Dinh Thesis

Harmonic Domain Modelling of
Direct Connected Generator and

HVDC Convertor units
Qinh Nhut-Quang, BE (Hons.)
/'
A thesis presented for the degree of
Doctor of Philosophy
III
Electrical and Electronic Engineering
at the
University of Canterbury,
Christchurch, New Zealand.
January 1998
ci'-lGINEERING
lIBRp,RY
'lIZ
\OlO
.1)584-
~ ~ ~
ABSTRACT
This thesis describes a new method of analysing the harmonic content of direct connected
generator-HVdc convertor units. Existing analysis techniques are reviewed, and their
strengths and weaknesses are discussed. A set of simultaneous equations that fully
describes the harmonic interaction among the generators, convertors and the dc system
is derived. The simultaneous equations are solved using Newton's method. By exploiting
the sparse nature of the Jacobian matrix of partial derivatives, the solution is obtained
quickly. The solution of a test system is verified by means of dynamic simulations. The
developed unified solution is a significant improvement on existing harmonic analysis.
A harmonic model of the salient-pole generator is developed and validated against
time domain simulations and harmonic measurement results obtained from the Benmore
convertor station when being operated as a group connected unit. The relationship
between the generator terminal voltages and currents in the presence of saliency can
be represented by analytical expressions in sequence components, derived from normal
machine parameters. The effect of generator saturation on the harmonic content is not
significant. The effect of rotor angle on the commutation process in the presence of
saliency is significant in direct connected units.
By choosing the generator terminal voltage as the firing angle reference, existing
firing control for conventional HV dc schemes can be used for direct connected units with
practically no modification. Provided that the nominal firing angles are kept within
the restriction limits, imposed by the overlap angle, generator terminal power factor,
equivalent continuous negative-sequence harmonic loading and the number of generators,
practically no extra capacity for harmonic current loading is required in direct connected
units. The generator harmonic ratings can be taken at the nominal speed and rated
power output. Conventional tuned dc filters are ineffective in adjustable speed operation.
Finally, detailed three-phase harmonic information of the Benmore convertor station
when being operated in the group connected mode is collected and discussed. The test
results are used for validating the developed models, and they also provide valuable
evidence of the feasibility of direct connected units.
ACKNOWLEDGEMENTS
Completion of this thesis marks the end of an invaluable, positive learning experience.
The most sincere thanks must go to Professor Jos Arrillaga, whose knowledge, super-
vising experience and enthusiasm is invaluable. I am extremely grateful for his time,
numerous advice and inspiration.
Many thanks are also due to Drs Alan Wood, Bruce Smith, Chris Arnold, Neville
Watson, Dave Watson, Pat Bodger, and Mr Mike Dewe for countless discussions and
advice; Ken Smart for being extra helpful all the time. I would like to thank the
computing staffs who have courageously kept the computer systems up and running;
my postgraduate colleages, particularly Drs Roger Brough, Shiun Chen, Wade Enright,
Maria-Luiza Lisboa, Stu MacDonald, Du Zhen-Ping; Graeme Bathurst, Dave Hume,
Thomas Keppler, Simon Todd and Andrew Ward for the many lively discussions and
functions. Special thanks to my many other friends who have been providing valuable
friendship and much needed distractions, especially Parapol Anusarnsunthorn, Michael
Burns, Dr Felix Hartanto, Dr David Hawkins, Dr Matt Hebley, Jos Li, Paul Johnstone,
Nu, Elwyn Smith, Sugar, and Tran Minh-Tri. Many thanks to Nu for proof-reading
parts of this thesis.
The financial support from the University of Canterbury through the Doctoral schol-
arship; from the University of Canterbury Power Systems Research group; from Trans
Power New Zealand, ECNZ, and the CHART development group for the Benmore con-
vertor harmonic measurements are gratefully acknowledged.
I wish to express my deepest gratitude to my Mom, Dad, my brother Tan and the
extended family in VietNam, whose invaluable life long encouragement and support have
helped lighten the long term effort. Thank you so much.
And finally, I wish to extend the warmest thanks and gratitute to my sponsor
families, especially Mr & Mrs Leo and Joan Clark; Michael, Susan and Bridget Coulter;
Charlotte and Andrew Robb; Paul, Jenny and Kelly Clark; Mr & Mrs Macann, whose
generosity and unassuming support have made this work possible.
CONTENTS
ABSTRACT
ACKNOWLEDGEMENTS
GLOSSARY
CHAPTER 1 INTRODUCTION
1.1 HVdc Transmission
1.2 Main Objectives
1.3 Thesis Outline
CHAPTER 2 DIRECT CONNECTION SCHEMES AND
HARMONIC MODELLING REVIEW
2.1 Introduction
2.2 Unit Connection Concept
2.3 Main Applications
2.4 A Review of Harmonic Analysis methods
2.4.1 Time Domain Simulations
2.4.2 Harmonic Domain Modelling
2.4.2.1 Simplified Convertor Formulation
iii
v
xix
1
1
2
3
5
5
5
8
8
9
10
10
2.4.2.2 Iterative Harmonic Analysis 11
2.4.2.3 Method of Norton Equivalents 14
2.4.2.4 Unified Solution using Newton's Method 14
2.5 Conclusions 15
CHAPTER 3 GENERATOR MODELLING IN THE HARMONIC
DOMAIN 17
3.1 Introduction 17
3.2 The Process of Frequency Conversion 18
3.3 Generator Harmonic Impedances in Sequence Components 19
3.3.1 Frame of Reference and Rotor Angle 19
3.3.2 Negative-Sequence Harmonic Impedances 20
3.3.3 Positive-Sequence Harmonic Impedances 21
3.3.4 Zero-Sequence Harmonic Impedances 23
3.3.5 Skin Effect 23
3.4 Synchronous Generator Harmonic Equivalents 24
3.5 Saturation Effects on Steady State Operation 26
viii CONTENTS
CHAPTER 4
CHAPTER 5
3.5.1 Saturation Effects on Generator Harmonic Im-
pedances
3.5.2 Generator Saturation Harmonics
3.6 Model Validation
3.7 Levels of Harmonic Impedances
3.8 Conclusions
CONVERTOR ANALYSIS IN THE HARMONIC
DOMAIN
4.1 Introduction
4.2 Commutation Analysis for a Unified Iterative Solution
4.2.1 Star Connection Analysis
4.2.2 Delta Connection Analysis
4.3 Convertor Control and Firing Process
4.3.1 Firing Process
4.3.2 Constant Current Control
4.3.3 Constant Power Control
4.4 Direct Voltage
4.4.1 Star Connection Voltage Samples
4.4.2 Delta Connection Voltage Samples
4.4.3 DC Voltage Construction by Convolution
4.5 Secondary Phase current
4.6 Transformer Modelling
4.7 Summary
A HARMONIC DOMAIN UNIT CONNECTION
MODEL
5.1 Introduction
5.2 System Variables
5.3 Generator-Convertor Interaction
5.3.1 Fundamental Frequency
5.3.2 Harmonic Frequencies
5.4 Convertor Interaction with dc System
5.5 Reactive Power and Voltage Control by Generator Excitation
5.5.1 Firing Angle Referenced to the Generator Terminal
Voltage
5.5.2 Firing Angle Referenced to the Generator Internal
Voltage E"
5.6 Functional Description of a Unit Connection
5.6.1 Phase Components Formulation.
5.6.2 Sequence Components Formulation.
5.7 Newton's Solution
5.7.1 Jacobian of Partial Derivatives
5.7.2 Initialization
5.8 Model Validation and Performance
27
30
32
33
39
41
41
41
42
43
45
45
45
46
47
47
48
50
52
53
57
59
59
60
62
62
62
63
63
63
64
65
65
66
68
69
72
74
CONTENTS
5.8.1 Validation
5.8.2 Convergence Properties
5.9 Conclusions
ix
74
77
80
CHAPTER 6 HARMONIC BEHAVIOUR OF DIRECT
CONNECTED UNITS 83
6.1 Introduction 83
6.2 Firing Angle References for Direct Connected Units 84
6.3 Overlap Angle Constraint 86
6.4 Generator Terminal Power Factor
6.5 Effects of Salient-Pole Generators
86
89
6.5.1 Effect of Generator Saliency on Harmonic Levels 89
6.5.2 Effect of Saliency on Generator Terminal Power Factor 92
6.5.3 Effect of Rotor Angle 95
6.5.4 Effect of Generator Saturation 95
6.6 Group Connection Implementation 97
6.6.1 Commutation Angle and Generator Power Factor
in Group Connections
6.6.2 Harmonic Content in Group Connections
6.7 Effects of Adjustable Speed Operation
6.8 Effects of dc Side Resonances
6.8.1 On dc Side Harmonics
6.8.2 On ac Side Harmonics
6.9 Conclusions
CHAPTER 7 HARMONIC EVALUATION OF BENMORE
CONVERTOR STATION WHEN OPERATED AS A
DIRECT CONNECTED UNIT
7.1 Introduction
7.2 Testing Procedure
7.3 Fundamental Frequency Variation
7.4 Three-Phase Harmonic Information
7.5 Harmonic Variation with dc Current Order
7.5.1 DC signals
7.5.2 AC harmonics
7.6 Generator Harmonic Ratings
7.7 Conclusions
CHAPTER 8 CONCLUSIONS AND FUTURE WORK
8.1 Conclusions
8.2 Suggestions for future work
8.2.1 Transformer Saturation Modelling
8.2.2 Subsynchronous Resonances and Interharmonic
99
99
105
106
106
113
114
115
115
116
117
117
121
121
123
123
131
133
133
135
135
Studies 136
x CONTENTS
APPENDIX A TEST SYSTEMS DATA 137
137
137
A.l Hwang's Generator Parameters
A.2 Benmore's Generators and transformers
APPENDIX B SALIENT-POLE GENERATOR HARMONIC
IMPEDANCE CALCULATION FROM MEASURED
HARMONIC VOLTAGES AND CURRENTS 141
APPENDIX C CUBIC SPLINE INTERPOLATION 143
APPENDIX D HARMONIC PHASOR REPRESENTATION
D.l Complex Fourier Series
145
145
146
147
D.2 Positive Frequency Harmonic Phasors
D.3 Convolution of Harmonic Phasors
APPENDIX E SIMPLIFIED CONVERTOR COMMUTATION
ANALYSIS AND HARMONIC CALCULATION 149
APPENDIX F JACOBIAN OF PARTIAL DERIVATIVES 151
F.l Voltage Control Equation Partial Derivatives 153
F.2 Terminal Voltage Partial Derivatives 153
F.2.1 With Respect to ac Voltage Variation 154
F.2.2 With Respect to dc Ripple Current Variation 157
F.2.3 With Respect to end of Commutation Variation 159
F.2.4 With Respect to Firing Instant Variation 161
F.3 Direct Current partial derivatives 162
F.4 End of Commutation mismatch Equation partial derivatives 168
F.5 Firing Instant mismatch Equation partial derivatives 171
F.6 Average delay angle partial derivatives 172
APPENDIX G PUBLISHED PAPERS 175
REFERENCES 177
LIST OF FIGURES
2.1 Conventional HVdc sending end arrangement. 6
2.2 Unit Connection configuration. 6
2.3 An example of a Group Connection arrangement. 7
2.4 Effects of arbitrary snubber capacitances on the harmonic levels III
simulations. 10
2.5 Matched reactance pair method for six-pulse convertors. 12
2.6 Matched reactance pair method for twelve-pulse convertors. 13
3.1 Response of a salient pole generator to the presence of a harmonic current. 18
3.2 Phasor diagrams of an unsaturated salient-pole synchronous generator
operating at a lagging power factor.
3.3 Generator Thevenin and Norton harmonic equivalents.
3.4 Structure of the generator impedance or admittance matrix in phase co-
ordinate and sequence components.
3.5 Modified equivalent circuit.
3.6 Graphical construction for finding the components of armature voltage
due to saturation Es corresponding to Potier voltage Ep.
3.7 Determination of rotor angle under saturation for salient-pole synchronous
19
24
25
26
28
generators operating at a lagging power factor. 29
3.8 Test system for model validation. 33
3.9 Model validation by comparison of time and harmonic domain solutions
for harmonic current injections. 34
3.10 Generator harmonic impedances obtained from the measurement and cal-
culation. 35
3.11 Generator harmonic impedances for a particular loading. 36
3.12 Magnitude and phase angle of cross coupling terms for various loadings. 37
3.13 Generator-transformer-convertor unit. 37
3.14 Variation of the 7th harmonic voltage in relation to the rotor angle and
firing delay angle. 38
xii LIST OF FIGURES
4.1 Circuit for star-g/star commutation analysis.
4.2 Circuit for star-g/ delta commutation analysis.
4.3 Constant current controller.
4.4 Circuit for a conduction period with a delta connected source.
4.5 Sampling functions used in convolution.
4.6 Equivalent circuit for star-g/star transformer.
4.7 Equivalent circuit for star-g/delta transformer.
43
44
45
49
50
54
55
5.1 Twelve-pulse Unit Connection. 60
5.2 Dependency among system variables. 61
5.3 Phasor diagram of a salient-pole generator in direct connected units. 64
5.4 Sparsity structure of the Jacobian in sequence components for 13 harmonics. 71
5.5 Flow diagram for the unified harmonic solution.
5.6 Model intervalidation by comparison of time domain and unified harmonic
solutions for generator voltage and current waveforms.
5.7 Model intervalidation by comparison of time domain and unified harmonic
solutions for generator harmonic voltages and currents.
73
76
77
5.8 DC voltage waveforms reconstructed from 50 harmonics. 78
6.1 Phasor diagram of a non-salient generator in direct connected units. 84
6.2 Relationship between firing angles referenced to the generator fundamen-
tal terminal voltage and internal voltage E" . 85
6.3 Comparison of commutation periods for direct connected units and perfect
ac filtering schemes.
6.4 Comparison of (a) internal displacement factors cos(e), and (b) generator
terminal power factors cos for direct connected units and perfect ac
86
filtering schemes. 88
6.5 Phasor diagrams of saliency effect on harmonic voltages. 90
6.6 (a) Maximum and minimum limits for the 11th current harmonics of a
salient-pole generator in direct connected units; (b) Voltage harmonics
corresponded to the current limits. 93
6.7 Commutation angles limits for salient-pole generators in direct connected
units. 94
6.8 Generator terminal power factor (cos ) for salient-pole generators. 94
6.9 Effect of rotor angle on the commutation angle (a), and on 11th harmonic
current level (b) in the presence of generator saliency. 96
LIST OF FIGURES
6.10 Simplification of a Group Connection to the basic Unit Connection for
harmonic analysis.
6.11 (a) Commutation angle and (b) generator terminal power factor in a
xiii
97
Group Connection being operated with various number of generators. 100
6.12 Generator harmonic currents in a Group Connection being operated with
various number of generators.
6.13 Generator harmonic voltages in a Group Connection being operated with
various number of generators.
101
102
6.14 Generator equivalent negative-sequence currents for a Group Connection. 103
6.15 Relationship between firing angles referenced to E" and to the generator
terminal voltage V for a Group Connection.
6.16 Typical hydraulic turbine power-speed characteristic.
6.17 (a) Commutation angle and (b) generator terminal power factor variation
with speed for various firing angle.
6.18 Harmonic current variation with speed for various firing angle.
6.19 Harmonic voltage variation with speed for various firing angle.
6.20 A typical dc side configuration with a dc filter, smoothing inductor and a
simplified representation of a dc line.
6.21 (a) Harmonic impedance of the dc filter. (b) Harmonic impedances of a
dc system consisting of a smoothing inductor (0.8 H), dc filter and a line
103
105
107
108
109
110
inductance of 0.2 H. 111
6.22 Harmonic voltages at the dc filter point for adjustable speed operation. 112
6.23 Generator current harmonics at different generator speed in the presence
of dc side resonances. 113
7.1 Benmore group connection and CHART measurement points. 116
7.2 Fundamental frequency variation during the test measurement. 118
7.3 Three-phase generator bus voltages and their harmonics in sequence com-
ponents for a dc current of 1016 amps. 119
7.4 Three-phase generator currents and their harmonics in sequence compo-
nents for a dc current of 1016 amps. 120
7.5 Variation of the zero-sequence voltage harmonics with respect to the fun-
damental16 kV bus voltage. 121
7.6 DC voltage level for various dc current orders. 122
7.7 DC current and harmonics for the current order of 1016 Amps. 122
xiv LIST OF FIGURES
7.8 Variation of the measured generator terminal power factor angle (lagging),
average firing angles (referenced to the measured generator terminal volt-
age) and commutation angles for a range of dc current.
7.9 Comparison of generator terminal voltage waveforms obtained from the
124
test measurement and harmonic domain solution. 124
7.10 Variation of the generator current harmonics for a dc current range. 125
7.11 Variation of the generator voltage harmonics for a dc current range. 126
7.12 (a) Commutation angle and generator terminal power factor at the rated
power output for the Benmore Convertor station operating as group con-
nected units; (b) Relationship between E" referenced and generator ter-
minal voltage V referenced firing angles.
7.13 Generator harmonic current and voltage levels for the Benmore Convertor
station operating as group connected units. Firing angles are referenced
128
to the generator terminal voltage. 129
7.14 Generator equivalent negative-sequence currents for the Benmore Conver-
tor station operating as group connected units.
7.15 Convertor transformer current THD for the Benmore Convertor station
operating as group connected units.
A.1 Benmore generator open circuit saturation curve.
C.1 Cubic spline interpolation of the Benmore generator open circuit charac-
130
130
139
teristic. 144
E.1 Star-star transformer primary current and its harmonic content, commu-
tation period is 13 degrees. 150
F.1 Partial derivatives in the Jacobian. 152
LIST OF TABLES
2.1 . Iterations to convergence for six-pulse unit connected IRA algorithms. 12
4.1 Construction of dc voltage and ac phase current samples. 48
4.2 Limits of convertor states for use in sampling functions. 51
5.1 Mismatch equations for a 12 pulse convertor in phase coordinates. 66
5.2 Mismatch equations for a 12 pulse convertor in sequence components. 67
5.3 The selection of the cut-off limit for Jacobian elements in order to maxi-
mize solution speed and preserving the robustness of the solution. 70
5.4 Iterations to convergence for the unified Unit Connection solution using
5.5
6.1
6.2
6.3
Newton's method. 78
Solution speed and convergence factor in relation to the magnitude of
generator subtransient reactances for solutions up to the fiftieth harmonic. 79
Minimum firing angles for various overlap angle limits. 87
Adjustment for the generator and convertor parameters of Group Con-
nections operating with a reduced number of generators. 98
Nominal firing angle restriction limits for group connected units imposed
by overlap angle, mis-firing, generator power factor and continuous equiv-
alent negative-sequence harmonic loading constraints. 104
7.1 Generator harmonic ratings for different group connected configuration
at the Benmore convertor station; *results from the 1993 measurement. 127
7.2 Adjustment of parameters for the Benmore system operating in group
connected and unit connected mode. 127
A.1 Generator harmonic ratings for twelve-pulse group connected operation
at the Benmore convertor station. 138
A.2 Benmore generator open circuit saturation characteristic. 138
F.1 The coefficient matrix Caoi which specifies the dependency among com-
mutation current i, terminal voltage phase, and ac current phase 155
F.2 Coefficient matrix Ei defining the contribution of the commutation cur-
rents to each phase current. i is the commutation number. 158
F.3 Construction of dc voltage partial derivatives. 164
F.4 Limits of convertor states for use in sampling functions. 166
GLOSSARY
Subscripts and Superscripts
+, -
0,+,-
a, b, c
b
D
d
e
9
k

