historical Growth of Solar PV Power Generation

1.1.4Technical Challenges of Solar PV Power Generation

1.1.5Possible Solution to Technical Challenges

1.2Major Objectives of the Book

1.3Contribution of the Book

1.4Organization of the Book

2.2Solar Photovoltaic Arrays

2.2.1Solar Photovoltaic Array Modeling

2.2.2Solar Photovoltaic Array Characteristics

2.2.3Solar Photovoltaic Array Maximum Power Point

2.3Inverters in Small-Scale Solar PV Systems

2.3.1Two-Stage Solar PV Inverters for Small-Scale Systems

2.3.2Multiple-Stages Solar PV Inverters for Small-Scale

2.3.3Single-Stage Solar PV Inverter for Small-Scale

2.3.4ABB and Siemens Solar PV Inverters for Small-Scale