n
a
p
q
p
S
S
Symbols
*
n
I
fJy
fJx
j
(3
(3
<5
phase connected to positive and negative dc rails respectively
Three sequence labels
Three phase labels
Phase beginning conduction
Associated with star-g/ delta group
Associated with direct-axis
Phase ending conduction
Associated with generators
Harmonic order
Firing number, i = 1...6
Integer n = 1,2,3 ...
Phase not participarting in a commutation
Conduction interval number, p = 1...12
Associated with quadrature-axis
Primary side
Secondary side
Associated with star-g/star group
Convolution operator
Conjugation operator
Real part of
Imaginary part of
Partial derivative of y with respect to x
A
Firing angle
Equidistant timing reference
An arbitrary phase angle
Rotor angle
Rotor angle under saturation condition
Firing instants
xviii
w
Fao
Fo
Frp
FI
F
Id
Fs
Fv
FVd
Fw
G
i (t)
ic(t)
~
~ q
I
Convergence factor
Average commutation duration
Power factor angle
End of commutation instants
Fundamental angular frequency
Flux linkage
Crossings of generator fundamental terminal voltages
Off nominal tap ratios
Commutating current DC offset
Generator Thevenin voltage source harmonics
Constant DC voltage source
Generator excitation voltage
Potier voltage
Magnetic saturation function
Full mismatch vector in Newton's method
Average delay angle mismatch
Firing mismatches
End of commutation mismatches
AC side harmonic current mismatches
Direct current harmonic ripple mismatches
Rectified power mismatch
AC terminal harmonic voltage mismatches
DC side harmonic voltage mismatches
GLOSSARY
Sequence components ac side harmonic voltage mismatches
Current transducer gain
Time domain current waveform
Time domain commutating current waveform
Time domain generator direct-axis current
Time domain generator quadrature-axis current
Convertor three-phase ac side harmonic currents
Commutating current harmonics
Convertor dc side harmonic currents
(chapter 3) Generator direct-axis current
N orton harmonic current source
Generator quadrature-axis current
Jacobian Matrix
Commutating inductance
Generator direct-axis inductance
Generator quadrature-axis inductance
GLOSSARY
p
p
Q
R
S
T
TJ
v(t)
V
Vd
Vdp
W
X
Xc
Xd
Xd
Xp
Xq
X"
X"
d
X"
q
Zgen
A b brieviations
CIGRE
HVdc
IHA
MHLF
Current controller proportional gain
Real power
Reactive power
Resistance
Complex power
Sequence transform matrix
Current controller integral time constant
Time domain voltage waveform
Convertor three-phase ac side harmonic voltages
Convertor dc side voltage harmonics
pth direct voltage sample harmonics
Convertor ac side terminal voltage in sequence components
Solution vector in Newton's method
Commutating reactance
Generator direct-axis reactance
DC side reactance
Potier reactance
Generator quadrature-axis reactance
Generator subtransient reactance
Generator direct-axis subtransient reactance
Generator quadrature-axis subtransient reactance
Generator harmonic impedance
Conference Internationale de Grands Reseaux Electriques
High Voltage direct current
Iterative Harmonic Analysis
Multiphase Harmonic Load Flow
xix
Chapter 1
INTRODUCTION
1.1 HVDe TRANSMISSION
DC transmission dated back from the dawn of electricity, when Marcel Deprez trans-
mitted 1.5 kW at2 kV for over 35 miles in 1882. As low voltage levels from generators
limited the transmission distance, stepping up the voltage through transformers was nec-
essary. This effectively eliminated the use of dc in power transmission until the advent
of the mercury arc valves that enabled high voltage rectification, and thus HV dc (High
Voltage Direct Current) transmission for longer distances. More reliable thyristor valves
gradually took over the mercury arc valves since the mid-seventies, reducing costs and
increasing reliability. Worldwide power transfer capacity has increased from a mere 20
MW in the first HV dc scheme connecting mainland Sweden and Gotland (in 1954) to
over 46 000 MW in recent time [Narain 1996].
HV dc has several distinctive advantages over ac transmission systems, particularly
when being used to transmit power over long distances, or interconnect large asynchro-
nous ac systems operating at the same or at different frequencies. The asynchronous
nature of a dc link does not appreciably increase the fault level, and subsequently the
required ratings of existing equipment. Moreover dc links can be used to improve ac
system stability as the power flow in a dc scheme can be controlled at high speed [Arril-
lag a 1983]. In the case of long submarine power transmission, HVdc is the only choice
since the voltage rating of the submarine cables makes ac transmission infeasible.
When the generated power is supplied exclusively to a HV dc transmission link, such
as in an isolated power station, it is advantageous to simplify the rectifier station, via a
direct connection of each machine set to a separate convertor group with series-parallel
combinations made on the dc side. In these schemes, ac harmonic filters, ac switching
and protection gears, and a transformation step are dispensed with. The convertor valve
halls and power station can be integrated into a HV dc Power Station, resulting in further
reduction in land requirement, civil and construction works. Cost savings in HV dc direct
connection schemes have been estimated at 30 to 40% in comparison with conventional
schemes [CIGRE JWG 14/11-09, 1997].
2 CHAPTER 1 INTRODUCTION
1.2 MAIN OBJECTIVES
The absence of ac filters and close proximity of the generators and convertors lead to
strong harmonic interactions in direct connected units. The convertor harmonic currents
flowing into the generator stator windings cause additional thermal stresses, parasitic
torques and produce harmonic voltage distortion at the convertor terminal, which in-
creases the insulation requirement for both generators and convertor transformers. Large
voltage distortion also influences the commutation process, and thus affecting ac har-
monic currents. In addition, the frequency conversion nature ofthe salient-pole generator
adds another dimension of complexity to the harmonic interaction.
So far, electromagnetic transient simulation oftwelve-pulse direct connected conver-
tor units to the steady state, based on the solution of differential equations using step
by step integration technique [DommeI1969], is the only satisfactory harmonic analysis
method for such systems. However this is a very inefficient approach considering the
long time constants of the synchronous generator, and the great number of simulations
needed to fully describe the harmonic behaviour. When modelling convertor operation,
extra snubber capacitance is often added to aid the numerical stability of such programs,
however that artificial capacitance can introduce substantial errors in the harmonic cal-
culation.
On the other hand, simplified convertor formulae [Kimbark 1971] can not represent
the effect of harmonic distortion on ac and dc convertor terminal, and the frequency
coupling caused by rotor saliency in the presence of current unbalance or distortion.
Such effects can be adequately modelled in the Harmonic Domain [Semlyen et al. 1988].
There are essentially two types of solution techniques, sequential and unified, that have
been developed for convertor harmonic analysis. Sequential solution techniques such as
Iterative Harmonic Analysis (IHA), which solves the convertor and ac system interaction
in two separate steps, have suffered convergence problems when being used to analyse
harmonics in direct connected units. Non-convergence or divergence is caused by large ac
terminal voltage distortion due to substantial generator harmonic impedances [Callaghan
and Arrillaga 1990]. Convergence can be improved when all the non-linearities are solved
together in a unified solution [Smith et al. 1996]. But so far, unified solutions for direct
connected units have not been developed.
The research effort described in this thesis was motivated by the desire to improve
the harmonic analysis techniques as well as the understanding of the harmonic interac-
tion in direct connected schemes. The principal goal was to develop a fast and accurate
harmonic model of direct connected generator-HVdc convertor units using a unified solu-
tion technique as a suitable tool for harmonic analysis of these schemes. The developed
model provides a platform for investigating the harmonic behaviour of direct connected
units under the influence of generator saliency and saturation, adjustable speed opera-
tion, dc system resonances, etc.
1.3 THESIS OUTLINE 3
A detailed harmonic generator model which takes into account the rotor angle ef-
fects- on saliency and saturation is also developed since the present Harmonic Domain
generator model [Medina and Arrillaga 1994] does not include these critical factors.
The generator model is expressed as a harmonically coupled impedance matrix with
elements derived from the normally available generator parameters such as subtransient
reactances.
Detailed three-phase harmonic information of the Benmore convertor station when
being operated in the group connected mode, collected in November 1995, is presented.
The test results are used to assess the accuracy of the developed generator and direct
connected models, and they also provide valuable evidence of the feasibility of direct
connected schemes.
1.3 THESIS OUTLINE
Chapter 2 summarizes the direct connection concept, and critically reviews the advan-
tages and disadvantages of existing methods for harmonic analysis of direct connected
units.
Chapter 3 derives a salient-pole generator harmonic model in sequence components.
Sequence components formulation is more desirable than phase components as the gen-
erator converts harmonics from one sequence to another in the steady state, thus gives
insight understanding into the harmonic couplings. The derived model is validated by
means of time domain simulations. The harmonic measurement results obtained from
the Benmore convertor station when being operated as a group connected unit is also
used to validate the developed generator model.
Chapter 4 summarizes a harmonic domain convertor model developed recently
by Smith et al. [1995], expressed in terms of simultaneous equations that fully describe
the convertor in the steady state.
The generator and convertor models are integrated into a twelve-pulse Unit Connec-
tion harmonic model in Chapter 5. The generator, convertor and dc system harmonic
interaction are formulated and solved in a unified algorithm using Newton's method.
Since unbalance is accounted for in the commutation process, the distorted generator
voltage can be chosen as a pseudo-commutation voltage. The voltage control by gen-
erator excitation in the absence of convertor transformer on-load tap-changers is also
modelled. The solution of a test system is verified against a time domain simulation.
In Chapter 6, the developed model is used to investigate the effects of generator
saliency, rotor angle and saturation, the number of generators in Group Connections,
adjustable speed operation, and dc side resonances on the harmonic levels. The firing
angle references, generator terminal power factor rating, and overlap angle constraints
in direct connected units are also examined.
4 CHAPTER 1 INTRODUCTION
Chapter 7 reports on the three-phase harmonic measurement of a HVdc Group Con-
nection, containing four 112.5 MVA hydro generators at the Benmore end of the New
Zealand HV dc link. The test measurement results, procedures, setup and configura-
tion of the CHART (Continuous Harmonic Analysis in Real-Time) harmonic monitor-
ing instrumentation are described. The results include measured waveforms and their
harmonic content in sequence components for a range of dc line current. Using the
developed direct connected model, generator harmonic levels are derived for the unit
connected mode, which has not been operated at the Benmore terminal, to provide evi-
dence in terms of harmonic ratings for the viability of this mode of operation. Finally,
the overall conclusions are drawn in Chapter 8.
Chapter 2
DIRECT CONNECTION SCHEMES AND HARMONIC
MODELLING REVIEW
2.1 INTRODUCTION
Since the steady state performance is critical in the operation of direct connected units,
it is necessary to be able to predict and quantitatively assess the harmonic behaviour of
such systems at the design stages. Existing methods of harmonic analysis applied to large
convertor installations are critically reviewed in this chapter, with particular attention
paid to the ability of the techniques to evaluate the strong harmonic interaction in direct
connected units.
The basic direct connection concept, typical configuration, main advantages, disad-
vantages, and possible applications of direct connected schemes are also summarized in
this chapter.
2.2 UNIT CONNECTION CONCEPT
In the conventional HV dc systems, generators are allocated to a common ac bus which
feeds all the convertors, as illustrated in Figure 2.1. The ac harmonic filters on the
convertor bus prevent characteristic convertor harmonics from penetrating into the gen-
erators or further into the ac network. However the presence of filters also imposes
steady state operation at a fixed electrical frequency. The large filter capacitances can
cause generator self-excitation on the occurrence of load rejection, resulting in substan-
tial dynamic overvoltages.
The possibilities of gaining efficiency and reducing cost by using adjustable speed
operation and simplifying the sending terminal are the main motivations behind the
direct connection concept [Calverley et al. 1973]. In its simplest form, the concept takes
the form of a "one-to-one" configuration, or a Unit Connection, in which the ratings of
the individual components are matched, as shown in Figure 2.2. The basic feature is
that the generator transformers and all the ac filters are removed and the generators are
directly connected to twelve-pulse convertors.
6 CHAPTER 2 DIRECT CONNECTION SCHEMES AND HARMONIC MODELLING REVIEW
To AC network
AC line
12 Pulse Convertor
Electrode
AC line
Filter bank
Figure 2.1 Conventional HVdc sending end arrangement.
12 Pulse Convertor
Electrode
Figure 2.2 Unit Connection configuration.
2.2 UNIT CONNECTION CONCEPT 7
12 Pulse Convertor
Electrode
Figure 2.3 An example of a Group Connection arrangement.
The absence of tuned ac filters enables the generating units to operate within an
adjustable speed/frequency band, rather than at a fixed speed. Since hydro turbines
are intrinsically adjustable speed machines, net energy generation, unit reliability and
service life duration of turbine can all be increased. Furthermore, the simplified layout
with substantially reduced equipment requirement leads to less land requirement at the
HV dc sending end. Operating and maintenance costs will lessen accordingly. The sav-
ing in capital costs for the convertor station is conservatively evaluated at 30 to 40%
in comparison with the conventional schemes [CIGRE WG 11/14-09, 1993]. Substantial
overvoltages due to the large filter capacitance during faults or power swing is elimi-
nated, leading to reduced insulation over-rating factors. In addition, current and power
flow are unaffected by ac system disturbances from major faults or oscillations. It is also
proposed that the voltage and reactive power control be exercised entirely by the gener-
ator excitation. As a result, the sending end transformers on-load tap changers are no
longer required. This control action is taken into account in the developed formulation
in Chapter 5.
Without ac filters the convertor harmonic currents flow through the generator stator
windings, causing additional thermal stresses and ripple torques. In the event that
generators are operated at frequencies other than power frequency, a separate source
of power for station auxiliaries must be arranged. The basic one-to-one arrangement
however, lacks flexibility on the account that each individual generating unit is dedicated
to a single convertor. In the case of a series connection on the dc side, for instance, dc
transmission voltage has to be reduced with loss of efficiency if one or more units are
down. Additional equipment will be required for the station to cope with outages. For
this reason, one-to-one arrangement is not recommended in general. It may however,
be possibly suitable for stations with only a few very large units and for small, compact
and isolated stations.
A variation ofthe basic one-to-one arrangement that offers more flexibility of opera-
tion and of maintenance is the Group Connection [Uhlman 1975], in which more than one
generator can be connected to a twelve-pulse convertor. When there are a large number
of generating units involved, a combination of parallel Group Connection could be used,
as illustrated in Figure 2.3. The generator transfer bus and ac switchgear are required.
The Group Connection can maintain a constant dc voltage at reduced dc power levels
which keeps the transmission efficiency high [Arrillaga et al. 1993b]. The matching of
generator and convertor ratings is no longer required. Adjustable speed operation is still
allowed although the group of synchronized generators will require some form of joint
speed control.
Nevertheless, the basic one-to-one arrangement is ideal for harmonic studies as other
configuration can be simplified to this arrangement with various modifications to power
plant ratings. This is further investigated in Chapter 6.
2.3 MAIN APPLICATIONS
Potential applications of direct connected HV dc schemes are in cases when the whole
output of the generating station can be transmitted via HV dc, such as:
Transmission of large power from remote sources of generation, such as hydro and
low grade coal fields, when the new development requires little or no local load.
Variable frequency operation such as pump storage schemes and wind power.
Connection of large generating stations to load centres using back-to-back HVdc
schemes. This practically eliminates the stability and short circuit level problems
associated with equivalent ac interconnections.
2.4 A REVIEW OF HARMONIC ANALYSIS METHODS
Harmonic analysis of the convertor can be divided into two broad categories: time
domain and frequency domain analysis. Although time domain has been developed
primarily for transient studies, it can be used for harmonic analysis when systems are
simulated to reach steady state. On the other hand, frequency domain techniques have
been designed for, rather than being applied to harmonic analysis.
The simplified convertor formulae [Kimbark 1971] cannot take into account the
effects of harmonics and frequency couplings due to generator saliency, thus is inadequate
for harmonic analysis of direct connected units. More detailed convertor models in
the Harmonic Domain, a frame of reference which consists of the fundamental and its
harmonic frequencies [Semlyen et al. 1988], are iterative and can be categorized into
two types based on the solution techniques: sequential and unified solution models. A
2.4 A REVIEW OF HARMONIC ANALYSIS METHODS 9
sequential solution model, the Iterative Harmonic Analysis (IHA) , has been primarily
developed for analysing harmonic interaction between convertor and ac system, in which
harmonic impedances are usually small. When extending IHAs to model direct connected
units, they suffer from convergence problems due to large ac voltage distortion in the
absence of harmonic filters. Convergence can be improved when a unified solution is
used, in which all the non-linearities around the convertor are solved together.
2.4.1 Time Domain Simulations
Time domain or dynamic simulation is the solution of differential equations to deter-
mine circuit voltages and currents at each time step. Electrical circuit simulation pro-
grams based on the trapezoidal integration technique [Dommel 1969] such as PSCAD-
EMTDC [Woodford et at. 1983] and EMTP [Dommel et at. 1980] are widely available,
and they have evolved primarily to investigate the response of the systems to transient
disturbances.
Basic to the EMTP method is the representation of each inductor and capacitor
as a resistor in parallel with a time dependent and a history dependent current source.
A thyristor switch is modelled as a resistor branch, and the switching is accomplished
by changing the resistor value, i.e. large values for non-conduction and small values
during conduction. Consequently, a network of lumped R, Land C components is built
up as an equivalent circuit of resistive branches and current sources. The trapezoidal
rule is used for integrating the ordinary differential equations of lumped inductors and
capacitors. Due to the switching action, the conductance matrix must be updated at
each switching instant. The node voltages are then calculated using nodal analysis, and
the current source values for the next time step are calculated from the history terms.
Harmonic information can be obtained by simulating the desired systems to reach the
steady state, a process that is computationally intensive. Nevertheless, the modular
and general purpose nature of these programs simplify the network construction process
considerably.
In PSCAD2-EMTDC simulations, it is necessary to include an artificial snubber ca-
pacitance across each thyristor to improve numerical stability. The artificial capacitance
used does not represent the actual snubber capacitance in the circuit. Numerical insta-
bility occurs when the rate of change of the voltages in the switching process are large.
However the often required large capacitance value can affect the harmonic results. Sim-
ulations with various values of snubber capacitance were carried out to study the extent
of the harmonic variation in a Unit Connection. Figure 2.4 shows the variation of the
11th and 13th ac current harmonics with respect to snubber capacitance values for a
particular loading condition. It is observed that the magnitudes of the harmonics vary
by up to 12% for various values of the snubber capacitance. An interpolation algorithm
applied to switching devices has been developed which could eliminate the need for using
Generator Harmonic Currents
0.08,,----.,---,----------.-----,---------,
0.07
" .1Jth ..
------------------------
0.06 ........................................................................ , .............. .
..--.
:::l
0.05
ci.
13th
----'-'------
'--'
a.>
"'C
0.04
:::l
+-'
:2
OJ
ro 0.03
:2:
0.02
0.Q1
0
2.5 10 25 50 75 100
Snubber Capacitance (pF)
Figure 2.4 Effects of arbitrary snubber capacitances on the harmonic levels in PSCAD2-EMTDC
simulations.
the artificial capacitance [Gole et ai. 1997]. This interpolation algorithm is going to be
incorporated into the 3.0 future releases of PSCAD-EMTDC.
The duration of time steps also influences the harmonic results. Ideally, switching
instants should fall exactly on one of the solution points. But in order to accomplish
this, the simulation step size may need to be changed, requiring a network reformation.
Alternatively the switching may be assumed to occur at the nearest solution point at
the expense of introducing a small error. It has been found that for small time steps,
I.e. 5p,s or smaller, the effect on harmonics is minimal.
In summary, provided that the snubber capacitance value and the time step size are
carefully selected, harmonic analysis may be performed by running dynamic simulations
to the steady state. Although frequency dependent components can be included as a
combination of R, Land C, the selection of these elements is a difficult and tedious
process.
2.4.2 Harmonic Domain Modelling
2.4.2.1 Simplified Convertor Formulation
The simplified convertor analysis assumes that the dc current is ripple free and the
commutation voltage is sinusoidal and balanced. The generator is simply represented by
a Thevenin equivalent with the impedance based on the subtransient reactance. Under
these ideal conditions, analytic expressions can be written for the transfers across the
convertor [Kimbark 1971].
2.4 A REVIEW OF HARMONIC ANALYSIS METHODS 11
A rigorous investigation by Sankar [1991] concluded that the simplified formulation
is inadequate for harmonic analysis of direct connected units, particularly in the presence
of generator saliency. Nevertheless it does provide a quick estimation of the harmonic
content and can be used as a starting point for an iterative solution.
2.4.2.2 Iterative Harmonic Analysis
As the terminal voltages are distorted in the absence of ac filters, an iterative solution
is necessary. Iterative Harmonic Analysis (IHA) , a general technique for assessing the
harmonic generation of non-linear devices, was first used to investigate the harmonic
interaction between convertor and ac system by Reeve and Baron [1971]. In this simple
algorithm, the convertor and ac system interaction was solved in two steps at each
iteration. Given the terminal voltages as pseudo-commutation voltages, ac currents were
derived from the convertor solution while the direct current was assumed to be flat. The
corresponding ac current harmonics, obtained by using a Fourier analysis, were then
impressed upon the ac system representation to obtain updated terminal voltages. More
detailed convertor models with dc system and control representation were developed
by Eggleston [1985] and Yacamini and de Oliveira [1980] [1986].
Since the convertor, ac and dc system non-linearities are solved sequentially, IHA
algorithms are essentially Gauss-Seidel, or fix point iteration in nature. Although ac-
curate results can be obtained when convergence occurs [Arrillaga et ai. 1987], IHAs
frequently suffer from non-convergence due to the Gauss-Seidel nature of the algorithm.
Yacamini and de Oliveira [1986] claimed that instability of an actual system was
characterized by divergence of the iterative algorithm. Rigorous investigation of IHA
algorithms applied to convertor analysis which was carried out by Callaghan and Ar-
rillaga [1990] has proved that divergence of IHA need not be representative of true
harmonic instability. Sufficient criteria for convergence was also derived analytically.
Non-convergence occurs when the ac system is weak and the commutating reactance is
small, leading to large variation of terminal voltage harmonics in consecutive iterations.
In the case of six-pulse convertors without ac filters, non-convergence of IHA has
been shown to occur when the generator impedance is 0.1 p.u. or more at fundamental
frequency, and the commutating reactance is 0.1 p.u. or less for a formulation containing
the first 50 harmonics. The iterations to convergence for a six-pulse IHA algorithm are
given in Table 2.1 [Callaghan and Arrillaga 1990]. The convergence threshold is very low
since typical values for generator subtransient and transformer leakage reactance are 0.2
and 0.1 p.u. respectively. Thus with no modification, IHA cannot be used to analyse
the harmonic interaction in six-pulse direct connected units.
Callaghan also showed that convergence could be improved by inserting a pair of
equal and opposite reactances, so called inserted reactance pair, between the transformer
primary and generator bus so that their series combination is equivalent to a 'piece of
Xgen Xc Iterations to convergence
0.05 0.15 8
0.07 0.13 14
0.08 0.12 20
0.09 0.11 36
0.10 0.10 Stagnated
0.12 0.08 Stagnated
0.15 0.05 Diverged
0.18 0.02 Converged to nonsense solution
Table 2.1 Iterations to convergence for six-pulse unit connected IHA algorithms; firing angle is 20
degrees and the number of harmonics considered is 50.
Figure 2.5 Matched reactance pair method for six-pulse convertors.
wire', as depicted in Figure 2.5. The positive reactance is then absorbed into the commu-
tating reactance, while the negative reactance is absorbed into the generator impedance.
As a result, the effective commutating reactance is increased and the reactive component
of the generator impedance is decreased without effecting the physical system. Therefore
by adjusting the parameters of the simulation, the same physical situation may be made
to converge or diverge at will.
However this method is not effective for systems consisting of more than one conver-
tor bridge such as twelve-pulse or higher pulse systems, since the harmonic currents are
phase shifted among the bridges. Considering a twelve-pulse convertor with an inserted
reactance pair as shown in Figure 2.6, the new commutation voltages are Vs and VD for
the star-star and star-delta valve group respectively. Although the terminal voltage Vt
is common to both valve groups, Vs and VD take on different values:
Vs
VD
Vt + j
2I
sX
Vt + j2IDX.
(2.1)
(2.2)
By injecting the currents into the generator representation, the terminal voltage is
2.4 A REVIEW OF HARMONIC ANALYSIS METHODS
Figure 2.6 Matched reactance pair method for twelve-pulse convertors.
expressed as:
Vt = -(Is + ID)Zgen + E.
Assuming for simplicity that Zgen = X, equations 2.1 and 2.2 become:
Vs
VD
j(Is - ID)X + E
j(ID - Is)X + E.
13
(2.3)
(2.4)
(2.5)
As a result, the 12n 1 characteristic harmonics are eliminated in the commutation
voltages Vs and VD due to the difference terms (Is- ID) and (ID -Is), but the 5, 7, 17, ... ,
6(2n - 1) 1 harmonics are introduced. Since the distortions due to the 6(2n - 1) 1
harmonics are higher than those from the 12n 1 harmonics, no improvement is gained
by using the inserted reactance pair method for twelve-pulse or higher pulse convertor
systems.
Most recent works have been directed toward improving the solution method itself,
rather than improving the fix point iteration of IRAs. By solving the convertor and the
interaction with the ac and dc system in a unified manner, convergence is improved. The
improvement lies in the fundamentally different solution techniques of the two analyses.
The unified solution is solved by Newton's method, in which all variables are related to
each other via the Jacobian matrix. For example, some variation from the set dc current
level affects all the other variables such that the required value is eventually restored. On
the other hand, IRA convertor variables are solved sequentially with a purely unilateral
relationship between the input and output variables. A large change in those variables
regarded as input could lead to divergence.
2.4.2.3 Method of Norton Equivalents
A variation of the IHA is the method of Norton equivalents. The convergence can be
improved when the convertor is represented in the ac system solution at each iteration
by a Norton equivalent instead of a constant current source. The Norton admittance
represents a linearisation of the convertor response to variation in terminal voltage.
Convertor and generator models have been developed for the Multiphase Harmonic
Load Flow (MHLF) program by Xu et al. [1990][1991][1994]. The harmonic load flows
are obtained from iterations between the Norton equivalent circuits of the nonlinear
elements and the linear network solutions at harmonic frequencies. The convertor model
in MHLF assumes a predetermined firing angle and dc current. Only the contribution of
the commutation current to the phase currents is linearised. The problem with modelling
the convertor and ac system by Norton equivalents is that the convertor is simply an
interface between the ac and dc systems, with only the ac system represented in the
overall solution process. If the convertor controller is modelled, a separate iterative
procedure is required to solve the convertor interaction with the dc system at each
iteration.
Harmonic studies in MHLF so far have only included filtered systems. While in
theory, the algorithm can be extended to model direct connected units, it has not been
attempted.
2.4.2.4 Unified Solution using Newton's Method
The first Newton type solution of the interaction of the convertor with the ac system
is the Harmonic Power Flow developed by Xia and Heydt [1982]. The load flow, har-
monic interaction between nonlinear loads, and firing angle for convertors are linearised
together in a unified Jacobian. This single phase program was extended to three phases
by Valcarcel and Mayordomo [1993]. However the described solution method is a fixed
point iteration of three separate Newton procedures; one each for the load flow, ac
system harmonic interaction, and commutation angles. The direct current is assumed
ripple free, and in common with Xia, the convertor operating point is specified in terms
of real and apparent power, rather than current order or dc power. Moreover the models
exclude the linearisation of an important interaction between the convertor and the load
flow, which may not be solvable by a fixed point iteration.
A full linearisation of the convertor requires that variables other than electrical
quantities, such as switching angles and control variables, are linearised in the same
Jacobian matrix. Since the variables are a mixture of real and complex quantities,
the solution is formulated in positive harmonics and real variables only. A system of
mismatch equations describing the interaction among the convertor, ac and dc system
is solved in a unified algorithm developed by Smith et al. [1995]. The harmonic transfer
2.5 CONCLUSIONS 15
across the convertor is modelled using convolution analysis, leading to a reduced set
of variables. The Jacobian is an estimation of the behaviour of the nonlinear system
in response to small perturbations, thus it gives insights into the transfer of distortions
around the convertor. The solution has been accelerated by employing sparsity methods,
and the convergence has been found to be robust and fast. However harmonic models
of direct connected units have not been developed.
2.5 CONCLUSIONS
In the absence of filters, detailed harmonic analysis is required to predict and evaluate the
strong interaction that exists between the generator and convertor. Such interaction can
be analysed using electromagnetic transient programs on reaching steady state. However
this is a very inefficient approach, considering the long time constants of the synchronous
generator. Moreover the large snubber capacitance value across each thyristor often
required to improve numerical stability can introduce a large error in the ac current
harmonics.
Although the simplified convertor formulae in the Harmonic Domain give direct
harmonic solution, they have been found to be inadequate for harmonic analysis of
direct connected units. Iterative Harmonic Analysis of the harmonic interaction between
convertor and an ac system has been found to give accurate results when convergence is
obtained. However due to the Gauss-Seidel or fix point iteration nature of the algorithms,
IHAs suffer from instability when being used to model the strong harmonic interaction
in direct connected units. This is caused by large ac harmonic voltage distortion due
to large generator harmonic impedances and the absence of harmonic filters. IHAs'
convergence can be improved in six-pulse convertor systems by inserting a matched
reactance pair between the generator terminal and the convertor transformer. However
this method is not applicable for twelve-pulse convertors since the harmonic currents are
phase shifted among the bridges due to the transformer connection.
Alternatively, convergence is improved when all the interactions around the conver-
tor are formulated and solved in a unified algorithm, as implemented for direct connected
units in Chapters 3, 4 and 5.
Chapter 3
GENERATOR MODELLING IN THE HARMONIC
DOMAIN
3.1 INTRODUCTION
A salient-pole generator acts as a harmonic convertor due to saliency and as a voltage
harmonic source due to saturation, generator slot and winding distribution. With the
increased size of modern generating plant and greatly improved synchronous machine
designs, the voltage distortion due to slot and winding distribution have become less
significant and will rarely need to be taken into account in system studies. It is also
assumed that the generator dc excitation current is ripple free. As a result, only accurate
representation of saliency and saturation is required in a generator model.
The harmonic representation of the synchronous machine as an emf in series with
k ~ or the average of ~ and k ~ [Clarke 1950] can not take into account the process
of harmonic conversion. The more advanced generator model in the harmonic domain
was formulated in phase components, and took the rotor saliency and saturation into
account [Medina and Arrillaga 1994]. This model however, did not include the important
effects of rotor angle on saliency and saturation. The transformer saturation formulation
was used to model generator saturation, but because of the difference in the nature of
saturation in generators and transformers, transformer saturation formulation is not
applicable to generator saturation modelling. Section 3.5 investigates the unsuitability
of this saturation model, and examines the generator saturation phenomenon in detail.
This chapter describes the derivation of a generator model in the harmonic domain
that takes into account the effects of rotor angle on the stator-rotor frequency conversion
phenomenon due to saliency and saturation. Since generators convert harmonics from
one sequence to another in the steady state, it is desirable to develop the model in se-
quence components, thus greatly simplifying the formulation and increasing the intuitive
understanding of the model.
The proposed model is validated against a time domain simulation using the
PSCAD2-EMTDC program [Woodford et al. 1983]. From the results of three-phase
harmonic measurements at the Benmore convertor station when being operated as a
group connected unit, described in Chapter 7, the generator impedances at the charac-
18 CHAPTER 3 GENERATOR MODELLING IN THE HARMONIC DOMAIN
teristic harmonics (12n 1) are obtained and compared with those calculated using the
proposed formulation.
3.2 THE PROCESS OF FREQUENCY CONVERSION
In a cylindrical rotor machine, a negative-sequence current of order k injected to the
stator will create a flux of order k + 1 in the rotor, which in turn will induce a negative-
sequence voltage of order k. For the case of a positive-sequence current of order k injected
to the stator, a flux of order k -1 is produced, which in turn produces a positive-sequence
voltage of order k.
In a salient-pole machine, a negative-sequence current of order k produces two
counter rotating fluxes of order k + 1 in the rotor; one inducing a negative-sequence
voltage of order k and the other a positive-sequence voltage of order k + 2 in the stator.
Similarly, a positive-sequence current of order k in the stator produces counter rotating
fluxes of order k - 1, which in turn induce a positive-sequence voltage of order k and
a negative-sequence voltage of order k - 2. The same process is repeated for a current
injection of order k - 2 and of higher harmonic orders in the stator. This is illustrated
in Figure 3.1 for cases where the stator current contains either positive or negative fifth
harmonic content.
6h-
Is = (5) = (5+) + (5_)
/
III
Figure 3.1 Response of a salient pole generator to the presence of a harmonic current.
In general terms, it can be noted that an unbalanced harmonic current of order k
injected in the stator will produce harmonic voltages of orders k - 2, k and k + 2.
3.3 GENERATOR HARMONIC IMPEDANCES IN SEQUENCE COMPONENTS 19
3.3 GENERATOR HARMONIC IMPEDANCES IN SEQUENCE
COMPONENTS
3.3.1 Frame of Reference and Rotor Angle
A positive frequency domain representation of a time domain quantity is equal to the
real or imaginary part of a rotating phasor. The real part yields a cosine referenced
waveform, whereas the imaginary part yields a sine referenced waveform. The cosine
referenced phase a voltage is chosen as the phase angle reference in this analysis.
The effects of saliency can be taken into account by the two-reaction theory proposed
by Blondel [1923] and extended by Doherty and Nickle [1926] and Park [1929]. The
armature current I is resolved into two components: Id in time quadrature with, and Iq
in time phase with the excitation voltage E f.
Under steady state operation, the excitation voltage is
where
I sin( J + ) e
j
(H-7r/2),
I cos (J + ) e
j
(H) ,
(3.1)
(3.2)
(3.3)
and Xd and Xq are the direct and quadrature axis synchronous reactances, and the
variation of air gap inductance is approximated by a second harmonic term.
The corresponding phasors are shown in Figure 3.2 for an unsaturated salient-pole
generator operating with a lagging power factor angle . The rotor angle J can be
expressed as
s; _ J(Xq cos - Ra sin)
tanu - (. ) .
Vi + J Xq sm + Ra cos
d-axis
Id - - - - - - --
I
Ef
jldXd
(3.4)
Figure 3.2 Phasor diagrams of an unsaturated salient-pole synchronous generator operating at a
lagging power factor.
3.3.2 Negative-Sequence Harmonic Impedances
Generator negative-sequence harmonic impedances are derived by injecting negative-
sequence currents into the generator Park's equations and taking the ratios of the ter-
minal voltages and currents at each harmonic. For a harmonic k, the current injections
are
where (3 is an initial angle.
I ej (kwt-(3)
I ej (kwt+27r /3-(3)
I ej (kwt-27r /3-(3)
} [
cos(kwt - (3) j
= I cos(kwt + 27f /3 - (3)
cos(kwt - 27f /3 - (3)
(3.5)
The field circuit is shorted, since it is not excited at harmonic frequencies, i.e.
Vfk = O. (3.6)
The field rotates at synchronous speed with the d-axis leading phase a positive-
sequence fundamental voltage as shown in Figure 3.2. The instantaneous position of the
rotor is
o = wt + 7f /2 + fJ .
Park's transformation from three-phase to dq coordinates is
[
::: j = ~ [ ~ ~ ~ e
~ O k 2"
cos(O - 27f/3) cos(O + 27f/3)
- sin(O - 27f/3) - sin(O + 27f/3)
1
2"
1
2"
(3.7)
(3.8)
Substituting the injected harmonic currents expressed in equation 3.5 into equa-
tion 3.8 and simplifying yields:
[
idk j [COS[(k + l)wt - (3 + 7f/2 + fJ)] j
~ = I - sin[(k + l)wt 0- (3 + 7f /2 + fJ)]
(3.9)
The Park's flux linkages are [Kimbark 1968]:
(3.10)
where ~ and ~ are the direct and quadrature axis subtransient inductances respec-
tively. Using the dq to abc transformation the flux linkage in phase a is
'l/Jak 'l/Jdk cos () - 'l/Jqk sin () + 'l/Jok
1" " 1" " )
2(L
d
+ Lq)I cos(kwt - fJ) + 2(L
d
- Lq)I cos[(k + 2)wt-fJ+7f+2J)] . (3.11
Phase a voltage is then derived:
Vak
(3.12)
Although the applied negative-sequence current is of order k, the armature flux
linkages, and subsequently the armature voltages, consist of harmonics of order k and
k + 2. Calculating phase band c indicates that the voltage harmonic k is of negative-
sequence, and voltage harmonic k + 2 is of positive-sequence. The ratios of voltages and
currents at the same harmonic order are the harmonic self impedances
. d + q
(
X" X")
Z-k,-k = Ra + Jk 2
(3.13)
The generator cross couplings from negative-sequence harmonic currents k to
positive-sequence harmonic voltages k + 2 are
Z = (k + 2) q d e
j
(
2w
t+2H7r/2)
(
X" -X")
+(k+2),-k 2
(3.14)
It is observed that the harmonic impedances are independent of the initial phase fJ
of the applied harmonic currents. The self terms, Z-k,-k, are also independent of the
rotor angle, whereas the cross coupling terms are not.
3.3.3 Positive-Sequence Harmonic Impedances
The generator positive-sequence fundamental impedance can be found by deriving the
excitation voltage in terms of the terminal voltage and load current. Substituting equa-
tion 3.2 and 3.3 into 3.1 yields
(3.15)
In a load flow or harmonic study, the value of the internal emf is not required, thus
Z+1,+1 can be assigned arbitrary.
For harmonic frequencies, injecting positive-sequence harmonic currents to the ar-
mature windings yields:
[
idk 1 [ cos[(k -l)wt - (3 - 7r/2 + 6)] 1
~ = I sin[(k -l)wt (3 - 7r/2 + 6)]
(3.16)
The flux linkage and voltage of phase a are:
'l/Jak = 'l/Jdk cos f) - 'l/Jqk sin f) + 'l/Jok
1" " 1" "
= 2(L
d
+ Lq)I cos(kwt - (3) + 2(L
d
- Lq)I cos[(k - 2)wt - 7r+28-(3)] (3.17)
. d'l/Jak
Vak R a ~ a k + at
R { [Ra + jk (X; ; X;) 1 Ie;(kW'-Pi}
{ (
X" -X") } +R (k - 2) q 2 d Ie
j
[(k-2)wt-,B-20+7f/2] (3.18)
Thus the positive-sequence harmonic parameters are:
(
X" +X")
Ra + jk d 2 q (3.19)
Z-(k-2),+k
(
X" -X") (k - 2) q 2 d e
j
(-2wt-20+7f/2)
(3.20)
The harmonic self impedance terms for positive and negative-sequence are found to
be the same. This is apparent as both sequences would have the same leakage path at
a particular harmonic frequency. Note the phase angle of the cross coupling terms is
now 7r /2 - 26. The cross coupling term is zero for an applied second harmonic positive-
sequence current.
For a different phase angle reference, modifications to the mathematical expressions
for the cross couplings are needed. For instance, if the phase a armature current is
chosen as reference, the instantaneous rotor position becomes
(3.21)
The self terms and the magnitude of the cross coupling terms are unchanged, whereas
the phase angles of the cross coupling terms become
L:Z+(k+2),-k = e
j
(2wt+2H2H7f/2)
L:Z-(k-2),+k = e
j
(-2wt-28-2H7f/2)
3.3.4 Zero-Sequence Harmonic Impedances
(3.22)
(3.23)
There is no cross coupling between the zero-sequence and other sequences in Park's
differential equations. Thus the generator zero-sequence can simply be represented by
an impedance at each harmonic frequency, i.e.
(3.24)
where ROI and X
01
are the fundamental zero-sequence resistance and reactance respec-
tively.
3.3.5 Skin Effect
A constant armature winding resistance value has been assumed so far. However due
to the skin effect the frequency dependence of the resistance accounts for the increased
winding losses with frequency. A skin effect correction factor can be directly incorporated
into the model by changing the resistance value at each harmonic. A commonly accepted
resistance characteristic is [CIGRE Joint Task Force 36.05.02/14.03.03, 1996]:
(3.25)
where Rl is the fundamental frequency resistance. Nevertheless, the contribution of the
armature resistance in the generator impedance remain small at harmonic frequencies as
the generator ratio X" / R is typically 40 at the fundamental, and decreases with higher
harmonic order.
3.4 SYNCHRONOUS GENERATOR HARMONIC EQUIVALENTS
The generator can be represented as a Thevenin or Norton harmonic equivalent, as
shown in Figure 3.3, and the terminal harmonic voltages are thus given as
(3.26)
where E is the excitation voltage and Zg is the generator harmonic impedance matrix.
I
I
or .Yg
Vt
IgE
Figure 3.3 Generator Thevenin and Norton harmonic equivalents.
vt and 1 contain all the relevant voltage and current harmonic phasors in sequence
components, i.e.
vt
1
[V01,01, V-l,-I, V+1,+I, ... , VOk,Ok, V-k,-k, V+k,+kl
[101,01,1_1,-1,1+1,+1, ... , 10k ,Ok, 1-k,-k, 1+k,+kl
(3.27)
(3.28)
For salient-pole generators, the impedance matrix consists of three diagonals as
shown in equation 3.29. The main diagonal contains the self terms and the other diag-
onals contain cross coupling terms between the harmonics.
Zgl,1 Zgl,3
Zg2,2 Zg2,4
Zgen =
Zg3,1 Zg3,3
(3.29)
Zg4,2 Zgk-2,k
Zgk,k-2 Zgk,k
3.4 SYNCHRONOUS GENERATOR HARMONIC EQUIVALENTS 25
Zg or Yg in phase coordinates Zg or Yg in sequence-components
a

0
.
b

c

+
2

2

3

3

4

4

5

5

6

6

a b c 2 3 4 5 6 0 - + 2 3 4 5 6
Harmonic order Harmonic order
Figure 3.4 Structure of the generator impedance or admittance matrix in phase coordinate and se-
quence components.
The elements of Zg are themselves sparse three-by-three matrices:
l
ZOk,Ok
0 0
j
Zgk,k 0
Z-k,-k
0
0 0
Z+k,+k
(3.30)
l
0 0
j
Zgk,k+2
0 0
0 0
(3.31 )
[
0 0 0
j
Zgk,k-2
0 0 0
0
Z+k,-(k-2)
0
(3.32)
where the harmonic impedances are given in equations 3.13, 3.14, 3.19, 3.20 and 3.24.
Zg is always invertible, thus Yg can be obtained by inverting Zg.
Figure 3.4 illustrates the structure of an unsaturated generator impedance or ad-
mittance matrix in both phase coordinate and sequence components form. It is clear
that the sequence components formulation provides more sparsity, leading to reduced
storage space and increased speed in system studies. More significantly, the sequence
components formulation is defined by simple analytical expressions obtained directly
from normally available machine parameters, whereas a transformation involving com-
plex numbers is required in obtaining phase component formulation.
The Norton shunt admittance will have a different value for each loading as the cross
I
Figure 3.5 Modified equivalent circuit.
couplings are dependent on rotor angle. However a constant admittance matrix can be
used, provided that the injected current is suitably modified to retain the accuracy of
the Norton equivalent. The constant admittance, Y
gO
, consists of only the main diagonal
(self impedance terms) and the saliency is taken into account by current injections, as
shown in Figure 3.5.
The armature current is:
1= Yg(E V) = Ygo(E - V) + Isal (3.33)
and hence
Isal = (Yg - Ygo)(E - V) = Ysal(E - V) (3.34)
where Isal is the current injection due to saliency and Ysal consists of the cross coupling
terms only.
U sing this approach, it is not necessary to update the Yg matrix at each step of
an iterative algorithm. Instead, changes in the rotor angle and consequently in saliency
terms are taken into account by the current injections.
3.5 SATURATION EFFECTS ON STEADY STATE OPERATION
The effects of magnetic saturation on the generator steady state behaviour is investigated
in this section. As the flux across the air gap increases, the core and teeth of the rotor and
stator saturate. Subsequently, the reluctance of the magnetic circuit increases, leading
to a decrease in the generator reactances, and thus an increase in the dc excitation level.
Based on the assumption that saturation is a function of the total air-gap flux only,
there are essentially three main effects on the generator steady state behaviour:
An increase in dc excitation level
Decreases in generator reactances lead to changes in rotor angle, and thus changes
in generator harmonic cross couplings.
3.5 SATURATION EFFECTS ON STEADY STATE OPERATION 27
The production of harmonics due to non-linear magnet ising characteristic.
In a steady state system study, the excitation requirement is not required to be
included in the formulation. The calculation of the generator dc excitation is readily
available in literature, where the excitation can be calculated from the terminal voltages
and currents obtained from the system study. The second and third effects are to be
examined in the following sections.
3.5.1 Saturation Effects on Generator Harmonic Impedances
As the subtransient reactances ~ and ~ of a synchronous machine with damper
windings are associated with fast transient flux that flows mostly in air, they remain
constant for a wide range of terminal voltage under saturation condition [Kilgore 1935].
Consequently, the self impedances at harmonic frequencies are unchanged according to
equation 3.13 and 3.19.
With saturation, the rotor angle is decreased [Kilgore 1935], regardless of motor or
generator action, or of power factor, leading to changes in the harmonic cross coupling
terms for all harmonics according to equation 3.14 and 3.20. The calculation of rotor
angle from the terminal voltage, load current and generator magnetic characteristics
has been developed using various techniques such as Potier reactance [Kimbark 1968],
saturation factors for d- and q-axis [Harley et al. 1980], etc.
The Potier reactance technique uses a saturation correction voltage E
s
, obtained
from the open-circuit saturation curve and the Potier voltage E
p
, to determine the
excitation voltage Ef. Ef indicates the position of the q-axis under saturation, and the
adjusted rotor angle is readily obtained since the terminal voltage is known. Although
it is assumed that saturation is the same on the d-, q-axis and between the axes, the
level of saturation will vary in practice. However this is a reasonable assumption in the
context of the steady state rotor angle calculation for large salient-pole generators with
damper windings, where the error obtained would be small. The significant advantage of
this technique is that the Potier reactance and open-circuit saturation curve are readily
available.
Calculation of rotor angle using Potier Reactance
Given terminal voltage lit, armature current I, load angle , Potier reactance Xp
and the open-circuit saturation curve, the rotor angle with saturation taken into account
can be calculated as follow:
Determine rotor angle 8 using unsaturated synchronous reactances, and E
q
, which
is the unsaturated Ef, as previously described in Section 3.3.1.

C<J
,::l

.
E %

-.:t:: E ............................................ .
p open-circuit
saturation clln'e
Field current
Figure 3.6 Graphical construction for finding the components of armature voltage due to saturation
E8 corresponding to Potier voltage Ep.
Find the Potier voltage as a vector sum of the terminal voltage and the voltage
drop across the Potier reactance, i.e. Ep = Vt + j XpI
Calculate the magnitude of the saturation correction voltage Es using the open-
circuit saturation curve and the Potier voltage E
p
, as illustrated in Figure 3.6. In
order to eliminate the manual calculation of E
s
, the saturation curve is represented
by a mathematical expression that can be obtained by using the cubic spline curve
fitting technique, summarized in Appendix C. Thus Es is directly expressed as a
function of Ep.
The saturated Ef is the sum of phasors Eq and E
s
, i.e. Ef = Eq + E
s
, where Es
is in phase with Ep.
The saturated rotor angle 5
sat
is obtained by subtracting the unsaturated 5 with
a correction angle i:J.5, where i:J.5 is obtained from the phasors relationship in
Figure 3.7, as to be described below.
With reference to Figure 3.7, the angle f3 can be found from:
tanf3 =
I(Xq - Xp) cos(5 + )
(3.35)
Vt cos 5 + IRcos(5 + ) + IXp sin(5 + )
............... /3 ................... ! ... !
, .... ]
~
1\\\\
....
\
Figure 3.1 Determination of rotor angle under saturation for salient-pole synchronous generators
operating at a lagging power factor.
The variation of rotor angle 1:::..8 can be obtained from Es and f3:
. A J: -1 E s sin f3
tiU = tan
Eq + Es cosf3
(3.36)
The adjusted rotor angle 8
sat
is therefore:
8
sat
= 8 - 1:::..8, (3.37)
where 8 and 1:::..8 are given in equation 3.4 and 3.36.
This completes the calculation of rotor angle under saturation using the Potier
reactance and the open-circuit (i.e. d-axis) saturation curve.
In general, regardless of calculation methods, rotor angle can be directly expressed as
a function of the terminal voltage, load current and generator magnetic characteristics,
f(>"), i.e.
8
sat
= f(vt, I, , f(>..)) (3.38)
and hence the rotor angle is readily incorporated into equation 3.14 and 3.20.
On the other hand, any rotor angle calculation technique using saturation factors
for d- and q-axis requires saturation curves for both d- and q-axis. However the q-axis
saturation characteristic is not often available, and it is also necessary to make an as-
sumption that the saturation characteristic can be interpolated between the axes. Since
superposition of individual magnetic circuit is not valid in the presence of saturation,
it is not clear at this time which of the many methods of interpolation is closest to the
practical test measurements. Nevertheless, generator saturation characteristic is usually
close to the air-gap line, thus changes in rotor angle due to saturation is relatively small.
3.5.2 Generator Saturation Harmonics
There are two kinds of space harmonic flux-waves exist in generators. The space har-
monic, which has been treated extensively in literature, is generated by the harmonic
content of the magnetising mmf wave and by the slot openings. The other type of
space harmonic is the result of magnetic saturation in the machine. The first space
harmonic is normally referred to simply as space harmonic and the second has been
named saturation harmonic. Although all harmonics should be considered together in
the determination of the air-gap flux distribution, saturation and space harmonics can
be examined separately without appreciable error.
The saturable portion of the generator magnetic circuit consists of stator teeth, rotor
teeth, stator core and rotor core. When a sinusoidal magnetising mmf is applied, the
resultant air gap flux wave may be either flat-topped or peaked depending upon whether
the teeth or cores are more saturated. Saturation of stator and rotor teeth causes the
air gap flux wave to be flat-topped, and saturation of the cores makes the wave to be
peaked. The flux in the cores is the space integral ofthe flux in the air gap; that is, the
air gap flux density is proportional to the space derivative of the core flux. If the core
is saturated and the teeth are not, the core flux distribution along the circumference is
a flat-topped wave. Since the derivative of a flat-topped wave is a peaked wave, the air
gap flux distribution is therefore peaked [Lee 1961].
In all practical machines, however, teeth are more saturated than cores. Air gap
flux waves in saturated machines are always flat-topped. Since tooth saturation and
core saturation have different effects on the shape of air gap flux wave, highly saturated
machines do not necessarily have high saturation harmonics. The saturation harmonic
content is high only when the teeth are highly saturated but the cores are not saturated.
As saturation harmonics are produced by the redistribution of the flux density in
the air gap when part of the magnetic circuit saturates, i.e. changing the shape of the
flux waves, the determination of generator saturation harmonics is extremely difficult
and requires a knowledge of the paths dimensions and the magnetic characteristics of
the material at every point in the circuit. The direct and quadrature saturation curves,
which are normally used to calculate the excitation requirement and values of saturated
reactances, cannot be used to determine the harmonics produced by saturation. This
is because these saturation curves only indicate the rms values of the terminal voltages
(and thus also of air gap flux), but not the shape ofthe flux waves.
A couple attempts have been carried out to model saturation harmonics in the syn-
chronous generator by Xu et al. [1991] and Medina and Arrillaga [1994]. However as the
d-, q-axis, or both axes saturation curves are used to derive the harmonic content, these
models will give inaccurate results. Moreover in the latter model, the harmonic currents
produced are assumed to be proportional to a fundamental magnetising current on the
stator (analogous to transformer saturation modelling) which is virtually non-existent
since most, if not all, ofthe magnet ising current is flowing in the rotor electrical circuit.
This is because the dc resistance of the rotor circuit is much lower than the impedance
of the stator circuit at fundamental frequency, and the amount of magnetising current
in each circuit is inversely proportional to the impedance of the circuit. Consequently,
these approaches are unsuitable for generator saturation harmonics modelling.
Although it is not straightforward to model saturation harmonics, it is possible to
estimate them and examine their effects.
Stator harmonic voltages due to saturation
Since the waveform of the air gap flux is repeated pole after pole, saturation har-
monics are odd-order harmonics, and the generator acts as harmonic voltage sources
under saturation. When a flat-topped air gap flux wave rotates in a machine, its shape
remains unchanged because of the symmetry of the magnetic structure. This means
that there is no relative motion among the fundamental and all harmonic flux waves.
In other words, all the harmonic flux waves rotate at the same speed and in the same
direction as the fundamental. The harmonic fluxwaves induce odd harmonic voltages in
the stator winding and harmonic currents on the line if there is a path for them to flow.
Note that there are no harmonic currents on the rotor due to saturation directly since
rotor harmonic currents would produce pulsating mmfs rather than space (or stationary)
harmonic waves, and the rotor is rotating at the same speed as the armature mmf wave.
In the induction motor case, harmonic currents are induced in the rotor due to the slip
speed differential.
Rigorous analysis of saturation harmonics of polyphase induction machines
by Lee [1961] concluded that the saturation harmonic content is low even though its
magnetic circuit is highly saturated, and it is because the saturation of the cores partly
offsets the flattening effect of teeth saturation. Quantitatively, the 3rd, 5th and 7th har-
monic components of a flat-topped flux-wave in the highly saturated 18kVA induction
machine are approximately 6-8%, 4% and 3% of the fundamental respectively [Lee 1961],
[Zagradisnik and Hribernik 1988].
Since an accurate analysis of generator harmonics required machine physical charac-
teristics and in the absence of information on synchronous generator saturation harmon-
ics, the induction motor case is used here as the worst case for synchronous generators
in terms of saturation harmonics. Synchronous generators are usually larger in physical
size, thus normally saturate less than induction motors.
In synchronous generators, the distributed winding and slot pitch factors (kd and
kp respectively), which are used to suppress space harmonics (particularly the 5th and
7th) [Say 1983] [McPherson and Laramore 1990], will also suppress saturation harmonics
since they both are of the same type of harmonics that rotate at the same speed as the
fundamental. Harmonic voltages are therefore only a fraction of the harmonic fluxes.
A typical value of kd and kp for 5th harmonics is 0.259 for a coil pitch of 150
0
, thus
the 5th harmonic voltage is only 6.7% (= kdkp) of the 5th harmonic component of the
flat-topped flux wave, i.e. V5 = 0.0027 p.u. in this case.
The third and triplen voltage harmonics are of zero-sequence, thus regardless of
whether the machine is connected in wye or delta, no third harmonic voltage will appear
at the line terminals. In normal system connections, the stator connection is wye-
grounded and the generators are connected to the transmission system through a delta-
star/grounded transformer, and thus no third harmonic current can flow. Consequently,
the only effect of the triplen voltage harmonics is an increase in the generator insulation
requirement.
For the 5th, 7th, 11th, 13th, etc. harmonic orders, the amount of harmonic current
flowing in the stator due to saturation is dependent on the external network connec-
tions and harmonic impedances of the circuit. They can be obtained from a harmonic
penetration study. However given the small harmonic voltage levels involved, the stator
harmonic currents due to saturation are insignificant. Therefore, the harmonics pro-
duced by generator saturation can be ignored in synchronous generator steady state
studies.
3.6 MODEL VALIDATION
The proposed formulation for the generator harmonic impedances is validated against a
time domain simulation for the test system of Figure 3.8 using the PSCAD2-EMTDC
program [Woodford et al. 1983]. The generator parameters, based on Hwang [1969] and
saturation characteristics are listed in Appendix A. The salient-pole generator in the
PSCAD2 program takes into account the direct-axis saturation, but does not include
the cross saturation effects where the direct-axis saturation changes quadrature-axis
parameters. As a result, the quadrature-axis synchronous reactance and thus the rotor
angle are unchanged under saturation. In this case, equation 3.38 can be used to calculate
rotor angle under saturated condition. Due to the insignificant effect of saturation
harmonics, they are not modelled in the PSCAD2 program, thus are neglected in the
validation process.
10% of negative-sequence harmonic currents of fundamental, 5th and 9th order were
injected simultaneously to the generator bus. Due to the generator saliency, positive-
sequence 3rd, 7th and 11th voltages are produced. Having run the simulation to steady
state, an FFT was performed over one fundamental cycle of the voltage and current
waveforms to obtained their harmonic contents, and the results were compared with
those calculated in the harmonic domain.
The time domain solution serves to verify the harmonic analysis if upon comparison
a close agreement is found in the terminal voltages. This is clearly the case, as illustrated
3.7 LEVELS OF HARMONIC IMPEDANCES 33
0.1 pu harmonic
current injections
PSCAD-
Ia
20QVv--
EMTDC
Ib 20Qyv--
Generator
Ie
20QA./'v-
model
---
Va Vb Vc
Figure 3.8 Test system for model validation.
by the results in Figure 3.9. Close agreement of phase angles as well as magnitudes indi-
cates that the generator impedances, including the saliency effect, have been accurately
modelled. It was found necessary to use a time step of 5/-ls, and to simulate for 700
fundamental frequency cycles to allow transients to decay. Therefore the analytical gen-
erator model in the harmonic domain is very efficient computationally in comparison
with the time domain for analysing harmonic behaviour.
One of the objectives of the harmonic measurements carried out at the Benmore
convertor station in 1995 when being operated as a group connected unit was to pro-
vide information for possible validation of the generator harmonic model. Three-phase
harmonic measurements of a generator currents and its terminal voltages were obtained.
From measured harmonic voltages and currents, generator harmonic impedances at the
characteristic harmonics 12n 1 such as self terms and cross-coupling terms are ob-
tained by using the developed saliency formulation. The derivation of the harmonic
impedances in the presence of saliency from the measured harmonic voltages and cur-
rents is described in Appendix B. The results are shown in Figure 3.10 together, for
comparison, with those derived from the model using the given generator parameters,
listed in Appendix A2. At the 11th and 13th harmonic orders which have the largest
harmonic voltages and currents (thus the data is most reliable), close agreement has
been obtained. The harmonic impedance trends are also consistent, particularly the
cross-coupling terms due to saliency. Thus the harmonic measurements also serve to
validate the developed generator saliency formulation.
3.7 LEVELS OF HARMONIC IMPEDANCES
An indication of the relative importance of the self and cross coupling harmonic im-
pedances is given in Figure 3.11 for a test machine, using the parameters given in
Appendix Al. The rotor angle is 2 for the circuit shown in Figure 3.8. Since the
34
0.4
........
:J
0.3
ci.
'-"
Q)
"0
0.2
:J
.....
c
Ol
0.1
co
2
0
200
(j)
Q)
100

Ol
Q)

0
Q)
(f)
co
-100
..c
D....
-200
0.2
........
:J
0.15
0..
'-"
Q)
"0
0.1
:J
.....
'c
OJ
0.05
co
2
0
200
(j)
Q)
100

OJ
Q)
"0
0
'-"
Q)
(f)
co
-100
..c
D....
-200
CHAPTER 3 GENERATOR MODELLING IN THE HARMONIC DOMAIN
rnn
2
2
2
2
Negative-sequence harmonic voltages
0
Harmonic domain
D
EMTDC
I:
3 4 5 6 7 8 9 10 11
Harmonic order
Negative-sequence harmonic voltages
3 4 5 6 7 8 9 10
Harmonic order
Positive-sequence harmonic voltages
51 Harmonic domain
D EMTDC
11
12 13
12 13
3 4 5 6 7 8 9 10 11 12 13
Harmonic order
Positive-sequence harmonic voltages
JIl
blU
3 4 5 6 7 8 9 10 11 12 13
Harmonic order
Figure 3.9 Model validation by comparison of time and harmonic domain solutions for harmonic
current injections.
3.7 LEVELS OF HARMONIC IMPEDANCES 35
Generator harmonic impedances - self terms
1 0 - - - - - - - - - - - - - - - - - - - - ~ - - _ - - - - - - - - _ . - - - - - - - - ~
8 Bars - Measured
Line - Calculated
""":'
:::l
0.. 6
"'-'
Q)
"0
:::l
+-'
'c
0)
4
cu
~
2
0
10 20 30 40 50
Harmonic order
(a)
Generator harmonic impedances - cross couplings
0.6
0.5
Bars - Measured
Lines - Calculated
.--..,
:::l
0.4
0..
"-"
..-
Q)
..-
"0
0.3
..-
:::l
:t::
C
0)
cu ..-
~ 0.2
..-
..-
..-
..-
..-
..-
0.1
..-
..-
..-
..-
0
10 20 30 40 50
Harmonic order
(b)
Figure 3.10 Generator harmonic impedances obtained from the measurement and calculation. (a)
Self harmonic impedance terms, (b) Dashed-dashed line and filled bars are the cross coupling terms from
negative-sequence k to positive-sequence k + 2, and dashed-dotted line and unfilled bars are the cross
coupling terms from positive-sequence k to negative-sequence k - 2.
36
9
8
7
...--- 6
::J
S,
5
Q)
"'0
.2
'c 4
OJ
<tl

3
2
0
91
(j)
90 Q)
Q)
...
OJ
89
Q)
2-
Q) 88
(J)
<tl
..c
a.
87
86
5
5
Generator harmonic impedances
0)
(ii)
--
(iii)
-.-
_.--
10 15 20
Harmonic order
Self terms, negative-sequence
10 15 20
Harmonic order
Cross coupling terms
(i)
25 30
25 30
100,,------,-------,-------,-------,-------,------,
I % __________________________ _______ _
0,
Q)
2- 90
Q)
l@ 85
..c
a.

5 10 15 20 25 30
Harmonic order
Figure 3.11 Generator harmonic impedances for a particular loading, 8 = 2 electrical degrees, where
(i) Self terms - Negative-sequences; (ii) Cross couplings - Negative to positive-sequences; (iii) Cross
couplings - Positive to negative-sequences.
3.7 LEVELS OF HARMONIC IMPEDANCES
Cross couplings between 5th and 7th harmonic

::i 0.8
00
Q) 0.6 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
"0
::l
1 0.4 (iii)
_._._._._.-.-.-.-'-'-'-'-'-._._._.-.-.-._._._.-.-.-.-.-.-.-.
0.2
0
0 10 20 30 40 50 60 70 80
Rotor angle
270
en ---
180 (ii)
....
OJ
Q)
r-.
7
':: -=- = -
90
Q)
- -.
en
ro
0 (iii)
..c
a..
-90
0 10 20 30 40 50 60 70 80
Rotor angle
37
90
90
Figure 3.12 Magnitude and phase angle of cross coupling terms for various loadings, where (ii) Neg-
ative to positive-sequences couplings; (iii) Positive to negative-sequences couplings.
Figure 3.13 Generator-transformer-convertor unit.
magnitudes and phase angles of the harmonic self impedances, apart from the funda-
mental frequency, are the same for both positive and negative-sequence, only the latter is
shown. The contribution of the resistance is negligible in comparison with the reactance
at high harmonic orders. The magnitude of the cross coupling terms due to saliency are
roughly 30% of the self impedances for this test generator.
Under a steady state operating condition, the phase angles of the cross couplings
from one sequence to another are the same for all harmonics. However these will vary
as a function of the rotor angle 8 for various loadings, as illustrated in Figure 3.12. On
the other hand, the magnitude of a particular cross coupling term does not change with
the rotor angle.
The effect of rotor angle on the cross couplings is emphasized in the case of a
direct connection of generators to HV dc convertors. In a simple system consisting of a
38
0.3
--:- 0.25
::l
E-
m 0.2
"0
::l
:t:
,0.15
ro

0.1
0.05
Positive-sequence 7th harmonic voltage
. . . . . . , . . . . ,
...... ' ......... -..... .......... .
Rotor angle
Figure 3.14 Variation of the 7th harmonic voltage in relation to the rotor angle and firing delay angle.
single generator-transformer-convertor unit, shown in Figure 3.13, an approximate direct
solution for harmonic voltages can be obtained by modelling the convertor as a fixed
injection of characteristic harmonic currents, and the dc current is assumed to be ripple
free. Under these conditions, the commutation angle and harmonic currents can be
calculated using the simplified convertor formulation [Arrillaga 1983]. The calculation
is shown in Appendix E and the results are used here.
Because of saliency, each 6k+ 1 positive-sequence voltage is made of two components
due to the corresponding positive-sequence 6k + 1 and negative-sequence 6k -1 currents.
Thus the harmonic voltages are dependent on the phase angle difference between the
currents as well as the rotor angle. Given the fundamental terminal voltage, dc cur-
rent, commutation reactance and firing angle, characteristic harmonic currents can be
calculated.
As an example to illustrate the rotor angle effect on saliency, the 7th harmonic
voltage can be approximated by injecting negative-sequence 5th and positive-sequence
7th currents, calculated using the simplified convertor formulation, into a salient-pole
machine with varying rotor angle. The machine parameters are listed in Appendix
Al. With 1 p.u. terminal voltage, fundamental current and 5% transformer leakage
reactance, the 7th voltage variation is shown in Figure 3.14. The commutating reactance
was estimated as the sum of transformer leakage and the average of the direct and
quadrature axis subtransient reactances. It is clear that the position of the rotor and
the firing delay angle do have significant effects on the harmonic voltages. It is observed
that for rotor angle less than 45, the 7th voltage is reduced with increasing rotor angle.
3.8 CONCLUSIONS 39
Consequently, any harmonic model for saliency that does not take into account the rotor
angle (i.e. zero rotor angle is assumed) will likely to over-estimate the 7th harmonic
voltage levels.
In practice, ac voltage harmonics and dc current ripple do influence harmonic cur-
rents, thus detailed harmonic interaction around a convertor is required to accurately
predict the commutation periods as well as ac and dc terminal harmonic distortions.
More detailed investigation of the saliency effect is carried out in Chapter 5 and 6.
3.8 CONCLUSIONS
A synchronous generator model in the Harmonic Domain has been developed. The model
uses a harmonically coupled impedance matrix which takes into account the effects of
saturation and the stator-rotor frequency conversion phenomenon due to saliency. It
has been found that the harmonic cross couplings vary as a function of the generator
rotor angle, whereas the generator harmonic self impedances, except the fundamental
positive-sequence impedance, are independent of rotor angle. The importance of rotor
angle in the saliency effect has been highlighted in the case of a direct connection of
generators to AC/DC convertors in which harmonic voltages vary not only with the
firing angle but also the rotor angle.
Without saturation effects, the relationship between the generator terminal voltages
and currents in the presence of saliency can be represented by analytical expressions in
sequence components. The expressions are derived from normally available machine
parameters such as direct and quadrature-axis subtransient reactances. As a result, this
analytical model is a powerful tool for a fast and accurate prediction of the harmonic
levels on the generator terminals in the presence of saliency and distortion.
Saturation in the machine air gap flux reduces the rotor angle which leads to changes
in the cross coupling terms for all harmonics. The self impedance terms are not affected
by saturation as they are solely dependent upon the subtransient reactances which are
constant since they are associated with fast transient flux that flows mostly in air.
Magnetic saturation also produces odd order harmonics due to the distortion of the
airgap flux wave. However saturation harmonics are small and can be ignored from
synchronous generator steady state studies.
The variation of generator air gap subtransient inductance has been approximated
by a second harmonic term described by Park's equations, which have also been used
extensively in other computer models. Alternatively, generator impedance can be ob-
tained by using a more rigorous analysis such as Finite Element Analysis. However this
is a computationally intensive and labourious process. The subtransient inductance has
also been assumed to be invariant for harmonic frequencies. Accurate inductance values
at harmonic frequencies can be obtained through test measurements and they can be
readily taken into account at each interested harmonic in the harmonic domain model.
The generator harmonic model has been validated against time domain simulations
and results from harmonic measurements of a group connected unit.
Chapter 4
CONVERTOR ANALYSIS IN THE HARMONIC
DOMAIN
4.1 INTRODUCTION
Essential to the Unit Connection model is the analysis of harmonic transfers across the
convertor. Of all the available convertor models in the Harmonic Domain, reviewed in
Chapter 2, the harmonic transfer analysis most suitable for a unified iterative solution
was developed by Smith et ai. [1995].
This model consists of a general set of non-linear equations that fully describes the
harmonic transfers through the convertor in the steady state. The formulation convolves
periodic sampled quantities in the harmonic domain with their sampling functions, so
that no Fourier transform is required. The sampling functions are defined in terms of the
exact switching instants, which are obtained as part of the overall iterative procedure
that accurately models the effect of ac voltage and dc current distortion on the valve
conduction periods. Distortions and unbalance are taken into account in the analysis of
the switching instants.
As thorough understanding of the convertor is neccessary to the overall Unit Con-
nection, the convertor harmonic domain model is summarized in this chapter.
4.2 COMMUTATION ANALYSIS FOR A UNIFIED ITERATIVE
SOLUTION
In the absence of ac harmonic filters at the generator terminal in the case of direct con-
nected units, the ideal sinusoidal commutation voltage moves inside the generator and
the commutating reactance consists of the transformer leakage reactance and generator
subtransient reactance. However since generator subtransient reactance varies with ro-
tor angle due to saliency, an iterative solution for the commutation process is necessary.
By choosing the generator terminal voltages as pseudo-commutation voltages, which can
be unbalanced or contain harmonic distortion, the effect of the varying generator sub-
transient reactance and transformer leakage reactance can be considered separately in
the iterative solution. Given the generator terminal voltage as a pseudo-commutation
42 CHAPTER 4 CONVERTOR ANALYSIS IN THE HARMONIC DOMAIN
voltage, the convertor harmonic currents can be obtained from the commutation analy-
sis, to be described in Sections 4.2.1 and 4.2.2, in which the reactances are solely the
transformer leakage reactances. The generator terminal voltages are then updated by
injecting the convertor harmonic currents into the generator harmonic impedance as
derived in Chapter 3.
In general, the ac terminal voltages and convertor transformer reactance are no
longer the ideal sinusoidal commutation voltages and commutation reactance in the
presence of distortion on the ac terminal. The commutation reactance consists of the
transformer leakage and a portion of the ac system reactance, which may contain phase-
dependent or frequency-dependent components. In such cases, the following commuta-
tion analysis can also be used, and the ac terminal voltages can be chosen as pseudo-
commutation voltages in the iterative solution so that the reactance in the commutation
analysis consists of solely the transformer leakage reactance.
Two separate analyses of the commutation process are given in this section. One
for a bridge connected to a star/star transformer, and the other for a bridge connected
to a star-g/delta transformer. Comnmtation reactances can be unbalanced between the
phases. The result of this commutation analysis is a set of non-linear equations that
can be solved to find the exact end of commutation angles given distortions on the ac
and dc terminals. The presence of resistance in the commutation circuit complicates the
analysis, however when the interaction of the convertor with generators or ac system is
solved, the effect of commutation resistance can be accounted for by placing it between
the ac system terminal and the convertor transformer primary windings.
4.2.1 Star Connection Analysis
The commutation circuit is shown in Figure 4.1, where Va, Vb, Ie and Id are sums
of harmonic phasors. In this diagram phase a is commutating off, while phase b is
commutating on. The commutation ends when Ie = I
d
.
Summing voltage drops around the commutating current loop at harmonic order k
yields
from which the commutation current can be obtained:
where Vabk = Vak - Vbk
I _ j kXaIdk - Vabk
ek - jk(Xa + X
b
) ,
(4.1)
(4.2)
Using the harmonic coefficients derived in 4.2, the periodic commutation current in
4.2 COMMUTATION ANALYSIS FOR A UNIFIED ITERATIVE SOLUTION 43
A
~
c
B
Figure 4.1 Circuit for star-g/star commutation analysis.
the time domain is expressed as
(4.3)
where
(4.4)
D ensures that the steady state commutation current is zero at the instant of firing,
ei. For incorporation into Newton's method, equation 4.3 is written as a steady state
mismatch equation. The mismatch is the current in the valve that is commutating off,
which at the end of the commutation should be zero. This is obtained by substituting
wi = <Pi into equation 4.3 and taking the difference between the dc and commutation
current. For the ith commutation
(4.5)
IdO and Idk denote dc and harmonic components of the dc current respectively. It
is evident from this analysis that the dc current ripples do affect the duration of the
commutation.
4.2.2 Delta Connection Analysis
The objective is to solve for the commutation current Ie in terms of the voltage and
current sources, as shown in Figure 4.2. Using phasor analysis, the following nodal and
A
V

c
Vd 1
+
ic
ia t
Id
Xc
C B
Figure 4.2 Circuit for star-g/delta commutation analysis.
mesh equations can be obtained for this circuit at each harmonic k:
Idk - Ick + ick - iak 0 (4.6)
Ick + iak - ibk 0 (4.7)
Idk + ick - ibk 0 (4.8)
Vak - jiakXak 0 (4.9)
Ybk - jibkXbk - Vdk =
0 (4.10)
Vck - jickXck + Vdk 0, (4.11)
where Xk = kX
1
. The commutation current IS obtained by combining equa-
tions 4.6, 4.8, 4.9 and 4.10
I _ _ Vak Ybk - Vdk
ck - x + x
J ak J bk
(4.12)
The dc voltage during the commutation is obtained from equations 4.6, 4.7, 4.10
and 4.11
(4.13)
A similar analysis holds for every separate commutation, with appropriate mod-
ifications to the phase subscripts, and the direction of the dc current. The end of
commutation mismatch equation for the star-g/delta connection is obtained as for the
star-g/star connection above.
4.3 CONVERTOR CONTROL AND FIRING PROCESS 45
4.3 CONVERTOR CONTROL AND FIRING PROCESS
4.3.1 Firing Process
Equi-distant firing control is normally employed in modern HV dc scheme. The firing
controller consists of a phase locked oscillator (PLO) tracking the fundamental compo-
nent of the terminal voltage and generating essentially equi-spaced timing references. A
well designed PLO is unaffected by harmonics in the terminal voltage as its time con-
stant should be in the same order as the fundamental. Consequently, the PLO is not
modelled, and the timing pulses are assumed perfectly equidistant, spaced by 60. The
effect of a non-ideal PLO would be to introduce uncharacteristic ac current harmon-
ics [Arrillaga 1983].
4.3.2 Constant Current Control
The result of this analysis is a set of non-linear equations that can be solved to find the
exact firing instant given distortions on the dc terminals.
The controller described here is a PI constant current control, as shown in Figure 4.3,
which will response to harmonics in the dc current.
controller
current transducer
I d
current order
Figure 4.3 Constant current controller.
Alpha
order
The alpha order can be expressed as a sum of harmonic phasors
where
(4.14)
(4.15)
A firing instant occurs when the elapsed angle f3 from a timing pulse is equal to the
instantaneous value of the alpha order:
a + f3i
f3i + ao + I {f aikejkOi} .
k=l
The mismatch equation for the firing instants are therefore:
(4.16)
(4.17)
In the steady state, the dc component of the alpha order, ao, takes on a value that
causes the dc component of the dc current to be equal to the current order. Thus the
constant current control mismatch equation is defined as
(4.18)
4.3.3 Constant Power Control
The constant power controller calculates and passes a current order to the current con-
troller that leads to the rectified power being the power order for the link. The time
constant of the power controller is assumed to be long enough that it is unresponsive to
harmonics in the instantaneous dc power. The power controller is therefore modelled as
Pmeasured - Porder = O.
(4.19)
The measured power is the constant component of the convolution of the measured
dc voltage with the measured dc current, where the measured quantity is a harmonic
weighting of the actual quantity due to the linear frequency response of the measurement
transducer:
and
(4.20)
(4.21)
(4.22)
Normally the harmonics contribute little to the real power, and the measured power
4.4 DIRECT VOLTAGE 47
can be simplified to:
Pmeasured = Vd01dO'
(4.23)
For a 12-pulse convertor, phasor analysis results in 25 simultaneous non-linear mis-
match equations in the same number of variables that fully specify the state of the
convertor as a function of the terminal harmonics. The 25 variables are the 12 firing
instants, the 12 end of commutation instants and the dc component of the alpha order,
ao. Although perfect equidistant firing is assumed, the actual firing instants are not
equi-spaced due to the current controller response to dc ripple. The end of commutation
angles are also unbalanced as a result of ac voltage and dc ripple harmonics.
4.4 DIRECTVOLTAGE
A six pulse bridge has twelve states per cycles. Six of these are commutation states and
six are normal conduction states. During normal conduction the positive and negative
rails of the dc side are connected to the ac side via two conducting thyristors.
Each state can readily be modelled by a linear circuit which has no initial transient.
For instance, during phase a to phase b commutation, the circuit is as shown in Figure 4.l.
The result is 12 harmonic spectra, one for each state and valids for the appropriate
conduction interval only. In order to obtain a single spectrum which is valid for one
complete cycle of dc voltage, the harmonic spectra are convolved with the spectrum of a
periodic square pulse that has value of one during the corresponding conduction interval,
and a value of zero everywhere else. The sum of generated spectra resulted from the
convolution is the spectrum of dc voltage across the convertor.
4.4.1 Star Connection Voltage Samples
During normal conduction the dc voltage is readily obtained as the appropriate phase
to phase voltage minus the voltage drops across the commutating reactances. Thus for
a harmonic k
(4.24)
where p is the conduction interval number.
During a commutation on the positive rail, analysis of Figure 4.1 yields
(4.25)
and
(4.26)
for a commutation on the negative rail, where e refers to phase ending conduction, b to
a phase beginning conduction, and 0 to the other phase.
From the know conduction pattern in each of the twelve states, equations 4.24, 4.25
and 4.26 are used to assemble the twelve samples of the dc voltage. These samples are
summarized in table 4.1.
sample (p) Phase currents DC Voltage Vdp
A B C b e 0
+
- eqn
1 lcl -ld ld - lcl A C B 4.25
2 ld
. -ld
0 A B 4.24
3 ld -lc2 - ld lc2
C B A 4.26
4 ld
0 -ld A C 4.24
5 ld - lc3 lc3 -ld B A C 4.25
6 0 ld -ld B C 4.24
7 lc4 ld -ld - lc4 A C B 4.26
8 -ld ld
0 B A 4.24
9 -ld -lc5 + ld lc5
C B A 4.25
10 -ld 0 ld C A 4.24
11 -ld - lc6 lc6 ld
B A C 4.26
12 0 -ld ld C B 4.24
Table 4.1 Construction of dc voltage and ac phase current samples.
4.4.2 Delta Connection Voltage Samples
The dc voltage during a particular commutation has been derived in section 4.2.2. The
general result is
( 4.27)
where p = 1,3,5,7,9,11. The coefficient matrix P is constant, and need only be calcu-
4.4 DIRECT VOLTAGE
lated once. During a commutation on the positive rail
P
epk
Xeo
Xeo +Xee
Popk
-Xee
=
Xeo + Xee
Pbpk
0
Pdpk
-jkXeoXee
Xeo +Xee
,
49
( 4.28)
(4.29)
(4.30)
(4.31)
where if i E {1 .. 6} is the number of a commutation on the positive rail, p = 2i - 1,
then the subscripts {b, e, o} are permutation of {a, b, c} according to i. A similar result
holds for a commutation on the negative rail.
A
v
c
Vd 1
+
Id
ic
ia t
Xc
B
Ce----I
Figure 4.4 Circuit for a conduction period with a delta connected source.
During a normal conduction period all three phases of the voltage source contribute
to the dc voltage, as illustrated in Figure 4.4 for a particular conduction period. This
circuit is analysed by writing nodal and loop equations at harmonic k:
iak - iek - Idk 0 (4.32)
Idk - iak + ibk 0 (4.33)
iek - ~ k
(4.34)
Vak - jiakXak - Vdk 0 (4.35)
Vbk - jibkXbk + Vdk + Vck - jiekXek 0, (4.36)
then the dc voltage sample at harmonic k is solved:
Vdk = (Vak - jXak1dk) (Xbk + Xek) - Xak(Vbk + Vck).
Xak + X bk + Xek
(4.37)
As for the star connected source, the solution for the dc voltage samples is generalized
over all twelve conduction periods into a matrix of coefficients of the ac and dc sources,
l.e.
(4.38)
where l = 2,4,6,8, 10, 12.
4.4.3 DC Voltage Construction by Convolution
p(wt)
o
2n
Figure 4.5 Sampling functions used in convolution.
The dc voltage spectrum is constructed by convolving each of the twelve voltage samples
(expressed as a function of dc and ac sides sources) with a square pulse sampling function,
as illustrated in Figure 4.5. The square pulse is periodic at the fundamental frequency
and delimited alternatively by the firing and end of commutation angles as listed in
table 4.2. The convolution described here is of positive frequency only, thus generates
phase conjugated terms.
The Fourier coefficients of the sampling function are:
1 127r
ak
- p(wt) cos(kwt)dwt
7r 0
11
bp
- cos(kwt)dwt
7r a
p
1 Ibp
"k sin(kwt)
7r a
p
~ [sin(kb
p
) - sin(ka
p
)], (4.39)
and
1 127r
bk
- p(wt) sin(kwt)dwt
7r 0
1 Ibp
-"k cos(kwt)
7r ap
1
k7r[cos(ka
p
) - cos(kb
p
)] (4.40)
4.4 DIRECT VOLTAGE 51
The complex Euler coefficient for the sampling function at harmonic k is
bk + jak
~ [cos (ka
p
) - cos(kb
p
)] + j ~ [sin(kb
p
) - sin(ka
p
)],
(4.41)
{
jbp:;:p
j [1 - a
p
2
-;b
p
] otherwise
( 4.42)
sample (p) a
p
b
p
1 (h (PI
2 (PI 8
2
3 (h (P2
4
(/;2
8
3
5 8
3 3
6
3
84
7 84
4
8
4
8
5
9 85
5
10
5
8
6
11 8
6 6
12
6
8
1
Table 4.2 Limits of convertor states for use in sampling functions.
Since the end of one conduction interval is the beginning of the next, all of the
trigonometric evaluations are used in two consecutive sampling functions, thus halving
the number of calculations. The dc voltage can now be expressed as
12
Vd = L Vdp 0 \[!p.
p=l
The convolution of two phasors is given by
as described in Appendix D.
(4.43)
(4.44)
The conjugate operator makes the convolution non-analytic, and so not differentiable
in the complex form. It avoids the need for negative harmonics however, and it is still
possible to obtain partial derivatives by decomposing into real and imaginary parts. the
sampling function spectra must be evaluated up to 2nh as is required to calculate voltage
harmonics up to nh. Since the convolution operator is linear, the twelve convolutions in
equation 4.43 can be decomposed into convolutions of the component phasors:
nh 2nh
Vdp 0 W p = L L Vdpk 0 W pl
k=II=O
(4.45)
This equation generates voltage harmonic components of order above nh which are
discarded. By using equations 4.43, 4.44 and 4.45, the kth harmonic phasor component
of Vd is
,k> 0(4.46)
(4.47)
This completes the derivation of the dc voltage harmonics in terms of a dc side
harmonic source and a three phase ac side voltage source connected in star or delta,
with source inductance.
4.5 SECONDARY PHASE CURRENT
The derivation of the dc voltage involves the convolution of twelve different dc voltage
samples, so by using the same sampling functions, 36 convolutions would be required to
obtain the three phase currents. However referring to table 4.1, and using the linearity
of the convolution, the phase a secondary current can be written as:
Ia = Id 0 {W2 + W3 + W4 + W5 - Ws - Wg - WID - Wn}
+Ic1 0 WI - Ic3 0 W5 + Ic4 0 W7 - Ic6 0 wn, (4.48)
and similarly for one of the other two phases. The third phase must always be the
negative sum of the first two, since there is no path for zero sequence into a bridge. This
leads to a total of 10 convolutions to calculate the three phase currents. As evident in
equation 4.48, the periodic samples for the phase current calculation are just the dc side
current, and the commutation currents derived in section 4.2. Thus the secondary phase
current can be expressed as a function of the dc current and its harmonics, terminal
voltage harmonics and the convertor switching angles. The calculation of the phase
current flowing into the transformer primary is addressed in the transformer modelling
section.
4.6 TRANSFORMER MODELLING 53
4.6 TRANSFORMER MODELLING
The convertor transformer is an essential element in a HV dc convertor as it simultane-
ously performs several useful tasks such as:
II Enabling efficient energy transmission by increasing the potential levels .
.. Voltage control and minimization of reactive power consumption by means of tap
changer control of the secondary voltage.
ID Cancelation of six-pulse characteristic harmonics in a twelve-pulse unit by phase
shifting through a star or delta connection .
Reduction of characteristic harmonics as transformer leakage reactance lengthens
the commutation period.
However, unbalance in the tap changer setting between two six-pulse groups will lead
to imperfect cancelation of six-pulse harmonics on the ac and dc sides of the convertor.
If the impedance in each phase of a three-phase bank is not equal, the convertor will
generate positive and negative-sequence odd triplen harmonics. The unbalanced star-
g/ delta connected transformer also acts as a sequence transformer, causing the convertor
to both respond to, and generate zero sequence harmonics. The transformer model is
thus required to take into account these non-ideal operation.
The transformer is modelled as a series connection of ideal tap changing transformers
on the primary and secondary sides, a resistance, a leakage reactance, and a star/delta
connection. Although it is proposed that tap changers can be eliminated in direct
connected units, they are included here for generalisation of the model. The resistance
and reactance may be unbalanced, the tap settings are assumed to be the same on
all phases although it can be unbalanced between the valve groups. The tap change
controller is not modelled as it does not respond to harmonics. The magnetising current
injection is approximated by a shunt to ground at the primary terminal. No attempt is
made to model non-linear transformer effects such as core saturation or hysteresis.
The result of this analysis is a transfer model of the transformer that relates the
primary currents to the secondary currents, and the secondary voltages to the primary
voltages. For a star-g/star transformer with off-nominal taps, shown in Figure 4.6, the
transformer and thyristor resistances are referred to an equivalent primary resistance,
Rae:
(4.49)
x
1
Figure 4.6 Equivalent circuit for star-g/star transformer.
The leakage reactance is referred to an equivalent on the secondary side:
Xeq = a ~ X (4.50)
The secondary voltage becomes independent of the current through the transformer
since all impedances have been removed from the transformer, i.e.
(4.51)
and similarly for the phase current:
(4.52)
These equations are repeated for all phases and all harmonics.
In the case of a star-g/delta transformer, two separate analyses are required for
transfers from the star to delta side and vice versa since part of the circulating zero-
sequence current in the delta winding would appear as a positive or negative-sequence
current on the primary side if the transformer is unbalanced.
The voltage transfer from star to delta is primarily concerned with setting up the
delta connected source for the voltage sampling and commutation analysis, as shown in
Figure 4.7. The delta connected source is scaled by a factor of V3, i.e.
(4.53)
and the equivalent reactance is:
Xeq = 3 a ~ X (4.54)
4.6 TRANSFORMER MODELLING 55
star-gldelta
x
1
..
star to delta transfer
..
delta to star transfer
Figure 4.7 Equivalent circuit for star-g/ delta transformer.
The transfer of thyristor resistance through the transformer is not affected by the
V3 scaling since it is not connected in delta. Thus the referred ac system resistance is
the same as described in equation 4.49.
In the current transfer from delta to star, the zero-sequence needs to be taken into
account separately. The admittance matrix for an unbalanced star-g/ delta transformer
is readily obtained:
I pa
a2Ya 0 0 -a(3Ya a(3Ya 0
VPa
I pb 0 a
2
Yb
0 0 -a(3Yb a(3Yb VPb
Ipc 0 0
a2yc
a(3Yc 0 -a(3Yc
VPc
ISa
-a(3Ya 0 a(3Yc. (32(Ya + Y
c
) -(3
2Y
a -(3
2y
c
VSa
ISb
a(3Ya -a(3Yb 0 -(3
2Y
a (32(Ya + Yb) -(3
2Y
b VSb
Isc
0 a(3Yb -a(3Yc -(3
2y
c
_(32Yb
(32(Yb + Y
c
)
VSc
(4.55)
where
Y
1
(4.56)
Rc+jX
1
(4.57) a
al
(3
1
(4.58) =
V3
a
2
Equation 4.55 is used to calculate the primary current, by assuming that Vp and
Is are known, and eliminating Vs. The admittance matrix in equation 4.55 is not
invertible, as the delta winding is floating. Thus there are infinite number of possible
potentials of the delta winding which are consistent with a given current injection into
the transformer. One such potential is that obtained by grounding phase c on the
secondary so thq,t VSc = O. This permits the removal of the last row and column from
equation 4.55, i.e.
Ipa
VPa
Ipb
1
VPb
Ipc VPc
(4.59)
Isa VSa
ISb VSb
where
[ a?r"
0
1
A
a
2
Yb
(4.60)
0
1
B = o -a(3Yb, (4.61)
a(3Yc 0
c
[
-a(3Ya 0 a(3Yc 1
a(3Ya -a(3Yb 0 '
(4.62)
D -
[
(32(Ya + Yc) -(3
2Y
a 1
-(3
2Y
a (32(Ya + Yb) .
(4.63)
4.7 SUMMARY
Eliminating Vs, the primary phase currents are:
Ip [A - BD-1C1Vp + BD-
1
Is,
def
YDVp +TDIs.
57
(4.64)
(4.65)
YD is a shunt admittance to ground at the convertor terminal that is added to the
filter shunt. TD is a transfer matrix across the transformer, of size 3 x 2, indicating
that there is no zero-sequence current on the secondary side, and that only the phase
a and b currents need to be calculated. If the transformer is balanced, the YD is a
zero-sequence shunt, and so is not invertible. This also implies that if the transformer
is nearly balanced, YD has a high condition number, and should not be inverted into
an impedance without first being combined with an admittance that offers a path for
positive and negative-sequence currents.
4.7 SUMMARY
Harmonic transfers around a twelve-pulse convertor unit in the presence of distortion and
unbalance on the ac and dc terminal have been described. In the presence of distortion
on the ac terminal, the ac terminal voltages and convertor transformer reactance are
no longer the ideal sinusoidal commutation voltages and commutation reactance. The
commutation reactance consists of the transformer leakage and a portion of the ac system
reactance, which could contain phase-dependent or frequency-dependent components.
As a result, an iterative solution for the commutation process is necessary. By choosing
the ac terminal voltage as a pseudo-commutation voltage, which can be unbalanced
or contain harmonic distortion, the effect of the ac system reactance and transformer
leakage reactance can be considered separately in the iterative solution. Given the ac
terminal voltage as a pseudo-commutation voltage, the convertor harmonic currents
can be obtained from the described commutation analysis, in which the reactances are
solely the transformer leakage reactances. The ac terminal voltage harmonics are then
updated by injecting the convertor currents into the ac system harmonic impedance.
Transformer transfer models for star-g/star and star-g/delta connection as well as the
convertor constant current, constant power control, the effects of the distortions on the
firing process have also been taken into account. Convolution in the harmonic domain
is employed to construct the dc voltage and ac currents in terms of the switching angles,
ac voltage harmonics and dc current harmonics. Thus the analysis is completely in the
harmonic domain.
The convertor model is formulated in terms of a set of non-linear mismatch equations
suitable for a unified iterative solution using Newton's method. These equations are
incorporated with the generator equations and solved simultaneously, as described in
the next chapter, for individual convertor switching angles and the interaction between
the convertor with both the generators and dc system.
Chapter 5
A HARMONIC DOMAIN UNIT CONNECTION
MODEL
5.1 INTRODUCTION
In this chapter, the generator model developed in Chapter 3 combined with the convertor
model described in Chapter 4, is solved in a unified algorithm to analyse the harmonic
interaction between the generators, convertors and the dc system.
A minimal set of variables is selected to fully represents the system in the steady
state. The variables are generator terminal voltage harmonics, dc current harmonics,
firing instants, end of commutation instants and a control variable. A set of equations
specifying the interaction between generator and convertor are developed and expressed
in terms of these variables using functional description in both phase and sequence
components, as detailed in section 5.6.
In the absence of the convertor transformer on-load tap-changers, the reactive power
control by minimizing firing angle has been proposed to be maintained by varying the
generator excitation [CIGRE JWG 14/11-09, 1997]. The use of sequence components
greatly simplifies the inclusion of this control action in the formulation as generator
excitation only controls positive-sequence fundamental voltage. Moreover the generator-
convertor harmonic interaction is best explained in this frame of reference as both the
generator and convertor convert harmonics from one sequence to another in the steady
state.
As there is a mixture of real (eg. firing instants) and complex variables, the variables
and equations must be decomposed into real quantities. The set of non-linear equa-
tions is solved by using Newton's method and a Jacobian matrix of partial derivatives,
whose sparse nature is exploited to accelerate the solution. Initial solution estimates
are obtained from the simplified convertor formulation [Arrillaga 1983], and convergence
properties as well as the performance of the unified algorithm are examined, especially
for difficult systems with high levels of distortion. Convergence threshold for the unified
solution method is also compared with IRA algorithms. Finally, the Unit Connection
formulation is validated against time domain simulations using the PSCAD2-EMTDC
program [Woodford et al. 1983].
60 CHAPTER 5 A HARMONIC DOMAIN UNIT CONNECTION MODEL
5.2 SYSTEM VARIABLES
Id Xd Rd
\j
+
E
Zg
Ig
V
D
YgD
P
L{
measured dc current alpha order
Figure 5.1 Twelve-pulse Unit Connection.
The basic Unit Connection to be solved, shown in Figure 5.1, consists of a twelve-pulse
controlled rectifier, and Thevenin equivalents for the generator and dc system. The
generator's harmonically coupled impedance matrix is that derived in Section 3.4. The
rectifier is under constant current or constant power control, and the phase a terminal
voltage is chosen as the phase angle reference.
The transformer and thyristor resistances are lumped into an equivalent resistance,
R
t
, and placed between the generator bus and transformers' primary. To eliminate the
resistive element from the commutation circuit, the voltages on the primary side of the
transformers V; and v/l are used as commutation voltages. With this arrangement, one
set of commutation voltage is required for each six-pulse bridge, doubling the number
of variables. However by relating the primary voltages directly to the generator bus
voltage through the voltage drops across the equivalent resistance, i.e.
(5.1)
the generator terminal voltages and their harmonics can be chosen as variables.
Figure 5.2 illustrates the inter-dependency amongst the system variables. The con-
vertor operation in the steady state can be completely defined in terms of the ac terminal
voltages, dc current harmonics and the switching angles. The ac currents and dc voltage
harmonics are expressed in terms of the chosen variables using convolutions as described
5.2 SYSTEM VARIABLES
Generator Terminal
Voltages + harmonics
Generator Harmonic
Impedance
AC Currents
+ harmonics
Convolution
Analysis
Convertor Switching
Angles
Convolution
Analysis
Figure 5.2 Dependency among system variables.
DC Voltage
+ harmonics
DC system
impedance
DC current
+ harmonics
61
in Chapter 4. The solution method uses estimates of the variables to calculate those
same quantities. The difference between the estimated and calculated values is zero or
smaller than a preset tolerance at the solution.
For an analysis of a 12 pulse convertor up to the nth harmonic, there are 8n + 25
variables:
3 x 2 x n ac voltage harmonics (Three-phase real and imaginary parts).
2 x n de current harmonics (Real and imaginary parts).
12 firing instants.
12 end of commutation instants.
A control variable, for instance, constant current control.
A set of equations is to be developed to fully specify the generator-convertor inter-
action, and the interaction between generator-convertor unit and the de system in the
steady state. The effects of the distortions on the firing and commutation process are
represented implicitly in the equations used for the calculation of the switching angles.
5.3 GENERATOR-CONVERTOR INTERACTION
5.3.1 Fundamental Frequency
In conventional power systems, the AVR action varies the excitation to keep the positive-
sequence fundamental terminal voltage at a set value. Thus in fundamental frequency
load flow studies, the generator internal positive-sequence emf E is not required to be
calculated. The positive-sequence terminal voltage can be specified, which makes it a
slack bus ("V, e" bus where e is the voltage angle) in the load flow.
Alternatively, when the firing angle instead of the voltage magnitude is specified,
the generator bus becomes a "e, a" bus. This implies that both the excitation and
fundamental positive-sequence voltage magnitude are not specified, but they will take
on values that satisfy the firing angle setting, as further explained in section 5.5.
Since the machine excitation acts symmetrically on the three phases, positive-
sequence voltage only is present at the internal busbar. Consequently, the fundamental
negative- and zero-sequence voltages are not influenced by the excitation or positive-
sequence impedance. They behave similarly to the harmonic voltages and can be mod-
elled in the same manner as these in the next section.
5.3.2 Harmonic Frequencies
The convertor harmonic currents flowing into the ac system will produce harmonic volt-
ages on the convertor terminal. The harmonic voltages in turn influence the commutation
and firing process. Given the transformer primary phase currents, the terminal voltage
can be found as
where the square brackets denote a three-phase quantity. The third and fourth terms on
the right hand side of equation 5.2 are due to generator saliency which causes cross cou-
plings between harmonics. Equation 5.2 is represented for all harmonics and expressed
in the mismatch form suitable for Newton's solution as
Since the generator de excitation is assumed ripple free, the harmonic component
of E is zero. Note that E is different from the ideal sinusoidal commutation voltage E",
which relates to the subtransient response of generators.
5.4 CONVERTOR INTERACTION WITH DC SYSTEM 63
5.4 CONVERTOR INTERACTION WITH DC SYSTEM
The mismatch equations derived so far are functions of the dc current harmonics, which
are required to be solved in the overall solution. The harmonic components of the dc
ripple can be found by solving the equations describing the interaction between the
convertor and the dc system. Impressing the calculated dc voltage upon the dc system
admittance leads to
where
5.5 REACTIVE POWER AND VOLTAGE CONTROL BY
GENERATOR EXCITATION
5.5.1 Firing Angle Referenced to the Generator Terminal Voltage
(5.4)
(5.5)
In conventional HV dc schemes, the tap changers on the convertor transformers are used
to control the reactive power by adjusting the taps to keep the firing angle, a, within a
limit while the ac terminal voltage is kept at a constant nominal level.
In the Unit Connection case, the terminaJ ac voltage magnitude becomes irrelevant
due to the absence of local loads. The generator excitation control could be used to
take over the function of on-load tap changers to keep the firing angle, a, within the
predetermined range. This control action is expressed as
12
def 1
Faset = aset - 12 L)Oi - Xi) = 0,
i=1
(5.6)
where aset is the desired firing angle referenced to the generator fundamental terminal
voltage, Oi and Xi are the firing instant and terminal voltage fundamental crossings
respectively. In order to minimize the reactive power consumption, the firing angle
would be set to the minimum allowable value and the generator excitation varies the
internal emf to give the desired dc line voltage, which is set by the invertor control.
In a frame of reference that uses real and imaginary components, it is necessary
to specify the angle of the fundamental positive-sequence terminal voltage which, being
used as a reference, is zero, i.e.
(5.7)
where subscript VI ps denotes the fundamental positive-sequence voltage.
The firing angle can be referenced to the generator terminal voltage, voltage behind
subtransient reactance E" or any other voltages in the unified iterative solution. As firing
angles are normally referenced to a sinusoidal commutation voltage, the formulation of
firing angles referenced to E" is to be described in the next section.
5.5.2 Firing Angle Referenced to the Generator Internal Voltage E"
In the absence of ac filters, the sinusoidal commutation voltage is the voltage E" behind
generator subtransient reactance as shown in Figure 5.3. Since E" leads the generator
fundamental terminal voltage VI for a lagging power factor of less than 90
0
, the valves
can be fired before the crossings of the terminal voltages provided that they are fired
after the crossings of E" to avoid mis-firing.
d-axis
I
I E

q-axis
Figure 5.3 Phasor diagram of a salient-pole generator in direct connected units.
With reference to Figure 5.3, the phase angle L.E"V between E" and VI can be
calculated as:
L.E"V
{
E" }
0- tan-
1

0- tan "
-1 {VI sino - II cos(o + }
Vi cos 0 + II sin(o + )Xd
(5.8)
The firing angle referenced to E" is the sum of the firing angle referenced to VI and
the phase angle L.E" V, i.e.
O',setE" = O',setVl + L.E"V,
(5.9)
and the E" referenced firing angle can replace the VI referenced firing angle in equa-
tion 5.6.
Since accurate measurements of the generator currents as well as voltages are re-
quired for the control system to calculate E" , the firing control of direct connected units
can be simplified considerably by using the generator terminal voltage as the reference.
5.6 FUNCTIONAL DESCRIPTION OF A UNIT CONNECTION 65
However it is necessary to use E" as the reference for comparison between direct con-
nected units and conventional HV de schemes.
5.6 FUNCTIONAL DESCRIPTION OF A UNIT CONNECTION
5.6.1 Phase Components Formulation.
From the commutation analysis in Section 4.2, the equations for end of commutation
instants can be expressed in terms of the commutation voltage, de current harmonics,
and the corresponding firing instants:
(5.10)
As the convertor current controller is responsive to de current harmonics, the equa-
tion for firing instants is a function of those harmonics:
(5.11)
By means of convolution analysis in the Harmonic Domain, the generator current
and de voltage of a Unit Connection scheme can be expressed in terms of the ac terminal
voltage harmonics, de current harmonics and the switching angles, i.e.
S . D
h(Vp , Vp ,Idk, Oi, i)
f4(Vj, vi, 1dk, Oi, i) (5.12)
The relationship between the commutation and generator terminal voltages de-
scribed by equation 5.1 can be rewritten as:
vj f5(V)
vi
(5.13)
The primary voltages can be dropped from the formulation by combining equa-
tions 5.12 and 5.13:
19 h(f5(V), fdV), 1dk, Oi, i)
- h(V, 1dk, Oi, i)
(5.14)
(5.15)
The derived equation 5.3 for the generator-convertor interaction is combined with
Equation No. of Vars. Function of
Fv
2x3n V - fg(V, I
dk
, Oi, i)
Fld
2n
Id - ho(V,Idk' Oi, i)
Frp 12 in (V, I
dk
, Oi, i)
Fe
12 h(I
dk
, Oi, ao)
Faa
1 h2(V, I
dk
, Oi, i, ao)
Table 5.1 Mismatch equations for a 12 pulse convertor in phase coordinates, where n is the number
of evaluated harmonics.
equation 5.14 to give
(5.16)
Similarly, the convertor-DC system interaction is expressed as
(5.17)
The primary voltages are also eliminated in the end of commutation instant equa-
tions, i.e.
(5.18)
The convertor constant current control equation Faa' a function of the dc voltage
and current, can be expressed as
(5.19)
and the equations in terms of the selected variables are summarized in Table 5.l.
It is difficult to incorporate the voltage control by generator excitation in the phase
component formulation since only the positive-sequence fundamental is controlled. On
the other hand, this control action is easily integrated into the sequence components
formulation to be described in the next section.
5.6.2 Sequence Components Formulation.
It has been necessary to model the convertor in phase coordinates since individual con-
duction intervals, related to switchings between the three phases, must be modelled
explicitly. However when interfacing with generators, the sequence components solution
5.6 FUNCTIONAL DESCRIPTION OF A UNIT CONNECTION 67
Variable No.ofVars. Function of
Fw
2x3n -1
W - T f9(T-
1
W,Idk,Ido, fJi, <Pi)
FId
2n Id - ho(T-
1
W,Idk,Ido, fJ
i
, <Pi)
F
12 fn(T-
1
W, Idk, fJ
i
, <Pi)
Fe
12 h(I
dk
, IdO, (}i, ao)
Fao
1 h2(T-
1
W, I
dk
, IdO, fJi, <Pi)
Faset
1 h3(T-
1
W,(}i)
Table 5.2 Mismatch equations for a 12 pulse convertor in sequence components, where n is the number
of evaluated harmonics. The reactive power control by generator excitation is also included.
is more suitable as only the positive-sequence is controlled by the generator excitation.
This also enables the modelling of the convertor reactive power control by generator
excitation in the absence of the convertor transformer on-load tap-changers. An addi-
tional advantage of sequence components is a substantial improvement in the sparsity
of the Jacobian, since the harmonic cross couplings are from one sequence to another.
Sequence to phase transformation is therefore required so that the equations can be
written in sequence components.
Using the sequence transform matrix, T, sequence components formulation in terms
of phase coordinates is
(5.20)
where a = e
j27r
/3. Setting
(5.21)
as the sequence components voltage yields V = T-1W. The existing phase coordinates
equations can now be written as sequence component mismatches, in terms of the se-
quence components terminal voltage, using the sequence transformation. Table 5.2 sum-
marizes the equations in sequence components. In order to model the reactive power
control by generator excitation, the equation relating to the positive-sequence terminal
voltage is replaced by the control equation F
aset
, in which the desired firing angle is
specified.
If a three phase quantity has been decomposed in real rectangular components, a
-
real components sequence transform matrix, T, can be constructed:
vo
R
1 0 1 0 1 0
VR
a
VJ
0 1 0 1 0 1
VI
a
V! 1 0 Ra
2
-'La
2
Ra -'La
V;R
W=
b
=TV (5.22)
V! 0 1 'La
2
Ra
2
'La Ra
V;I
b
VR
+
1 0 Ra -'La Ra
2
-'La
2
VR
c
VI
.+
0 1 'La Ra 'La
2
Ra
2
VI
c
The chain rule readily yields the modifications that must be made to the phase
coordinates partial derivatives, to yield the sequence components partial derivatives:
[
8FW]
8W
[8:;]
~ i ]
(5.23)
5.7 NEWTON'S SOLUTION
Since the ac voltage and dc current mismatch equations are not differentiable in complex
form, the Newton solution is formulated in terms of real functions and variables. The
variables are assembled into a real vector X as
(5.24)
where Rand 'L denote the real and imaginary components respectively. The mismatch
equations are likewise assembled into a real vector:
F(X) = [R{Fv } ,'L{Fv }, R{FId} , 'L{FId} , Fe, Frp, Fao, Faset]T. (5.25)
Newton's method is an iterative process for finding a solution Xs that causes the
mismatch vector to be zero, i.e.
(5.26)
Given an initial estimate of the solution, X
o
, the iterative process is described
as [Burden et at. 1981]
69
(5.27)
(5.28)
with convergence deemed to have occurred when some norm of the residual vector F(XN)
is less than a preset tolerance, and J is the Jacobian matrix of partial derivatives of the
equation set F(X). The convergence tolerances selected are listed below:
IFVkl
<
0.001
IVkl
1F1dk I
<
0.001
IIdkl
IFBil <
5e - 8
IFil <
5e - 8 (5.29)
As the convergence tolerance for the complex mismatches, Fv, FI d is expressed in
terms of the magnitude of the mismatch, the error in an estimated value for a variable,
for instant V+u, is smaller than 0.1% of its own magnitude. Thus all harmonics are
treated equally. This convergence test is only applied to harmonics that have a size
larger than 1e - 5 per unit in order to prevent any attempt to converge to an absolute
error of zero for harmonics that are not present.
5.7.1 Jacobian of Partial Derivatives
By linearization around an operating point using the Jacobian of partial derivatives,
Newton's method effectively reduces a set of non-linear equations to linear equations,
which are solved to improve the accuracy of the solution estimates. Equation 5.28
represents a set of linear equations that must be solved at each iteration. This represents
the majority of the processing required to obtain a solution. By introducing sparsity
into the Jacobian matrix the Newton's solution can be accelerated.
If a mismatch equation is only weakly dependent upon one of the variables, then the
partial derivative with respect to this variable will be small. Thus setting the correspond-
ing Jacobian element to zero still represents an accurate linearization of the equation
in terms of that variable. For a solution up to the fiftieth harmonic, the Jacobian is
typically 91% sparse with elements smaller than 0.005 per unit set to zero. The cut-off
limit of 0.005 per unit was selected to maximize the solution speed while preserving the
robustness of the solution, as demonstrated in Table 5.3. Although more sparsity can
Cut-off Limit Number of Sparsity CPU time Number of
(p.u.) elements (%) (seconds) iterations
0.05 15126 91.6 21.1 21
0.01 16564 90.8 24.3 21
0.007 16753 90.7 21.2 16
0.005 16964 90.6 11.9 7
0.001 18151 90.0 13.1 6
1e-4 22041 87.8 16.9 6
1e-6 40642 77.5 30.2 6
Table 5.3 The selection of the cut-off limit for Jacobian elements in order to maximize solution speed
and preserving the robustness of the solution. Subtransient reactances = 1.0 p.u., transformer leakage
reactance = 0.05 p.u., Idc = 1.0 p.u., 50 harmonics.
be obtained by setting larger elements to zero, the convergence rate is slower and more
iterations are required.
The asymmetric sparse linear solver [Zlatev 1991], y12m, is used to solve equa-
tion 5.28 as the Jacobian is neither symmetric nor diagonally row dominant. Instead of
inverting the Jacobian, it is factorized to LU components and equation 5.28 is solved by
back substitution. The pivoting in the y12m solver is selected to compromise for both
sparsity and numerical stability. The y12m solver computer program was extensively
tested at the University of Copenhagen 1.
Symmetric bifactorization methods, such as by Zollenkopf [1970], require the Jaco-
bian to be diagonally row dominant since it was designed to solve admittance matrix
systems, in which the diagonal elements, i.e. self-admittance terms, are sums of other
terms in the corresponding rows. Consequently, the symmetric bifactorization method
has been found to be unsuitable as the Jacobian is not diagonally row dominant although
it has a large diagonal.
The Jacobian elements can be obtained by using numerical partial differentiation or
by evaluating analytical expressions of the partial derivatives. In the numerical approach,
the elements are approximated by calculating the function with a small perturbation
(L.Xj) to each of the variables sequentially [Gerald and Wheatly 1985], i.e.
(5.30)
where Xo denotes the variables besides Xj. Numerical calculation of the Jacobian is
IFORTRAN codes are available on the World Wide Web at http://www.netlib.org/y12m/
5.7 NEWTON'S SOLUTION 71
.. .. .. .. .. .. .. .. .. ..
Generator saliency
. \ . . ,
. .... iii .. ' .... . :
.... .... ..
......
.... .... : .. ..
.... : .. ..
......
.......... , .....
26 : .......... .
F 52

.. ..
.. '
.. .. ..
.. .. ..
. .... .... ' ...... .
78
104 : .........
F<jl
Fe ' 1
, 4-
.. .. .. .. .. .. .. .. .. .. .. .. .. ..
e do
Figure 5.4 Sparsity structure of the Jacobian in sequence components for 13 harmonics.
a powerful method that is easy to apply, and programs are widely available [Press
et al. 1992]. However it is approximately twenty times slower than the analytical ap-
proach for a formulation of 13 harmonics, since each column of the numerical Jacobian
requires a computation of all the mismatch equations. Although the analytical method
requires considerable effort to obtain the partial derivatives in analytic form, the gain
in solution speed is substantial. Detailed derivation of the analytical Jacobian is given
in Appendix F, and the obtained partial derivatives have been verified against the nu-
merical solution.
For a formulation of 13 harmonics, the Jacobian is 129 elements square, as illustrated
in Figure 5.4. In this example, there is no coupling between the zero and other sequences
as the transformer phase impedance are assumed to be balanced, which is usually the
case in a Unit Connection. Due to generator saliency, the elements representing the
harmonic cross couplings are shifted up by two harmonics from negative-sequence k to
positive-sequence k + 2, and down by two harmonics from positive to negative-sequence,
as indicated in Figure 5.4. Moreover the Jacobian also gives insights to the harmonic
transfers across the convertor through its structural features such as:
The average delay angle mismatch is related to the average dc current, thus it
is sensitive to changes in the fundamental terminal voltage and the harmonics
coupled to the fundamental, such as 11th and 13th ac voltages.
11 The firing instants are sensitive to the dc current, its harmonics, and the constant
current controller parameters.
III The end of commutation mismatches are very sensitive to harmonics in the ac
voltage and dc current. They are also sensitive to the firing instants.
IV A change in harmonic k on the dc side affects positive-sequence harmonic k - 1
and negative-sequence harmonic k + 1 on the ac side, and vice versa for the ac to
dc conversion. The dc current harmonics are sensitive to the ac voltage harmonic
variation as they are obtained by impressing the dc voltage on dc side admittance,
where dc voltage harmonics are essentially ac voltage harmonics transfered through
the convertor. These are called three-port terms by Larson et al. [1989].
v The elements in rows that are parallel and perpendicular to the main diagonals
are the result of the commutation period modulation process [Wood and Arril-
lag a 1995]. Convolutions ofthe dc current harmonics with the sampling functions
produce sum and difference harmonic terms; the sum terms give rows parallel
with the main diagonal, and the difference terms give rows perpendicular to the
diagonal.
5.7.2 Initialization
The simplified convertor analysis [Arrillaga 1983] is employed to calculate the first es-
timation of the system variables. An initial estimate of the convertor delay angle is
obtained using equation 5.31,
2 X 3V2 3X
VdO = IViI cos a - -Ida,
1f 1f
(5.31)
where VI is the phase to phase rms fundamental voltage referred to the secondary of the
convertor transformers. Since there are two six-pulse valve groups, multiplication factor
of 2 is required in the first term on the right hand side of equation 5.31.
read system data
calculate system constants
estimate starting point
using simplified equations
calculate system equations
y
>----;...1 output results
y
N
calculate and bifactorise the Jacobian
update system variables
update secondary variables
Figure 5.5 Flow diagram for the unified harmonic solution.
73
The dc voltage is estimated from the voltage drop through the dc system and the
dc voltage source:
The average commutation angle I'" is obtained from:
3V2
VdO = -IVll[cosa + cos(a + 1"')]
7f
(5.32)
(5.33)
These angles are then used to assemble rough estimates of the individual firing and
end of commutation angles:
(5.34)
(5.35)
Figure 5.5 illustrates the steps of the unified harmonic solution in which the New-
ton's method is implemented in the main loop. The secondary variables are dc voltage
harmonics and ac current harmonics, which are calculated using equations derived in
Section 4.4 and 4.5 respectively.
5.8 MODEL VALIDATION AND PERFORMANCE
5.8.1 Validation
Using the developed formulation, the system of Figure 5.1 is solved and compared with
the results obtained from a time domain simulation using the PSCAD2-EMTDC pro-
gram. The unit connected mode of a Group Connection containing 4 generators has
been selected to provide a high level of harmonic distortion for validating purpose. The
Group Connection implementation is to be described in Section 6.6. The firing angle
referenced to the generator fundamental terminal voltage is 15. The generator and
convertor transformer parameters are listed in Appendix A2.
Despite the high level of distortion, close agreement in both magnitudes and phase
angles has been obtained for the ac voltage, current harmonics and dc voltage as shown
in Figure 5.6, 5.7 and 5.8. It was found necessary to use a time step of 5p,s, and to
simulate for 700 fundamental frequency cycles to allow transients to decay.
It has been found that large values of the snubber capacitance (across thyristors),
which is required to aid numerical stability in the time domain simulations, can give in-
accurate harmonic results. The ac harmonic currents vary by upto 12% for various value
of the snubber capacitance. In order to minimize the effect of the snubber capacitances,
5.8 MODEL VALIDATION AND PERFORMANCE 75
1.2 ...
Generator terminal voltages
. ................ , .......... " ........... , .............................. .
o 45 90 135 180 225 270 315 360
Degrees
Generator phase currents
0.5
.-.
:::J
ci.
--
CD
"0
0
:::J
:!:::
C
0)
cO
2
-0.5
_ 1 L L ~ ~ ~ L L ~ ~
o 45 90 135 180 225 270 315 360
Degrees
Figure 5.6 Model intervalidation by comparison of time domain and unified harmonic solutions for
generator voltage and current waveforms, reconstructed from 50 harmonics. Harmonic domain: solid,
PSGAD: dashed.
76
---
:::J
CL
--
Q)
'0
:::J

C
Ol
CO

---
(J)
Q)
Q)
.....
Ol
Q)
'0
--
Q)
(J)
CHAPTER 5 A HARMONIC DOMAIN UNIT CONNECTION MODEL
0.2
0.1
...............
0
11
Generator voltage harmonics
:EJ Harmonic domain
.................. n ... pyn,ami.c. s.irrMCl,ti.on ............ .
23
Harmonic order
35 47
180nr---------.------------,------------.-----------,,--,
90 ............ .

CO -90 ............ .
..c
a..
--:-
:::J
ci.
--
Q)
'0
:::J

C
Ol
CO

.--.
(J)
Q)
Q)
.....
Ol
Q)
'0
--
Q)
(J)

1 11
0.1
0.05
0
1 11
23
Harmonic order
35
Generator current harmonics
Jill Harmonic domain
47
..................... n ... pyn.ami.c.s.irrju.lCl,ti.on .............. .
ril m
23
Harmonic order
m m
35 47

90
CO -90 .................................... .
..c
a..
.................... ..

1 11 23
Harmonic order
35 47
Figure 5.7 Model intervalidation by comparison of time domain and unified harmonic solutions for
generator harmonic voltages and currents. PSCAD: left; Harmonic domain: right.
5.8 MODEL VALIDATION AND PERFORMANCE
300 ............... .
Q)
-g 200
:!:::
C
0)
ctl
~
150
DC voltage
1 0 0 L L L ~ ~ L L ~ ~
o 45 90 135 180 225 270 315 360
Degrees
77
Figure 5.8 DC voltage waveforms reconstructed from 50 harmonics. Harmonic domain: solid,
PSCAD: dashed.
a small capacitance value of 2pF was used. This introduced high frequency spikes in the
time domain waveforms due to numerical oscillations. However, the harmonic content
closely matched those from the harmonic domain solution. For comparison purpose, the
waveforms were reconstructed from their first 50 harmonics.
5.8.2 Convergence Properties
If the initial estimation is sufficiently close to the final solution, the convergence of the
Newton's method is quadratic [Gerald and Wheatly 1985]. However the convergence is
slower, or in some cases divergence can occur, when the initial estimation is considerably
far from the solution. The convergence of Newton's method is also dependent upon the
gradient of the iterating function F(X). The gradient is given by the magnitudes of the
Jacobian elements. Divergence occurs when the gradient exceeds some "norm" value.
The iterations to convergence for various values of commutation reactance and gen-
erator subtransient reactance is listed in Table 5.4. The formulation consists of the
first 50 harmonics and firing angle is set at 15
0
For typical systems consisting of 0.1
p.u. commutation reactance (which is the transformer leakage reactance in the devel-
oped formulation), the convergence is obtained within 5 iterations for a wide range of
generator subtransient reactance up to 0.5 p.u.. In comparison with sequential IHA
algorithms, the unified solution is much more robust since IHA algorithms diverge when
the commutation reactance is less than 0.1 p.u. and generator subtransient reactances
Generator 'Thansformer leakage reactance (p. u.)
X" (p.u.) 0.02 0.03 0.1
0.0 2 2 2
0.1 5 5 3
0.2 - 5 4
0.3 - 6 5
0.5 - 6 5
Table 5.4 Iterations to convergence for the unified Unit Connection solution using Newton's method.
Hyphen denotes divergence.
are larger than 0.1 p.u., as reviewed in Chapter 2. The improvement lies in the unilateral
relationship of the variables in the unified solution.
For smaller leakage reactance values, the convergence threshold is lower due to a
substantial increase in gradient of the iterating function. Divergence occurs frequently
when the commutation reactance is less than 0.02 p.u .. As time domain simulations of
these systems give valid operating condition, divergence of the unified solution does not
indicate harmonic instability.
The factors that significantly affect the gradient are the commutation reactance and
generator subtransient reactances. Their influence on the Jacobian magnitude can be
examined by considering a term on the Jacobian main diagonal for phase a at harmonic
k. This term can be obtained by differentiating the voltage mismatch, represented by
equation 5.3, with respect to ac voltage variation. Generator saliency is neglected for
simplicity, and the differential is:
(5.36)
The ac current I is made of commutation currents during commutation periods that
involve phase a, and of dc current during normal conduction period. Since dc current
harmonics are chosen as independent variables, they are held constant when differen-
tiating other variables with respect to ac terminal voltages. Thus only contribution of
the commutation current is required to be considered. The commutation current for
commutation period i is given by:
(5.37)
5.8 MODEL VALIDATION AND PERFORMANCE 79
Generator subtransient CPU time Number of
Ac
reactance (p.u.) (seconds) iterations
0.0 1.6 2 1.2 e-5
0.2 6.3 4 2.8 e-3
0.4 6.6 5 5.1 e-3
0.6 8.6 5 1.0 e-2
Table 5.5 Solution speed and convergence factor in relation to the magnitude of generator subtransient
reactances for solutions up to the fiftieth harmonic.
where
D -I {t ICikejkOi} ,
k=l
jkwLc1dk - Vabk
jkwLab
and phase a and b are the phase ending and beginning conduction respectively.
Differentiating Icik with respect to ac voltage yields:
1
jkwL
ab
(5.38)
(5.39)
(5.40)
The firing instants, Oi, are also independent variables, thus are held constant when
differentiating with respect to ac voltage. Consequently, the differential of ac voltage
mismatch with respect to ac voltage can be expressed in terms of generator harmonic
impedance and commutation reactance:
(5.41)
where A is a constant coefficient. The derivation of ~ r ~ } } for all harmonics is described
in Appendix F.
It is apparent from equation 5.41 that the gradient of F(X) is directly proportional
to the generator harmonic impedance and inversely proportional to the commutation
reactance. Hence for large generator subtransient reactance and small commutation
reactance values, the gradient is large which can lead to divergence. However it has
been shown that convergence is fast and robust for typical reactance values and thus
the developed formulation is sufficient for analysing the harmonic interaction in direct
connected units.
A convenient measure of the rates of convergence is the sum of the magnitudes
of all the mismatch equations, which is called i-norm [Gerald and Wheatly 1985]. At
each iteration the i-norm is reduced by an approximately constant convergence factor,
A
c
, depending upon the difficulty of the system, and whether the Jacobian is updated.
The smaller A
c
, the faster the convergence. Table 5.5 shows the solution speed and
convergence factors for various generator subtransient reactance values. The transformer
leakage reactance is 10%. Clearly, the convergence is rapid when there is no voltage
distortion, i.e: with perfect ac filtering (generator harmonic impedance is set to zero).
For typical subtransient reactances of 0.15 to 0.25 per unit, the convergence is fast
and robust, taking only 4 iterations and roughly 7 seconds on an UltraSparc processor.
The solution speed will no doubt be reduced significantly as computer processing power
improves, whereas the robustness of the algorithm, implicitly represented in the small
convergence factor A
c
, is the most significant advantage of the unified solution using
Newton's method.
Although the implemented Newton's method is quite powerful for solving sets of
non-linear equations, obtaining the analytical Jacobian is a formidable task which re-
quires detailed understanding of the formulation. On the other hand, numerical evalu-
ation of the Jacobian is simple, but computationally intensive. Alternatively, there are
quasi-Newton methods that provide computationally inexpensive approximations to the
Jacobian for solution finding. The best of these methods is the Broyden's method [Broy-
den 1965].
Broyden's method converges superlinearly once the solution estimate is close enough
to the solution, and it has been claimed that it is almost as robust as Newton's
method [Press et al. 1992]. The solution speed would be slower than the Newton's
method employing the analytical Jacobian approach, but faster than the numerical Ja-
cobian approach. However, as the Jacobian approximation matrix used in the Broyden's
method is not always close to the true Jacobian at the solution, the insights into the
harmonic interaction would be compromised. Further work is required to implement
Broyden's method for direct connected units to determine the convergence threshold as
well as to further substantiate those claims on the robustness and speed of Broyden's
method when being applied to direct connected systems.
5.9 CONCLUSIONS
A twelve-pulse Unit Connection harmonic model has been developed using a unified
algorithm, in which all of the system variables and equations are solved simultaneously.
The effects of the convertor control, generator saliency and saturation, ac terminal volt-
age and dc current harmonics are formulated using a set of mismatch equations which is
solved by using Newton's method. The reactive power control by minimizing firing angle
5.9 CONCLUSIONS 81
is maintained by varying generator excitation in the absence of convertor transformer
on-load tap-changers. The mixture of real and complex variables as well as equations
has been decomposed into real quantities, and the Newton's solution is formulated using
real values in positive harmonics only. By setting the small elements of the Jacobian of
partial derivatives to zero and exploiting its sparsity, the solution is greatly accelerated
while preserving the robustness of Newton's solution.
The initial estimation based on the simplified fundamental frequency formulation
has been shown to provide a good starting point for the Newton's solution. The gra-
dient of the iterating function of the Newton's solution, represented by the Jacobian of
partial derivatives, is proportional to the generator subtransient reactance, and inversely
proportional to the commutation reactance. Thus the gradient is large for large gen-
erator subtransient reactance and small commutation reactance values, which can lead
to divergence. However the unified solution has been found to be fast and robust for
typical reactance values and no modification is needed for harmonic analysis of direct
connected units. This is a significant improvement in comparison with sequential IHA
algorithms, in which divergence occurs when the commutation reactance is smaller than
0.1 p.u. and generator impedance is larger than 0.1 p.u ..
The unified solution has been intervalidated with time domain simulations using
the PSCAD2-EMTDC program. For the first time, it is possible to intervalidate dy-
namic simulations with a harmonic domain analysis for direct connected units. It has
been found that the large snubber capacitance used in time domain simulations to aid
numerical stability does have significant influence on the harmonic levels.
Chapter 6
HARMONIC BEHAVIOUR OF DIRECT CONNECTED
UNITS
6.1 INTRODUCTION
In the absence of ac harmonic filters, the generator terminal voltages are distorted and
the commutation voltages are the sinusoidal voltages E" behind generator subtransient
reactance. However as E" is not physically accessible and varying with the generator
terminal voltages as well as currents, the determination of firing angles in direct con-
nected units are more complex than in conventional HV dc schemes. Alternatively, firing
angles can be referenced to the generator terminal voltages, leading to a simpler firing
control mechanism since generator currents are not required in the determination of fir-
ing angles. Section 6.2 discusses the advantages and disadvantages of using E" and the
generator terminal voltage as firing angle references.
Since the absence of ac filters also leads to a longer commutation period (also called
overlap angle), firing angle limits due to overlap angle constraints are examined in Sec-
tion 6.3. As transmitting power from remote hydro generating stations is one of the
most suitable applications for the direct connected concept, the involved generators are
usually of salient-pole type, and they would normally be connected in a Group Con-
nection arrangement. The effects of generator saliency, rotor angle and saturation, and
the number of generators in Group Connections on the harmonic levels are investigated
in this chapter, with particular attention paid to the generator harmonic ratings. The
generator terminal power factors in direct connected units are also compared with those
in perfect ac filtering HV dc schemes in Section 6.4.
In the absence of ac filters and local load, generator speed can be adjusted to suit
the optimal operation of the turbines. The economic advantages of the adjustable speed
operation of Unit Connections have been well documented [Naidu and Mathur 1989].
However since the transformer leakage and generator subtransient reactance vary with
speed, the harmonic content on the ac and dc side of the convertor are also varied with
speed. Moreover the dc side harmonics will be varied over a wide range of frequency,
and consequently direct connected units are more likely to be operated under resonance
conditions on the dc side of the convertor than conventional schemes. Sections 6.7
84 CHAPTER 6 HARMONIC BEHAVIOUR OF DIRECT CONNECTED UNITS
and 6.8 examine the impact of adjustable speed operation on the ac and dc harmonic
levels, particularly in the presence of dc side resonances.
6.2 FIRING ANGLE REFERENCES FOR DIRECT CONNECTED
UNITS
In the presence of perfect ac filtering, the generator terminal voltages are the sinusoidal
commutation voltages to which the firing angles are commonly referenced. On the other
hand, the generator terminal voltages are distorted in direct connected units and the
sinusoidal commutation voltage is the voltage E" behind subtransient reactance as shown
in Figure 6.1 for a non-salient generator. The valves must be fired after the crossings of
E" to avoid mis-firing, but can be fired before the crossings of the terminal voltages as
E" leads the terminal voltage V for a lagging power factor angle of less than 90
0
, which
is common in direct connected schemes. The cosine of the angle <Pe, measured from E"
and II, is defined as the internal fundamental displacement factor; cos <P is the generator
terminal power factor.
E"
_____
Figure 6.1 Phasor diagram of a non-salient generator in direct connected units.
The advantage of using the ideal sinusoidal voltage behind subtransient reactance
E" as the firing angle reference is that the limit for mis-firing is simply zero degree.
However since E" is not physically obtainable, accurate measurements of the currents as
well as generator terminal voltages are required to calculate E". The extra monitoring
signals required can lead to errors in the determination of firing angles in the control
system. The determination of firing angles referenced to E" is further complicated in
the presence of saliency, as depicted in Figure 5.3, since E" varies with the generator
terminal voltages, currents as well as rotor angle. On the other hand, by choosing
the generator terminal voltage as the reference for firing angles, firing control process is
simplified considerably as the currents are no longer required to be monitored. Moreover
the existing firing control for conventional HV dc schemes, in which the reference is the
generator terminal voltage, can then be used for direct connected units with practically
no modification.
Figure 6.2 shows the relationship between firing angles referenced to the generator
6.2 FIRING ANGLE REFERENCES FOR DIRECT CONNECTED UNITS
Firing Angles - Terminal Voltage Reference
50,----,----,----,----,----,----,----,----,----,
40
30
20
10
. ~
. ~
(i)
/' /(iii)
,/
_ 2 0 L L L L ~ ~ ~ ~ __ J_ __
o 5 10 15 20 25 30 35 40 45
Firing Angles - E" Voltage reference
85
Figure 6.2 Relationship between firing angles referenced to the generator fundamental terminal volt-
age and internal voltage E"; (i) Perfect ac filtering, (ii) and (iii) direct connected units with non-salient
generators with X" = 0.2 and 0.3 p.u. respectively.
fundamental terminal voltage and internal voltage E" for typical X" values of 0.2 and
0.3 per unit at generator rated power. As E" typically leads VI by 5 to 15, a zero firing
angle referenced to the terminal voltage implies a firing angle of 5 to 15 with reference
to E" at rated power.
For comparison between direct connected units and conventional HV dc schemes
which are assumed to have perfect ac filtering, it is necessary to use E" as the reference.
But for the operation of direct connected units, it is simpler to use the generator terminal
voltage as the reference. Thus both references are used where appropriate. Firing angles
referenced to E" are simply referred to as firing angles (which has been conventionally
defined in textbooks), whereas firing angles referenced to the terminal voltage would be
clearly indicated.
It should be noted that the firing angle must be referenced to the sinusoidal commu-
tation voltage E" in the simplified convertor formulae [Kimbark 1971] since the generator
terminal is an intermediate node and inaccessible when using the simplified formulation.
On the other hand, the firing angle can be arbitrarily referenced, for instance to the
terminal voltage or any other voltage, in the unified iterative solution. This is possible
because the effects of generator subtransient reactance and the transformer leakage reac-
tance on the commutation process can be considered separately in the iterative solution
by allowing for distortion on the pseudo-commutation voltages, which have been chosen
as the generator terminal voltages in the described model.
Commutation Angles
50
45
,
,
,
40
,
,
,
,
",,(iii)
(f) 35
,
, ,
(])
, ,
(])
,
(") "
....
30
' .... II ".
J..l= 28 Ol .... .....
(])
....
Q
....
ca
25
()
(i)
....
.;::
.... ,
...... 20
....
()
(])
jjJ
15
10
5
0
0 5 10 15 20 25 30 35 40 45
Firing Angles (Degrees)
Figure 6.3 Comparison of commutation periods for direct connected units and perfect ac filtering
schemes; (i) Perfect ac filtering, (ii) and (iii) direct connected units with non-salient generators with X"
:::: 0.2 and 0.3 p.u. respectively.
6.3 OVERLAP ANGLE CONSTRAINT
The absence of ac filters lengthens the commutation period since the generator sub-
transient reactance is also part of the commutating reactance. This is illustrated in
Figure 6.3 for generators with typical X" of 0.2 and 0.3 per unit, and transformer with
typical leakage of 10%. The firing angle in the perfect filtering case is referenced to the
sinusoidal terminal voltage whereas the internal voltage E" is the reference in direct con-
nected units. In order to keep the commutation process of the two bridges forming the
twelve-pulse convertor group independent from each other, the overlap angle should be
kept under 30. The usual practice is to base the system design on a maximum overlap
angle between 26 and 28. For a /J = 28 limit, the minimum firing angles are 10 and
14.9, as shown in Figure 6.3, for direct connected units containing non-salient genera-
tor with X" of 0.2 and 0.3 per unit respectively. The minimum firing angles for other
overlap angle limits are listed in Table 6.1. The overlap angle limit is not a problem in
the presence of perfect ac filtering for the same system.
6.4 GENERATOR TERMINAL POWER FACTOR
The cos if; (or fundamental frequency displacement factor between the generator terminal
voltage and current) is commonly used by generator manufacturers as power factor. By
comparing the power factors of generators used in direct connected units and in perfect
6.4 GENERATOR TERMINAL POWER FACTOR 87
J-l
Ctmin
X"
= 0.2 p.u.
X"
= 0.3 p.u.
26 12.7 17.8
27 11.3 16.3
28 10 14.9
Table 6.1 Minimum firing angles for various overlap angle limits. 'Ifansformer leakage is 10%.
ac filtering schemes, this section investigates if any extra capacity in terms of generator
terminal power factor is required for direct connected units.
Figure 6.4(a) shows the internal fundamental displacement factors cos(e) measured
between the ideal sinusoidal commutation voltage behind subtransient reactance and the
generator current. It is apparent that the larger the subtransient reactance, the longer
the commutation period and thus the smaller the internal displacement factor. On the
other hand, the generator terminal fundamental power factors, shown in Figure 6.4(b),
do not follow the same trend. Although more total reactive power is required by the
convertor for the larger subtransient reactance, the machine own internal reactive power
(the amount corresponding with the subtransient reactance) is larger, leading to a higher
generator terminal power factor for commutation angles of 30 or less. Both of the
internal displacement factor and generator terminal power factor increase with smaller
firing angles for commutation angles of 30 or less, but decrease when the commutation
angles exceed 30.
For J-l < 30, the generator terminal power factor in direct connected units is higher
than in perfect ac filtering schemes for a given firing angle. Thus for a generator with
the rated terminal power factor of 0.915 at a typical 15 firing angle in the presence of
perfect ac filtering as indicated by the line iv in Figure 6.4(b), the same generator can
be used in direct connected units with a higher nominal firing angle, for instance, 22
or 26 for X" of 0.2 or 0.3 per unit respectively. It is assumed that the reactive power
contributed by the ac filters are not taken into account when calculating the reactive
power capacity of generators in the perfect ac filtering case. However, this gain in power
factor capacity could be offset by the need to have a higher nominal firing angle to keep
the overlap angle under 30 in direct connected units. For instance, the nominal firing
angle should be within 16.3 < Ctnominal < 26 to keep the overlap angle under 27 and
generator terminal power factor greater than 0.915 (obtained from the nominal 15 firing
angle in the presence of perfect ac filtering). As long as the firing angle is kept within
the limits corresponding to the overlap angle and power factor constraints, practically
no de-rating of generators in terms of power factor for direct connected operation is
required.
88
0.95
0.9
0.85
0.8
/
0.75
0.7
0.65
0.6
0
0.95
0.9
0.85
0.8
0.75
CHAPTER 6 HARMONIC BEHAVIOUR OF DIRECT CONNECTED UNITS
/
/
/
/
5
/
/
/
Generator Internal Displacement Factor
(i)
(ii)
.... --- ....
,
,/ ".
'.
'.
.....
,
,
" ....
.... ,
". ""
'. ,
'. ,
, ,
" ,
" ,
". ",
'. ,
'. ,
, ,
" ,
" ,
".'
,
10 15 20 25 30 35 40 45
(a)
Generator Terminal Power Factor
;:;:,,::::'-'-'-
'.
'.
,
....
(iv)
, '.
.... ,
", "'" (iii)
, ,
, ,
" '
, ,
, ,
(
")' "
II " "-
, \
(i)
, \
\ ,
" \,
\ '.
, \
\ "
\
\
\
0.7L-----L-----L---__ L-__ ____ ____ ____ ____
o 5 10 15 20 25 30 35 40 45
(b)
Figure 6.4 Comparison of (a) internal displacement factors COS(;e), and (b) generator terminal power
factors cos ; for direct connected units and perfect ac filtering schemes; (i) Perfect ac filtering, (ii) and
(iii) direct connected units with non-salient generators with X" = 0.2 and 0.3 p.u. respectively; (iv)
Terminal power factor rating in the presence of perfect ac filtering for a typical 15 nominal firing angle.
6.5 EFFECTS OF SALIENT-POLE GENERATORS 89
6.5 EFFECTS OF SALIENT-POLE GENERATORS
6.5.1 Effect of Generator Saliency on Harmonic Levels
In a salient-pole machine the 12n -1 negative-sequence voltages are made of two compo-
nents due to the corresponding negative-sequence 12n - 1 and positive-sequence 12n + 1
convertor currents. This can be expressed in matrix form for 11th and 13th har-
monies as:
[
Vll ] = [Z91l,1l Z91l,13] [ -Ill ]
Vi3 Zg 13,11 Zg 13,13 - h3
(6.1)
where the minus sign indicates that the currents are flowing into the generator. Ex-
panding equation 6.1 using the saliency cross coupling terms derived in Chapter 3 yields
(resistive terms are neglected):
where
ViI Vll,ll + Vll,13
-Zgll,llIll - Zgll,13 I 13
= -jAIIllle
jLIll
-
A
B
x" + X"
11 q d
2
X"-X"
11 q d
2
(6.2)
(6.3)
Let P be the difference in phase angles between the harmonic current pair 12n 1,
I.e.
P = LIll - Lh3, (6.4)
then the 11th harmonie voltage components caused by the 11th and 13th currents can
be expressed in terms of the 11th current phase angle as:
Vll,ll
Vll,13
-jAIIllle
jLIll
-jBlh3Ie
jLIll
e
j
(-P-26) .
(6.5)
(6.6)
Thus the cross coupling term (ViI,13) is lagging the self-term (Vi1,1l) by an angle of
P+28.
I
I
....... -T- ......
,; I " 11113
I I \'
I iP+2o..:.
I ~ 1 \
I
"
'{1,1l
"""',------1110-
~ 3
~ 1
"lf3,lt - - - .... /
, .....
, "
P+28 , \
"-+rr----___ 1
I
,;
Figure 6.5 Phasor diagrams of saliency effect on harmonic voltages.
Similarly for the 13th voltage
V13,13 + V13,1l
-jClh3le
jLh3
- jDllnlejLh3ej(28+P)
where
C
x"+x"
13 q d
2
X"-X"
13 q d
2 '
D
\
\
(6.7)
(6.8)
and the cross coupling term (V13,1l) is leading the self-term (V13,13) by an angle of P+2J.
It is easier to visualize equations 6.2 to 6.8 by using phasor relationships, shown
in Figure 6.5, with all phase angles referenced to the generator fundamental terminal
voltage phase angle. The addition of the two voltage components for each harmonic will
result in larger or smaller voltage levels depending on the rotor angle and the difference
between the harmonic current phase angles P. The rotor angle is dependent on the
power output, synchronous reactances, terminal voltage, and generator saturation. The
phase angle difference P is dependent on the firing and commutation angles, the later
being in turn a function of the ac terminal voltages and their harmonics, de current and
its harmonics, firing angle and commutation reactance. As the ac terminal harmonic
voltages depend on generator saliency, all these quantities are inter-dependent. There-
fore, an accurate harmonic solution in the presence of saliency can only be obtained by
solving all the non-linearities in a unified iterative solution as described in Chapter 5.
However it is possible to estimate the worst effects due to saliency on the harmonic
current levels by using the derived saliency formulation in Chapter 3. Consider the
harmonic voltages in the case where the total phase angle 26 + P is 180
0
With reference
to Figure 6.5, the 11th harmonic voltage can be obtained from equations 6.2 and 6.3 as:
Wnl
(6.9)
Since typically lIn I n lId, and k ~ ::; X ~ ::; X ~ Wnl can be approximated as
(6.10)
with upto 5% error only. Similarly,
(6.11)
As a result, the salient-pole generator can be considered as a non-salient generator
with its subtransient reactance equal to the direct-axis subtransient reactance of the
salient-pole generator. As the direct-axis subtransient reactance is smaller than the
quadrature-axis subtransient reactance, the commutation period is the shortest and
thus harmonic current levels are the largest. It is noted that this observation is valid
only for direct connected systems in which no other harmonic besides the characteristic
harmonics is produced by generator saliency.
On the other hand, when the angle 26 + P is 0
0
, the harmonic voltage levels are:
1 1 X ~ IInl
1 3 X ~ I I d
(6.12)
(6.13)
and thus the salient-pole generator can be considered as a non-salient machine with sub-
transient reactance equal to the quadrature-axis subtransient reactance. Consequently,
the current harmonic levels are the lowest in this case.
By using a non-salient generator with a subtransient reactance equal to the ~ of the
salient-pole generator, the maximum limit for the harmonic currents can be obtained. On
the other hand, the minimum limit for the harmonic currents in a salient-pole machine
can be obtained by using a non-salient generator with its subtransient reactance equal
to X ~ Figure 6.6 (a) shows the 11th harmonic current levels for a test salient-pole
generator (solid lines) when operated at rated power, and the maximum and minimum
limits for harmonic currents. The limits are obtained by using non-salient generators
with direct-axis or quadrature-axis subtransient reactance values. The results for the
salient-pole case are obtained from the unified solution developed in Chapter 5; the
direct and quadrature-axis subtransient reactances are 0.2 and 0.367 p.u. respectively.
Other parameters for the salient-pole generator are listed in Appendix AI. The firing
angles are measured from the ideal sinusoidal commutation voltage behind subtransient
reactance. Figure 6.6(b) shows the harmonic voltage levels corresponding to the currents
limits. Since the harmonic currents are dependent on the subtransient reactance and
the harmonic voltages are dependent on the currents as well as the generator harmonic
impedance, the minimum current limit does not always correspond to the maximum
voltage limit.
For commutation angles of 30 or shorter, the harmonic levels reduce with smaller
firing angles. However, this trend is not applicable to cases where the commutation
angles are longer than 30.
6.5.2 Effect of Saliency on Generator Terminal Power Factor
The maximum and minimum limits for commutation angle in direct connected units
containing salient-pole generators can also be found by using non-salient generators
with appropriate subtransient reactance values as shown in Figure 6.7. In order to keep
J-t < 28, the firing angle should be approximately 10 or larger when the salient-pole
generator acts as a non-salient generator with X" = X ~ ; 18 or more when X" = X ~
With reference to Figure 6.8, the corresponding generator terminal power factors for
a = 10 in the case where X" = X ~ and a = 18 for X" = X ~ case are approximately
the same, i.e. 0.96. In practice, the commutation angle would be in between the limits as
shown in solid line in Figure 6.7 for the considered salient-pole generator. In this case, the
minimum firing angle should be 12.5, which also results in a similar generator terminal
power factor of 0.96, as shown in Figure 6.8. Thus in the consideration of generator
terminal power factor rating, it is not necessary to minimize generator saliency by having
full cage damper windings. However full cage damper windings could be recommended
due to its positive effect on the generator dynamic response, for instance, to damp
oscillations, suppress hunting, provide additional torque for synchronizing generators,
etc. [Kimbark 1968].
6.5 EFFECTS OF SALIENT-POLE GENERATORS
11 th Harmonic Current
o.o8,-----,-----r---,----.-----,--.,---,----,-----,
0.07
0.06
0.05
0.04
0.03
--..--
0.02
- - - - -
0.01
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
,," (ii)
"
"
"
"
"

o 5 10 15 20 25 30 35 40 45
(a)
11 th Harmonic Voltage
0.3,-----,-----r---,----.-----,--.,---,----,-----,
0.25
...-;- 0.2
::J
0.
.......
Q)
" 0.15
:t::::
C
OJ
eel
:2: 0.1
0.05
"
"
"
"
"
"
"
(ii) ,,"
"
"
(i)
"
"
"

(b)
93
Figure 6.6 (a) Maximum and minimum limits for the 11th current harmonics of a salient-pole gener-
ator in direct connected units; (b) Voltage harmonics corresponded to the current limits. Solid: salient-
pole generator, Dashed: non-salient generator with (i): X" = and (ii) X" = X
t
= 10%.
Figure 6.7 Commutation angles limits for salient-pole generators in direct connected units. Solid:
salient-pole generator, Dashed: non-salient generator with (i): X" = ~ = 0.2 p.u., and (ii) X" = X;
= 0.367 p.u.; X
t
= 10%.
....... -.- .......
0.95
~
,
,
,- ,-
,-
,-
,
,
/
"
/
,
....
....
0.9
/
"
....
"
/
" " /
"
"
(ii)
"
" "
" " 0.85
"
\
\
"
"
"
\
\
\
"
\
0.8
\
(i)
\ \
\
\
\
\
\
\
0.75 \
\
\
\
\
0.7
0 5 10 15 20 25 30 35 40 45
Figure 6.8 Generator terminal power factor (cos ) for salient-pole generators. Solid: salient-pole
generator, Dashed: non-salient generator with (i): X" = ~ = 0.2 p.u., and (ii) X" = X; = 0.367 p.u.;
X
t
= 10%.
6.5.3 Effect of Rotor Angle
With reference to Figure 6.5, the generator harmonic voltages are directly related to the
rotor angle, and thus the commutation periods which depend on the terminal voltage
distortion are also related to the rotor angle in the presence of saliency. The higher the
degree of saliency, the larger the rotor angle effect on the harmonic levels. Figure 6.9
shows the effect ofrotor angle on the commutation angle and the 11th harmonic current
for various degrees of saliency. The degree of saliency in the subtransient reactances is
defined as:
(6.14)
In this case study, ~ and ~ are varied so that their average is constant and
equal to 0.2835 p. u. to enable the results to be compared with those calculated from
the simplified convertor formulae. The dc current order and firing angle are also kept
constant and a large smoothi'ng inductor (lOR) is used to minimize the effect of the
dc ripple. The vertical line indicates the saliency level for a machine with ~ and ~
equal to 0.2 and 0.367 p.u. respectively. The 5% transformer leakage is obtained when
a Group Connection of 2 generators with X
t
= 10% is operated in the unit connected
mode, i.e. the output power is approximately half of the total rated power of the group
connected unit. The Group Connection implementation is examined in Section 6.6. The
firing angle is referenced to the generator terminal voltage so that the firing angle control
is effectively independent of rotor angle.
Since the simplified convertor equations use the average of ~ and ~ in the com-
mutation reactance, the commutation angle and the harmonic levels remain constant
regardless of the degree of saliency as long as the average of the subtransient reactance
is unchanged. On the other hand, a saliency formulation that does not include the
rotor angle effect, e.g. with the rotor angle assumed to be zero, such as by Eggleston
et al. [1988], can lead to large errors in the solution as also shown in Figure 6.9. For
small six-pulse direct connected schemes, the effect of rotor angle is more substantial due
to the larger harmonic levels involved in the six-pulse convertor operation. Therefore, it
is important that the rotor angle is taken into account in the harmonic solution of direct
connected schemes in the presence of saliency.
6.5.4 Effect of Generator Saturation
As examined in Chapter 3, generator saturation produces three main effects on the gen-
erator steady state behaviour. Firstly, saturation causes an increase in the dc excitation
level, but since the excitation level is not used in the harmonic formulation, this effect
can be ignored. Secondly, magnetic saturation produces odd order harmonics due to the
.-....,
CJ)
Q)
Q)
....
0>
Q)
o
co
()
.;;:
"0
Q)
[jJ
---
Commutation Angle
17
16
15

o M M
Saliency (%)
(a)
11 th Harmonic Current
0.07,-------------.,-------------,------------,---------------,
0.065
.-... 0,06
:J
c..
---
Q)
-g 0.055
-
'c
0>
co
0.05
0.045
1
1
____________ 1 __________ __
1
0.04 '--__________ ...1...-__________ ---'---__________ --'-__________ ---'
o 60 80
Saliency (%)
(b)
Figure 6.9 Effect of rotor angle on the commutation angle (a), and on 11th harmonic current level (b)
in the presence of generator saliency; Q = 10 referenced to the generator terminal voltage, X
t
= 5%.
The vertical line indicates the saliency level for a machine with = 0.2 p.u. and = 0.367 p.u.
6.6 GROUP CONNECTION IMPLEMENTATION 97
Lpu
IDC
Zleak
Pu
r VDC
Sbase = n x S Gl base
n
Gn
S tranf base = '2' X Sbase
-=-
~
Lpu
IDC
Zleak
Pu
ZGPU =ZGl
PU
r VDC
Sbase = n X SGI base
n
Stranfbase= '2' X Sbase
-=-
Figure6.10 Simplification of a Group Connection to the basic Unit Connection for harmonic analysis.
distortion of the air gap flux waves. However it has been shown in Chapter 3 that the
harmonics caused by generator saturation are negligible. Thirdly, saturation causes a
decrease in generator synchronous reactances which lead to a smaller rotor angle, and
thus changes in generator harmonic cross couplings. However since the decrease in ro-
tor angle is in the order of a few electrical degrees for generators operated at full load,
the changes in the harmonic levels are small. Moreover as the generator subtransient
reactances ~ and ~ of a synchronous machine with damper windings are associated
with fast transient flux that flows mostly in air, they remain constant for a wide range
of terminal voltages under saturation conditions. Thus in the context of the harmonic
analysis of a Unit Connection containing generators with damper windings, the effect of
the generator saturation is insignificant.
6.6 GROUP CONNECTION IMPLEMENTATION
In a Group Connection of n generators, shown in Figure 6.10, the system power base is n
times larger than the generator power base, and the convertor transformer nominal power
rating is half of the system power base, i.e. i times larger than individual generator
power rating.
As the generators are connected in parallel, their combined actual impedance is
reduced by n times. However since the power base of the system is now n times larger,
the impedance base is reduced by a factor of n. Thus the combined per unit generator
impedance is the same as the per unit impedance of each generator. Consequently, a
Group Connection can be simplified to the basic Unit Connection with no adjustment
Generators System base Trans. base
X"
d
X
t Ide
(MVA) (MVA) (p.u.) (p.u.) (p.u.)
3 300 150 0.2 0.1 1
2 300 150 0.3 0.1 0.67
1 300 150 0.6 0.1 0.33
k nx100 0.1
lx
n
1
3 300 150 0.2 0.1 1
2 200 100 0.2 0.067 1
1 100 50 0.2 0.033 1
k kx 100 0.2 0.1 1
Table 6.2 Adjustment for the generator and convertor parameters of Group Connections operating
with a reduced number of ge:p.erators. Generator base = 100 MVA, transformer rated power = 150 MVA;
n, k are the number of nominal generators, and remaining generators respectively.
to the per unit values, as demonstrated in Figure 6.10. The harmonic currents on the
primary side of the convertor transformers are shared evenly between the generators,
given the assumption that the generators are identical and operated at similar power
level, which is normally the case in a Group Connection. Consequently, the harmonic
currents in a generator when being connected in a Group Connection are the same as
when it is used in a Unit Connection. Note that when calculating the actual harmonic
currents flowing in an individual machine, the power rating of the machine is used as
the base value instead of the system base.
However when a Group Connection containing n generators is being operated with
k generators, the harmonic levels will be affected considerably due to changes in the dc
power order and reactances. If the system power base is chosen as n times the generator
power base, the combined generator subtransient reactance is increased by The dc
output is reduced to while the transformer leakage remains unchanged. The required
conversion of the per unit reactances and power order is given in the top half of Table 6.2
for a Group Connection consisting of three 100 MVA generators. The transformers are
150 MVA units with 0.1 p.u. leakage reactance.
Alternatively, the system base can be chosen as k instead of n times generator power
base. In this case, the per unit transformer leakage decreases to while the combined
generator subtransient reactance, and dc output remain the same in per unit quantity.
The conversion in this option, given in the bottom half of Table 6.2, is simpler than
the other option since only one parameter which is the per unit transformer leakage
reactance is required to be adjusted in the simulations. Consequently, a Group Connec-
tion containing n generators when operating with only k generators is equivalent to a
Unit Connection with the transformer leakage reactance adjusted by a factor The
system base is chosen as k times the generator base.
6.6.1 Commutation Angle and Generator Power Factor in Group Con-
nections
As the effective per unit transformer leakage and thus commutation reactance decreases
with the smaller number of generators, the commutation angles are also reduced, as
shown in Figure 6.11(a). This is advantageous in terms of satisfying the overlap angle
limit, i.e. the firing angle can be reduced when a Group Connection is being operated
with a reduced number of generator while maintaining the overlap angle to less than
a set limit. With reference to Figure 6.11(a), the minimum firing angle to satisfy a
f-L = 28 limit can be reduced from 10 for the nominal 3 generators unit to 2 in the
unit connected mode. However a safe firing angle limit of 10 to 15 should be maintained
to avoid mis-firing.
The contribution of generator subtransient reactance to the overall commutation
reactance is larger when generators are taken out of a Group Connection. This leads to a
higher generator terminal power factor for a given firing angle as shown in Figure 6.11 (b).
For the considered generator, the generator terminal power factor increases from 0.96 in
the group connected mode with 3 generators to 0.98 in the unit connected mode for a
firing angle of 10. Hence the generator power factor rating can be obtained from the
group connected mode with the nominal number of generators connected.
6.6.2 Harmonic Content in Group Connections
Upon disconnection of a generating unit, the harmonic current loading in the remaining
generators is significantly increased for a given firing angle. For a Group Connection
containing 3 generators, the increase in the 11th harmonic current and the current THD
is about 0.01 per unit for each generator taken out of service, as shown in Figures 6.12.
The parameters of the generator and transformer used are as listed in Table 6.2. The
number of generators in a Group Connection is thus an important factor in the cal-
culation of the generator and transformer harmonic ratings. The generator harmonic
ratings increase proportionally to the maximum number of generators in a Group Con-
nection. However it is possible to reduce the generator harmonic loading by specifying a
smaller nominal firing angle when a Group Connection is operated with fewer than the
nominal number of generators, provided that the overlap angle limit and the safe limit
for mis-firing are satisfied. Moreover the reduction in the generator fundamental power
factor loading can provide extra capacity to accommodate the increase in the genera-
tor harmonic current loading. Figure 6.13 shows the harmonic voltage variation with
respect to the number of generators in service. Although the increase in the harmonic
current loading can be accommodated by the reduction of the generator terminal power
factor, the increase in harmonic voltages must be taken into account in the insulation
requirement of the generators.
1'.,. ..
OF CANTERBURY
CHRISTCHURCH, N.Z.
100
CJ)
(])

0)
(])
0
ca
C,)
C
-
C,)
(])
W
Commutation Angles
/.L= 28
,
, ,
25
,
,
,
3 generators
,
20
, ,
',2
,
,
,
, ,
".,1
,
,
15
,
,
10
5
0
0 5 10 15 20 25 30
(a)
," -'-'- -
/
/
..... - - - --.-
35 40 45

':',.,1 generator
0.9
0.85
0.8
0.75
5
,
.......... ".
, ,
... .. .............. , ...
3
, "
..... " ".
, ,
, "
"" ".
, '.
, ,
, "
, "
, .
, '.
, '
, "
/
, "
Perfect ac filtering
10 15 20 25 30 35
(b)
, ,\,
, \
40
, ,
, "'
, "'
, .
,
45
Figure 6.11 (a) Commutation angle and (b) generator terminal power factor in a Group Connection
being operated with various number of generators. Xt = 10%, = 0.2 p.u.
Generator 11 th Harmonic Current Generator 13th Harmonic Current
0.09
0.08 O,OB
0.07 0,07
:;;
0."
S
Q)
0.05
'0
:J
:g 0.04
0>
III
::l:
0,03
0.02 0.02 3 generators
,
/
0.Q1 0.01
/
30 35 40 45
0
0

10 15 20 25 30
3'
40 45
Firing Angles (degrees) Firing Angles (degrees)
Generator 23rd Harmonic Current Generator 25th Harmonic Current
0.03
0.025 0.025
:;;
0,02
S
Q)
, '0
0,015
:J ,

,
,
0> ,
III ,
::l: 0.01 ,
,
,
/2 /
"".-.-.; ....... ;-,- - .....
,
, 0.005 0.005
,'-
3 generators
30 35 40 4'
0
0
__ __
o 10 15 m
3 generators
10 15 20 25 30 35 40 45
Total Generator Harmonic Current Distortion
0.12
0.1
:;;
0.08
S
Q)
'0
0.06
.a
'c
0>
III
::l: 0.04
0.02
0
0 10 15 20 25 30 35 40
4'
Firing Angles (degrees)
Figure 6.12 Generator harmonic currents in a Group Connection being operated with various number
of generators. Xt = 10%, X" = 0.2 p.u.
102
::i
.&
Q)
"0
.a
'c
Ol
<U
:2
::i
.&
Q)
"0

C
Ol
<U
:2
Generator 11 th Harmonic Voltage Generator 13th Harmonic Voltage
0.2,---,-------,---,----,---,-----,--,---,--,
0.1"5
0.1
::':._.-'
0.05
0
0
0.14
0.12
0.1
0.08
0.06
0.04,
.' ,

'-
,
0.02
0
0
0.15
/1
,
,
::i
,
.&
, ,
, ,
Q)
,
........... 2
"0
0.1
::::J
,
'-g .' ,
,
,
,
3 generators Ol
,
<U
:2
ODS
,
,
,
0
10 15 20 25 30 3S 40 46 0
Generator 23rd Harmonic Voltage
0.14
0.12
0.1
::i
.&
0.06
Q)
"0
.a
'c 0.06
Ol
<U
:2
0.04
0,02
0
0
, ,
,
, ,
,
"./1
,
,
,
,
,
,
,
,
" 2
,
,
,
,
,
,
3 generators
10 15 20 25 3() 3S 40
Generator 25th Harmonic Voltage
10 15 20 25
i
i
,
,
,
,
/2
,
,
,
,-
/
3 generators
35 40
Total Generator Harmonic Voltage Distortion
0.35
0.3
,
,
,
0.25 ,
,
,
,
,
::i
, ,
,
.&
,
,
0.2
Q)
,
,
"0
/;-
,
,
.a
,
,
,
'c 0.15 , ,
,
Ol
,
2
<U
:2 ,
3 generators
0.1 ..... -.- ..... ,
,
----.- ....
0.05
0
0 10 15 20 25 30 3S 40 45
46
45
Figure 6.13 Generator voltages in a Group Connection being operated with various number
of generators. Xt = 10%, X = 0.2 p.u.
6.6 GROUP CONNECTION IMPLEMENTATION
...-;-
::J
C.
----
(])
"'0
::J
+-'
c
OJ
co

Generator Equivalent Negative-Sequence Currents
o.2,----,---,----,----,----,---...... --,---,-----,
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
-----
"
,," 2
3 generators

103
Figure 6.14 Generator equivalent negative-sequence currents for a Group Connection. Xt = 10%,
X" = 0.2 p.u.
(]) 40
Q
C
(])
.....
(])
-
(])
30
a:
(])
0)
!
+'
a
20
>
<a
c
'E
10
.....

(/)
(]) 0
OJ
c

0)
c
-10
.;;::
u::
o
y
,,"
y"
w
?-'
"",if
<"
//
---
?
" /
V

V

/
3 generators
2 generators
1 generator
5 10 15 20 25 30 35 40
Firing Angles - E" Voltage reference
45
Figure 6.15 Relationship between firing angles referenced to E" and to the generator terminal voltage
V for a Group Connection. X
t
= 10%, X" = 0.2 p.u.
No. of Nominal firing angle D:
nom
restriction limits
Gen- fL ::; 28 mis-firing cos 1> 0.915
i
len::; 10% Overall limits
erators constraint constraint constraint constraint
1 2 10 ::; 26.6 ::; 16 10 ::; D:
nom
::; 16
2 6 10 ::; 24.5 ::; 22 10 ::; D:
nom
::; 22
3 10 10 ::; 22.5 ::; 26 10 ::; D:
nom
::; 22.5
No. of Nominal firing angle D:
nom
restriction limits
Gen- fL ::; 26 mis-firing cos 1> 0.935
ii
len::; 10% Overall limits
erators constraint constraint constraint constraint
1 4.4 10 ::; 23.2 ::; 16 10 ::; D:
nom
::; 16
2 8.6 10 ::; 20.2 ::; 22 10 ::; D:
nom
::; 20.2
3 12.7 10 ::; 18 ::; 26 12.7 ::; D:
nom
::; 18
Table 6.3 Nominal firing angle restriction limits for group connected units imposed by overlap angle,
mis-firing, generator power factor and continuous equivalent negative-sequence harmonic loading con-
straints. X" =0.2 p.u. and X
t
= 10%; (i) Generator terminal power factor for a nominal 15 firing angle
in the presence of perfect ac filtering; (ii) for a 10 firing angle.
The rotor heating effect caused by convertor loads can be compared with standard
specification ratings of generators by using the equivalent continuous negative-sequence
current len [Krishnayya 1973]:
(6.15)
Figure 6.14 shows the variation of equivalent negative-sequence currents with vary-
ing delay angle for a Group Connection when being operated with various numbers of
generators. In the unit connected mode, the total harmonic loading is typically 10% for
a firing angle of 16, but increases to about 15% for firing angles of 30. As generators
are normally designed to sustain an equivalent of 10% continuous negative-sequence cur-
rent loading, extra capacity for harmonic losses would not be required if the nominal
firing angle is carefully selected. In this case, the nominal firing angle should not be
more than 16, 22 and 26 for the unit connected, group connected with 2 and 3 gener-
ators mode respectively. The restriction limits of nominal firing angles derived from the
overlap angle, generator terminal power factor, continuous equivalent negative-sequence
harmonic loading and mis-firing constraints are given in Table 6.3. Provided that the
nominal firing angles are kept within the derived limits, practically no extra capacity for
generator power factor and harmonic current loading are required in group connected
units.
Firing angles referenced to the generator terminal voltages (for control system that
6.7 EFFECTS OF ADJUSTABLE SPEED OPERATION
1.4
1.2
""""
d
8
[) 1.0
~
0
0.-
0.8
0.6
Linear approximation /. /.
/.
/.
/.
<
/.
/.
/.
/.
........ ~ ~ ~ ~ P C ? ~ ~ ~ . . . . . . . . . . . . . . .. : ................... .
Optimal efficiency /. /.
\ /./.
/.
/.
/.
/.
/.
/.
0.8
7
~
""0:
o
0:
0;.
",.
Oil:
. :
o
Z:
0.9 1.0 1.1
Speed (p.u.)
1.2
Figure 6.16 Typical hydraulic turbine power-speed characteristic.
105
uses V as the reference) can be derived from Figure 6.15, which shows the relationship
between firing angles referenced to E" and to the generator terminal voltage V. For
instance, a 10 firing angle referenced to E" corresponds to a -1 firing angle referenced
to V; 16 to 6, etc. In per unit quantities, the transformer leakage reduces with fewer
generators in service while the dc current order and generator subtransient reactance
are unchanged, resulting in smaller commutation period and generator terminal power
factor angle. However the phase angle between E" and V only increases slightly with
fewer generators since the reduction in power factor angles is small, as indicated by the
closeness of the plots shown in Figure 6.15.
6.7 EFFECTS OF ADJUSTABLE SPEED OPERATION
In the absence of ac filters and local loads, generators can be operated with adjustable
speed to maximize turbine efficiency. The speed of generators are controlled so that they
operate at the optimal efficiency for a machine loading given by the dc power order. A
typical hydraulic turbine power-speed characteristic for optimal efficiency is shown in
Figure 6.16. Within the 20% speed range the optimal efficiency characteristic can be
linearly approximated as indicated by the dotted line.
In theory, the convertor transformers and generators can be designed and rated at
any frequency other than the nominal frequency. But this can incur extra cost in the
106 OHAPTER 6 HARMONIC BEHAVIOUR OF DIREOT OONNEOTED UNITS
design process. Alternatively, convertor transformers and generators that are rated at
the nominal frequency can be used. Within the interested speed range, the transformer
leakage and generator subtransient reactances will vary linearly with speed [CIGRE JWG
14/11-09, 1997]. As a result, the commutation angle is increased with higher speed as
illustrated in Figure 6.17(a). The firing angles are referenced to E". The power base
is taken to be rated power at the nominal 50 Hz frequency, and the dc power order is
taken from the approximated power-speed characteristic for optimal efficiency as shown
in Figure 6.16 .. However since the intended application of adjustable speed operation is
to utilize the generator optimal efficiency in the region where the speed is lower than
the nominal, the commutation period in this region is smaller than those at the nominal
speed. Thus the overlap angle limit can be taken at the nominal speed.
The generator terminal power factors, shown in Figure 6.17 (b), are larger at lower
speeds, i.e. less reactive power is required by the convertor. Hence the generator terminal
power factor rating can also be taken at the nominal speed.
Figures 6.18 and 6.19 show the ac harmonic current and voltage variation with re-
spect to generator fundamental frequency for various firing angles. The total ac harmonic
current loading, indicated by the equivalent continuous negative-sequence current load-
ing, increases as the fundamental frequency reduces due to a reduction in the reactances
and subsequently commutation angle. However since the dc power order also reduces
with lower speed, the increased harmonic loading is not a problem. On the other hand,
the changes in the generator voltage THD is not significant at speed below the nominal
speed. Thus generator harmonic ratings can be taken at the nominal speed and rated
power output.
6.8 EFFECTS OF DC SIDE RESONANCES
6.8.1 On de Side Harmonics
The Unit Connection model developed in Chapter 5 is used to investigate the effects of
dc system resonances of a test system, shown in Figure 6.20, on the dc and ac harmonic
levels. The test system consists of a typical double-tuned filter, smoothing inductor
and a dc line. The smoothing inductor, dc filter parameters, generator and transformer
parameters are taken from the New Zealand Benmore convertor terminal and are listed
in Appendix A2. For simplicity, the dc line is represented by a resistance in series
with an inductance. A more detailed dc line representation can be readily incorporated
into the Harmonic Domain solution by using its harmonic impedances. The chosen line
inductance of 0.2 H corresponds to an overhead dc line of 150 km with typical series line
impedance of 1.33 mH/km 1.
lThe line impedance is obtained from the New Zealand HVdc transmission line
6.8 EFFECTS OF DC SIDE RESONANCES 107
Figure 6.17 (a) Commutation angle and (b) generator terminal power factor variation with speed for
various firing angle. X
t
= 0.1 p.u., ~ = 0.2 p.u. at the 50 Hz nominal frequency.
Generator 11 th Harmonic Current

0.04
0.035
0.03 -'-'-'
,
OJ
,
0,025
,
,
c:
g>
:2:
:i
&
OJ
"
:::J
.-g
Ol
ct1
:2:
:i
&
OJ
"
:::J
'2
Ol
ct1
:2:
0.02
0.015
0,01
Q.Q05
4
L
O
-----',,=-----"'50,-----5=5--------:'50
0.Q1
0.006
/
0.006
,
,
0.004 ,
, ,
, ,
, ,
,
/
0.002
Fundamental Frequency (Hz)
Generator 23rd Harmonic Current
,
,
,
,
'-' ....
/."
, ,
,
I
I
I
,
I
I
/
/
".
---- ...
"""''''' ex = 10
"
..... -.-.-
"' .... 0::;;: 20
,
,
,
,
,
,
,
,
,
,
,
,
,

0.05,------,------.-----,-----,
0.05
cx= 15
0.04
cx= 10
,
,
,
0,03
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
". "
'. "
,
,
,
O.02
4
'=-0--------c
4
""5
50
Generator 13th Harmonic Current
0.045,------.------.-----,-----,
0.04
0.035
0.03 -
OJ
0.025
c:
g> 0.02
:2;
:i
&
OJ
"
:::J

Ol
ct1
:2;
0.016
0,01
O.OO5
4
L
O
-----',,:-----=50,-----5=5------:'50
0.01
Generator 25th Harmonic Current
I
,
,
,
,
I
,
:"'"
,
,
,
/
,
,I ''''''''
",a = 20
,
,
,
,
,
a= 15 \
,
,
,
,
,
,

0.09
0.08
0.07
0.05
0.05
0.04
0.03
40
'.
'.
'.
'.
,
,
'.
,
,
,
,
,
45 50 65
,
,
,
,
,
,
,
,
,
,
,
,
50
Figure 6.18 Harmonic current variation with speed for various firing angle. Xt = 0.1 p.u., = 0.2
p.u. at the 50 Hz nominal frequency.
:;;
8,
Q)
"0
:E
t:
Ol
'"

Generator 11 th Harmonic Voltage Generator 13th Harmonic Voltage

O.OB
,
0.07
,
,
,
,
,
0,06
0,05
0.04
,
,
,
,
,
,
,
0.06
0.07
0.06
0.05
0,04
,
,
,
,
,
,
u= 15
0
,
,
,
,
,
,
,
,
,
,
- _._._._., --------->-1
0.03
4
':-0-----c,,::------::SO:------,::'--------:!50
0.05
0.04
0.03
Generator 23rd Harmonic Voltage
.-.- .J.
,
,
, "
,
,
,
/ / / / - - ,u = 20
/ '
/ '
, '
, '
, "
,
,
,
,
,
,
,
0.05
0.04
:;;
8,
0.03
Generator 25th Harmonic Voltage
...... = 20
,
,
,
,
,
,
,
,
,
,
,
,
.",,' I Q) '-',.(
, "
,/
0.02
,
, ,
0.01 ..... - '"
,
,
,
,
,
,
,
,
,
"0
.-2
-'-'-'-'-'-'- t:
Ol 0.02
'"

,
0,01
,
/
/
,
,
,
,
,
"
,
,
,
,
,
,
,
,
I ".
'.

50
0
40 45 50
"
Fundamental Frequency (Hz) Fundamental Frequency (Hz)

0.12
0.1
0.06
u=200
--------- ....
u= 15
0
'.
'.
'.

,
,
,
,
,
,
50
Figure 6.19 Harmonic voltage variation with speed for various firing angle. X
t
= 0.1 p.u., = 0.2
p.u. at the 50 Hz nominal frequency.
O.8H 10.95 Q
0.2H
0.
923
Il
F
+
0.037 H
1400Q
0.019H
Figure 6.20 A typical de side configuration with a de filter, smoothing inductor and a simplified
representation of a de line.
The 12th and 24th dc harmonics are shunted by the low dc filter impedance shown
in Figure 6.21(a), where the fundamental frequency is 50 Hz. However new resonance
frequencies occur when the de filter is combined with the smoothing inductance and the
dc line impedance as shown in Figure 6.21(b). Parallel resonances occur at 325 and
875 Hz, and series resonances are at 350 and 885 Hz. Of particular importance is the
series resonances at which the dc harmonic current ripples are large since the convertor
acts as harmonic voltage sources. These large harmonic currents could cause harmonic
interference in the dc line, and raise the required ratings of the filter.
The series resonances at 350 and 885 Hz are significantly away from the 12th har-
monic (600 Hz) to have any substantial effect on this harmonic in the interested speed
range of 35 to 65 Hz. However, they will have significant effects on the 24th, and the
6th as well as the 18th which could exist under unbalanced operation. Considering an
unbalance case where the convertor transformer leakage reactance are 11.2 and 11.7% (at
the nominal 50 Hz) for the two six-pulse valve groups respectively, Figure 6.22(a) shows
the harmonic voltages at the dc filter point for the 6th, 12th, 18th and 24th for a speed
range from 35 to 65 Hz in 0.25 Hz steps. At 36.9 Hz fundamental frequency, the 24th
harmonic is drastically magnified due to the 885 Hz series resonance. The generators
are operated at the optimal efficiency with the approximated power-speed characteristic
as depicted in Figure 6.16.
The effects on the 6th and 18th harmonics are more critical as the generator speeds
at which the series resonances occur are within the typical 20% speed variation range
of a Unit Connection. For instance, the small 18th voltage harmonic is greatly magnified
at about 49 Hz fundamental frequency for this test system. The resonance frequencies
Dc filter impedance

15
--
a
C-
Q)
"0
10
:::::J
-
'c
0)
Cd

5

0 10 20 30 40 50
Harmonic order
(a)
Dc side impedance
20
15
--
a

--
Q)
"0
10
:::::J
+-'
'c
0)
Cd

5
____ ________ -L ________ ________ L_ ______
o 10 20 30 40 50
Harmonic order
(b)
Figure 6.21 (a) Harmonic impedance of the de filter. The fundamental frequency is 50 Hz. (b)
Harmonic impedances of a de system consisting of a smoothing inductor (0.8 H), dc filter and a line
inductance of 0.2 H.
Harmonic voltages at dc filter point
10
24th
8
-----
>

6
'-"
Q)
"0
::::I I
+-'
II
'c
II
0'1
4 II
ro
:2:
II
II
6th
, I
, I
, I
2 , I
18th
I 12th
ii
,
ii.
0
35 40 45 50 55 60 65
(a) Line inductance = 0.2 H
Harmonic voltages at dc filter point
10
I
II
II
II
II
8
II
II
II 24th
II
-----
II
>
II

6
'I
'-"
II
Q)
II
"0
"
::::I
...... II
6th
'c II
0'1
4
II
ro
II
:2: II
II
, I
I I
2
I I
12th
,
, \
0
'.
- - --
------
35 40 45 50 55 60 65
(b) Line inductance = 0.3 H
Figure 6.22 Harmonic voltages at the dc filter point for adjustable speed operation. a = 20.
6.8 EFFECTS OF DC SIDE RESONANCES
Generator Harmonic Currents
0.045,-------,------.-----,----,---------,----,
0.04
0.035
;- 0.03
0..
---- Q) 0.025
'0
:J
:g 0.02
OJ
<tS
:2: 0.015
0.01
0.005
11th
13th
23rd
.......... - --,/
5th
o '-'-'-'- -_.-.-._._._.-.-.-'-._._._._._.-.-.-
35 40 45 50 55
17th
60 65
113
Figure 6.23 Generator current harmonics at different generator speed in the presence of de side
resonances. Line impedance = 0.2 H, a = 20
0

of the dc impedances are also varied with the line impedance. Figure 6.22(b) shows the
harmonic voltages at the dc filter point for a line inductance of 0.3 H. The magnification
of the 6th harmonic voltage now occurs at generator speed of about 49.5 Hz. Thus
the conventional tuned dc filter becomes ineffective for the frequencies at which series
resonances on the convertor dc side occur.
Alternatively, other filtering schemes have been proposed such as active filters [Zhang
et aZ. 1993] [Cheuksun et aZ. 1989], pulse multiplication [Arrillaga et aZ. 1993a] maybe
more applicable for adjustable speed operation. Nevertheless, it has been demonstrated
that the developed Unit Connection model is a suitable tool for the investigation of dc
side resonances regardless of filtering schemes.
6.8.2 On ac Side Harmonics
For dc side resonances at harmonic frequencies, the magnification of dc side harmonic
currents at dc resonance frequencies is reflected as spikes on the ac harmonic currents
as shown in Figure 6.23. However since the magnitudes of the affected ac harmonic
currents are small, the impact of dc resonances on the ac harmonic levels is insignifi-
cant. In conventional HVdc schemes, a dc resonance at the fundamental frequency can
interact with an ac resonance at the second harmonic caused by ac filter capacitances
and ac system impedance. This could result in convertor transformer core saturation
instability [Chen et aZ. 1996]. However this is not a problem in direct connected units
since ac filters are absent.
6.9 CONCLUSIONS
By choosing the generator terminal voltage instead of the voltage E" behind generator
subtransient reactance as the firing angle reference, the firing control mechanism can
be simplified considerably since the generator currents are not required to be included
in the firing angle determination. Moreover the existing firing control for conventional
HV dc schemes, in which the reference is the generator terminal voltage, can be used for
direct connected units with practically no modification.
It has been shown that the rotor angle has a significant effect on the commutation
process in the presence of saliency and the close proximity of generators and conver-
tors. Substantial improvement in the accuracy of the harmonic prediction for direct
connected units has been obtained by taking into account the position of the rotor. An
accurate harmonic solution in the presence of generator saliency can only be obtained
by solving the non-linearities in a unified solution since all of the system variables are
inter-dependent. However it has been shown that the maximum harmonic current levels
in direct connected units containing salient-pole generators can be estimated by using
a non-salient generator with its subtransient reactances equal to ~ of the salient-pole
generator.
The harmonic loading on individual generator when connected in Group Connection
is the same as in Unit Connection. However when a Group Connection is being operated
with a reduced number of generators, the remaining generators are subjected to large
increases in harmonic loadings. This is due to a reduction in the per unit convertor
transformer leakage reactance and thus commutating reactance. The total generator
harmonic loading in terms of equivalent continuous negative-sequence current is the
largest in the unit connected mode and could exceed the normal 10% thermal design
rating. However, it has been shown that by keeping the firing angle under a certain limit,
which is dependent on power plant parameters, the harmonic loading can be kept under
the normal 10% design rating. The range for nominal firing angles in direct connected
units are also restricted by the overlap angle, generator terminal power factor, and mis-
firing constraints. Provided that the nominal firing angles are kept within the imposed
limits, practically no extra capacity for generator power factor and harmonic current
loading are required in direct connected units.
The generator harmonic content increases as the generator speed reduces. However
the increased harmonic loading is not a problem since the dc power order also reduces
with lower speed. The generator harmonic ratings can be taken at the nominal speed
and rated power output. In the adjustable speed operation, conventional tuned dc filters
become ineffective at frequencies where series resonances in the dc system occur, and
the harmonic levels on the dc side can be magnified to unacceptable levels at those
resonances frequencies. The effect of dc resonances on ac harmonic levels has been
found to be insignificant due to the small magnitudes of the affected ac harmonics.
Chapter 7
HARMONIC EVALUATION OF BENMORE
CONVERTOR STATION WHEN OPERATED AS A
DIRECT CONNECTED UNIT
7.1 INTRODUCTION
The Benmore convertor station of the New Zealand HVdc link occasionally operates as
an islanded or group connected system during the maintenance period of the intercon-
necting transformers, which link the convertor station and the rest of the ac system.
In 1993, a set of single-phase harmonic measurements of a direct connected HVdc con-
vertor scheme was collected [Macdonald et al. 1995]. While some of the results were
of considerable assistance to substantiate the claims of the direct connection, they did
not provide accurate harmonic information due to the limitations of the available, at
the time, monitoring techniques, i.e. the single phase nature of the test, lack of phase
referencing, manual triggering, etc.
Recent advances in harmonic monitoring, and in particular the availability of
CHART instrumentation [Continuous Harmonic Analysis in Real Time: Technical Ref-
erence 1996], encouraged 'Irans Power New Zealand Limited to repeat the earlier field
test to obtain more reliable information. CHART's improved dynamic range and contin-
uous updating of the fundamental frequency improves the detection of harmonics with
small magnitudes, especially the non-characteristic harmonics. Its accurate GPS (Global
Positioning System) based time stamping capability and synchronization combined with
multi-channel and multi-unit configuration permits the computation of the harmonic
phase angles and the signal sequence components at different locations.
This chapter describes the measurement procedures, the setup and configuration of
the CHART instrumentation, and the results. The results include the measured wave-
forms and their harmonic contents in sequence components for a range of dc line current.
The harmonic ratings for generators in a Group Connection are also investigated. The
measured and calculated firing angles described in this chapter are referenced to the gen-
erator terminal voltage since the internal voltage E" is inaccessible for measurements.
The test results are also used to assess the accuracy of the developed harmonic domain
Unit Connection models.
116CHAPTER 7 HARMONIC EVALUATION OF BENMORE CONVERTOR STATION WHEN OPERATED AS A DIREC
To South
Island
Network
16 kV bus
G2
'VI---+----l
Pole 1A
T3
To the North Island
O.SH

.
1M3
,
,
,
GROUP CONNECTED HYDC SYSTEM
Figure 7.1 Benmore group connection and CHART measurement points.
7.2 TESTING PROCEDURE
By arrangement with Trans Power New Zealand Limited, the Benmore convertor station
became available for harmonic tests in group connected mode during a recent mainte-
nance period. The test system, shown in Figure 7.1, involves four generators (G3 to G6)
and the convertor group connected to pole-lB. Three-phase measurements were made
of the:
16 kV bus voltages
Generator G5 currents
Convertor transformer T4 and T6 currents
As one of the objectives of the measurements was to provide information for pos-
sible validation of the generator harmonic model, three-phase currents of the generator
7.3 FUNDAMENTAL FREQUENCY VARIATION 117
G5 were recorded. Single-phase measurements of the red-phase currents were made at
generators G4 and G6. Pole-IB dc line voltage, current and the firing angles of the valve
groups were also recorded.
Due to the geographical separation of the measurement points, two CHART units
were used. One unit was placed in the pole-l dc control room and the other in the
power station relay room which are over 200 meters apart. The CHART units were
synchronized via GPS satellite, and the red phase 16kV bus voltage was chosen as the
harmonic phase angle reference for both units.
CHART adjusts the sampling rate when the fundamental frequency drifts away from
its previously measured value so as to maintain exactly the specified number of sampling
pulses within a single fundamental cycle throughout the measurement. CHART was
programmed to sample and store the signals every three minutes. This would produce
several sets of reliable data for each dc current setting, while allowing time for the system
to stabilize.
Two types of data were gathered from the waveforms: 256 cycles of 128 sample points
and 8 cycles of 1024 sample points. The longer duration intervals provide information
on possible sub-synchronous swings and helps determining if the system is stabilized.
The higher sampling rate enables the switching instants of the convertor to be resolved
accurately.
The pole-IB mercury-arc valve convertor was deblocked and operated as a group
connection in the constant current control mode. The dc current was varied from 750
to 1190 amps in approximately 50 amp steps over a four hour period.
7.3 FUNDAMENTAL FREQUENCY VARIATION
The fundamental frequency, computed and recorded for each set of data, varied from
49.0 to 51.2 Hz during the test, as shown in Figure 7.2. The system load restriction
required that the dc current be decreased from 1190 to 1016 amps, and then increased
from 750 to 950 amps. Due to over compensation of the control system action, the
machine speed was faster at heavier loads.
7.4 THREE-PHASE HARMONIC INFORMATION
Figure 7.3 and 7.4 display the 16 kV bus voltages and generator G5 currents as well as
their harmonics in sequence components for a dc current setting of 1016 amps (270 MW).
The high sampling rate of 51.2 kHz reveals the high frequency ringings of approximately
10 to 15 kHz at the start of commutation periods in the voltage waveforms. This
indicates that the total stray capacitance in the circuit is approximately 0.14 to 0.2 nF,
given that the smoothing inductance is 0.8 H.
Fundamental frequency
51.5,...-----,--.....,----,----,--.........::....,..-'--,---....------,
51
. .
--. . . . .. ...........,..
50.5
50 ..... ..
950A
850A
49.5
49
750A
10 20 30 40 50 60 70 80
Data Set
Figure 7.2 Fundamental frequency variation during the test measurement.
The measurement results confirm the rotational sequences of the characteristic har-
monics of 12-pulse HV dc operation, ie. the (12n-1 )th harmonics are of negative-sequence
and the (12n+1)th harmonics are of positive-sequence. Small amounts of the 6-pulse
convertor characteristic harmonics, such as the 5th, 7th, etc., caused by slight imbalances
among the two 6-pulse valve groups, were also detected.
The measurements show a modest amount of third and smaller amounts of higher
triplen zero-sequence voltage harmonics. These must be attributed to generator satu-
ration and space harmonics because the convertor transformers are delta-connected on
the generators' side. Thus there is practically no zero-sequence harmonic current path
in the group connection configuration. These zero-sequence voltages due to generator
saturation and space harmonics would also exist in the conventional operating mode as
the interconnecting transformers (T5 and T2) are delta connected onto the generator
bus, as shown in Figure 7.1. Figure 7.5 illustrates the variation of the zero sequence
voltage harmonics with respect to the 16 kV bus fundamental voltage. In terms of the
equipment insulation requirement, the voltage levels are quite small however, and would
have little influence on the insulation specification.
On the other hand, the harmonic currents are more serious as they create extra
losses in the generator stators. Thus it was decided to obtain the variation of harmonic
currents for a range of dc line current. In addition, the availability of multiple data sets
should help with the comparison between measurement and simulation results.
7.4 THREE-PHASE HARMONIC INFORMATION
(!J
I
.c
CL
>
.::s:.
..-...
:::s
ci.
--
(!)
"'0
:::s
.......
c
0)
ctl

16kV bus voltages
15,----,----,----,----,----,----,----,----,-----,---,
.'\. :(\
+... ,'Vi". .... . \'"
,': \1)" - , , , - , ,
10 '\'
, \,\ : : \,
"'Y" .......... ..;: .... ;;. ....
, /''''\f,l' : \ I ....
,. \. : \' I :
5 ..... \

. , ,I
/' ......... , ..... ,.,.,.'., ... , ";.-1:"
,./.,r' \ V\!., '
Vi
o
-5
-10
: ,
\
\
, :,
... .1.. . ........... .' .... 1 .. .' ....... : .. \ ..
. I
\, ;/
\ IJ: I:
r\'
:'-, ,. r
.,.,., .. : .. ,/
I
/\ I .
:/ I
...... ........ ................. : ....
j.... I":'
,: / :
R.\. / /'
....... '. \... : hoi

i j
:,Ij
:

o 0.002 0.004 0.006 0.008 0,01 0.012 0.014 0.016 0.018 0.02
Time (seconds)
16 kV bus voltage harmonics
0.06n:r--,--,---,---,---,-----,---,--,-----,--,
0.05
0.04
0.03
0.02
0,01
...... , .................. [1] . pcisliivesequehces'
sequ$nces :
. , , .
I ill m: n I. I : fl] I

5 10 15 20 25 30 35 40 45 50
Harmonic order
119
Figure 7.3 Three-phase generator bus voltages and their harmonics in sequence components for a dc
current of 1016 amps.
Q)
Generator G5 currents
4,----,----,----,----,----,----,----,----,----,----,
,
I
3 .. \
2
o
\
\
... \
\,;"':
\:
.r...(, .... '.".
/. ".
/:
;" \.
.', ...... -r-; ..... ,', ......... :-\ .. , ... :., ... ,( .. :
I : I: 1\ t
:.. / :..... 1:;/
, , I \.
i'-' ,., .... ,',.,. ,., ..... ,. ,:.,
.. .. 1. ,',\.\
, . .
,,,,,':i
j
,<
: I
,j
I
" ,:
I :
T
... \
\ :
\.:
.. \
:\,-.
I

o 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Time (seconds)
Generator G5 current harmonics
0.06 rrr-----,----,----,----,-----,-----,----,-----.-----,-------,
0.05
0.04
Positive sequences
sequEilnces
"'0 sequences
-g 0.03
......
'c
0')
co
0.02
0.01
o I", n_1iI
1 5 10 15
",'

.'
I III n
20 25 30 35 40 45 50
Harmonic order
Figure 7.4 Three-phase generator currents and their harmonics in sequence components for a dc
current of 1016 amps.
7.5 HARMONIC VARIATION WITH DC CURRENT ORDER
0.04
0.035
0.03
--:-
::J
0.025
S
Q)
"0
0.02
::J
.....
c
Ol
0.015 co

0.01
0.005
0
0.995
Zero-sequence 16 kV bus voltage harmonics
- --
._ . .....;._.-,_.-.
1.005
.,' h<irrn()niGj,
9th
, ,
"15th'harmoniC"
--:--
...... ..
-'-'-'-'-i-'-'-'-'_'_':_'_'_'_
1.01 1.015
Fundamental 16 kV voltage (p.u.)
121
1.02
Figure 7.5 Variation of the zero-sequence voltage harmonics with respect to the fundamental 16 kV
bus voltage.
7.5 HARMONIC VARIATION WITH DC CURRENT ORDER
7.5.1 DC signals
The dc voltage and current are required for the calculation of the total dc power output,
and for the intervalidation with simulation results.
The dc voltage was obtained through a filtered channel, thus containing unreliable
harmonic information. However if the dc current harmonics are obtained, the dc voltage
harmonics are not required. The voltage level was found to be almost constant at
approximately 266 kV for various loads, as shown in Figure 7.6. This is to be expected
since the firing angle and the ac terminal voltage magnitude are kept within a narrow
band, while the invertor is operated in a minimum gamma control mode, and the line
resistance is small.
As dc current harmonics influence the commutation and firing process, it was de-
sirable to obtain the DC current waveform to confirm that the harmonics were small as
anticipated. This is clearly the case as illustrated in Figure 7.7 for a dc current order
of 1016 Amps. The presence of the 6th, 18th, 30th, etc. harmonics also indicates that
there is a slight unbalanced between the two six-pulse valve groups.
122CHAPTER 7 HARMONIC EVALUATION OF BENMORE CONVERTOR STATION WHEN OPERATED AS A DIRE(
DC Voltage

265
.............. ' ........... -
750 805 855 1150 1190
DC Current (A)
Figure 7.6 DC voltage level for various de current orders.
DC Current
1050,------------,------------,------------.-----------.
C/)
0.
E 1025'

o 0.005 0.01 0.015 0.02
Time (seconds)
Harmonics for DC current of 1 016 Amps
5,,---.-----.----,-----,----,-----.-----,----,-----,----,
C/)
Q.
E 2.5

Harmonic order
Figure 7.7 DC current and harmonics for the current order of 1016 Amps.
7.6 GENERATOR HARMONIC RATINGS 123
7.5.2 AC harmonics
Since the ideal sinusoidal commutation voltage E" behind generator subtransient re-
actance is not accessible for measurements, the firing angles have been derived from
the crossings of fundamental components of the measured generator terminal voltages.
Figure 7.8 shows the derived firing angles, C, commutation angles, f.l, and the gener-
ator terminal fundamental power factor angles for the tested range of the dc current
in pole-lB. It should be noted that the firing angle derived from the crossings of the
ideal commutation voltage E" behind subtransient reactance would be larger than those
shown since the generator terminal voltage is lagging behind the ideal commutation
voltage. Generator terminal voltage waveforms obtained from the test measurement
and harmonic domain solution are compared in Figure 7.9 for 11 firing angle and dc
current order of 750 amps. Close agreement in the waveforms is obtained as the fir-
ing angle is derived from the generator terminal voltages in both the measurement and
unified iterative solution.
Generator current and voltage harmonic variation with respect to dc current order
was obtained using the developed Unit Connection model and the firing angle derived
from measured waveforms. Relevant information used in the system models is given
in Appendix A2. The recorded and simulated harmonic trends, shown in Figure 7.10
and 7.11, are consistent and reasonably close, particularly the 11th and 13th current
harmonics, considering the uncertainties in generator parameters and the tolerance of
the voltage and current transformers used.
7.6 GENERATOR HARMONIC RATINGS
The possibility of a direct connection was envisaged in the design of the Benmore gen-
erators although restrictions were imposed on the nominal power rating (lowered from
112.5 to 102.3 MVA) and on the current harmonic content under such operating con-
ditions. The design ratings as well as the recorded current harmonics at the maximum
power output during the test measurements, which include the 1993 single-phase mea-
surement, are listed in Table 7.1. The table shows that none of the design ratings has
been exceeded in group connected operation so far. It is observed that the harmonic
current levels, particularly the 11th and 13th, were higher when three instead of four
generators were used. This increase in harmonic currents is consistent with the Group
Connection analysis in Chapter 6, which showed that when a Group Connection is being
operated with a reduced number of generators, the remaining generators are subjected
to an increase in harmonic loadings.
The Benmore convertor station has not been operated in the unit connected mode.
This is partly due to the lack of accurate understanding of the harmonic variation in this
operation mode. Using the developed model, harmonic variations in the unit connected
en
Q)
25
20
~
OJ 15
- - - Power factor angles
. -. -. - Commutation angles
-- Firing angles
Q)
o
10 ------------------------------
--
5L-__ __ _L __ __ L _ __ __ __ _L __ ____ _ ~
0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15
DC current (kA)
Figure 7.8 Variation of the measured generator terminal power factor angle (lagging), average firing
angles (referenced to the measured generator terminal voltage) and commutation angles for a range of
dc current.
16 kV bus voltage from measurement and modelling
15 ...
10
5
(9
':
I
..c
D... 0 ............... .
~
~ :
-5
-10 ............... -.......................... .
_ 1 5 L L ~ ~ ~
o 90 180
Electrical degrees
270 360
Figure 7.9 Comparison of generator terminal voltage waveforms obtained from the test measure-
ment and harmonic domain solution. Recorded data = solid, harmonic domain solution = dashed,- both
reconstructed from 200 harmonics.
7.6 GENERATOR HARMONIC RATINGS
Generator 11 th harmonic current
0.04,.--,.--,.--,.--,.--,---,---,.--,---,----,
0.035
0.03
::; 0,025
.e,
0.01
0,005
DC current (kA)
Generator 23rd harmonic current
0.04,.--,.--,.--,.--,---,---,---,---,---,----,
0,035
0.03
::; 0.025
.e,
0.02
0,015
0,01
0,005
DC current (kA)
Generator 35th harmonic current
0.04,---,---,.--,---,---,---,---,---,---,----,
0,035
0.03
::j 0.025
.e,
~
~
c
0.02
g> 0.015
::z:
0,01
0,005
DC current (kA)
125
DC current (kA)
0.04,.--,.--,.--,.--,.--,.--,.--,.--,.--,.-----,
0.035
0,03
::i 0,025
.e,
CD
-g 0.02
~
g> 0.015
::z:
0.01
0,005
DC current (kA)
0.04,---,---,.--,---,---,---,---,---,---,.---,
0.035
0.03
::j 0.025
.e,
CD
~ 0,02
c
g> 0.015
::z:
0,01
0.005
DC current (kA)
Figure 7.10 Variation ofthe generator current harmonics for a de current range; Recorded data: solid,
harmonic domain: dash-dot.
0.08
0.'07
0,06
:;;
0,05
8,
<ll
"0
0.04
.a
'c
OJ
<U 0,03
::;;;
0.02
0.01
0
0.7 0.75
0.06
0.07
0,06
:;;
0.05
8,
<ll
"0
0,04
1
c
OJ
<U 0.03
::;;;
0,02
0,01
0
0.7 0.75
0,08
0,07
0.08
:;;
0.05
8,
<ll
"0
0.04 .
:E
c
OJ
<U 0,03
::;;;
0,02
0.01
0
0.7
Generator 11 th harmonic voltage
:;;
8,
<ll
"0

C
OJ
<U
::;;;
1
DC current (kA)
Generator 23rd harmonic voltage
0.8 0.85 0.9 0.95 1
DC current (kA)
Generator 35th harmonic voltage
0.08
0,07
0.06
0,05
0.04
0.03
0.02
0.01
0
0.7 0,85 0.9 0.95 1 1.05 1.15
DC current (kA)
0L-__ __ __ __ __ -L __ -L __ __
0.7 0.8 0,85 0.9 0,95 1 1.06 1,15
DC current (kA)
0.08r---r--,---,---,----,--.,-----,---,----,---,

DC current (kA)
Fi'g'l.ire-'Z.lT-- Variation of the generator voltage harmonics for a dc current range; Recorded data: solid,
harmonic domain: dash-dot.
Generator Design Recorded harmonic currents (p.u.)
harmonic ratings 4 generators 3 generators*
current (p.u.) (@316 MW) (@212 MW)
11th 0.0744 0.0356 0.0435
13th 0.0595 0.0204 0.0298
23rd 0.0193 0.0127 0.0084
25th 0.0154 0.0078 0.0049
Table 7.1 Generator harmonic ratings for different group connected configuration at the Benmore
convertor station; 'results from the 1993 measurement.
Gen- System base X"d X"
q Xleakage
P
dc
erators . (MVA) (p.u.) (p.u.) (p.u.) (MW)
4 409.2 0.1582 0.1728 0.112 360
3 409.2 0.2110 0.2304 0.112 270
2 409.2 0.3164 0.3455 0.112 180
1 409.2 0.6328 0.6911 0.112 90
k 4 X 102.3 tWH X 0.174
4102.3 0 19
k ~ .
0.112 k X 90
Table 7.2 Adjustment of parameters for the Benmore system operating in group connected and unit
connected mode. Generator rated power = 102.3 MVA, transformer rated power = 187.5 MVA; k is the
number of generators in the group connected mode.
mode can be examined. Simulations were carried out for group connected systems
containing 4, 3, 2 or 1 generator. The generators are operated at rated power of 90MW
at 50 Hz in all systems. Since the system parameters are changed due to the reduction
in the generator power rating for direct connected operation, the modified parameters
are listed in Table 7.2.
Figure 7.12(a) shows the variation of commutation angles and generator terminal
power factor with respect to firing angles and the number of generating units. Firing
angles have been measured from the generator terminal voltages so that they are consis-
tent with the measurement results described in Section 7.5.2. The commutation angles
are typically below the 28 limit provided that the firing angle referenced to V is greater
than 2 for the group connected mode with 4 generators. The generator terminal power
factors are also well above the design rating of 0.88. Firing angles referenced to E" can
be derived from the relationship between E" referenced and generator terminal voltage
V referenced firing angles, as shown in Figure 7.12(b). The phase angle between E"
and V only increases slightly with fewer generators since the reduction in power factor
angles is small.
The harmonic current variation with respect to the firing angle as well as individual
128CHAPTER 7 HARMONIC EVALUATION OF BENMORE CONVERTOR STATION WHEN OPERATED AS ADIREC'
Commutation Angles
30,----,-----.----.---.11-=--:-28"'0,---.-------,

0,95
3
20 2
0,9
15
0.85
'0
O.B
1 gen unit
4
Design rating
,
,
,
,
,
,
,
,
,
,
0L----':----:',O,----'Cc:'---=20,-----:27' ---:30 0.75
0
=-------:-----:'':-0 ---'=S----:':20:------:2':-S ---:'30
Firing Angles (Degrees) Firing Angles (degrees)
Q)
u
c

Q)
0>
:m
o
>
Lu
I
81
OJ
!i
0>
C
IT:
(a)

Firing Angles - Terminal Voltage Reference
(b)
Figure 7.12 (a) Commutation angle and generator terminal power factor at the rated power output
for the Benmore Convertor station operating as group connected units; (b) Relationship between E"
referenced and generator terminal voltage V referenced firing angles.
Generator 11 th Harmonic Current Generator 13th Harmonic Current
0.1 0.1
0.08
Desi n ratin
0.08
:::i :::i
Design rating
S
0.06
S
0.06
OJ OJ
'0 '0
.a
~
'c ", '2 c
0> 0.04 0> 0.04
'" '" :::E
3
:::E
_ ' ' -- 4 e n units
''''"2
3
,
,
,
0,02 0.02
,
,
4 gen units
0 0
0 10 15 20 25 30 0 10 15 20 25 30
Generator 23rd Harmonic Current Generator 25th Harmonic Current
0.04 0.04
0.03 0.03
:::i :::i
S S
OJ
Design rating 1 gen unit
OJ
1 gen unit
'0
0,02
'0
0.02
.-E
::::J
c
:g
Design rating
0> 0>
'"
;/
/
'"
:::E :::E
?/
/
/ /
0.01 /
/
0.01
/
:---.... -
/
/
/
/
-
/
3
/
/
-.-:-
-
,
/ ~ /
-
,
,
-
- ---_ ....
--'!_- - __ 4.,'
0 0
0 10 15 20 25 30 0 10 15 20 25 30
0.12 0.3
0.1 0.25
:::i
0.08
:::i
0.2
S S
OJ OJ
'0
0.06
'0
0.15
.a
" /'2
. ~
'c C
0> 0>
,,"2
'" '" :::E 0.04 3
:::E 0.1
,
, __ ' ' 4 gen units
3,,' -
,
4 gen units
0.02 0.05
0 0
0 10 15 20 25 30 0 10 16 20 25 30
Figure 7.13 Generator harmonic current and voltage levels for the Benmore Convertor station oper-
ating as group connected units. Firing angles are referenced to the generator terminal voltage.
130CHAPTER 7 HARMONIC EVALUATION OF BENMORE CONVERTOR STATION WHEN OPERATED AS A DIRE(
0.2.------,-----.---,----,-------,-----,
0.18
0.16
Design rating
0.14
--:-
::J
E::
0.12
CD
-0
0.1
::J
:!::!
C
0)
0.08
co
:2:
0.06
0.04
- 4 gen units
0.02
0
0 5 10 15 20 25 30
Figure 7.14 Generator equivalent negative-sequence currents for the Benmore Convertor station op-
erating as group connected units.
Transformer Current THD
0.3
0.25
---
---
4 gen units
--:-
0.2
::J
E-
CD
-0
0.15
::J
:!::!
C
2
0)
co
:2: 0.1
0.05
0
0 5 10 15 20 25 30
Figure 7.15 Convertor transformer current THD for the Benmore Convertor station operating as
group connected units. 'ITansformer power base is 187.5 MVA.
7.7 CONCLUSIONS 131
harmonic ratings are plotted in Figure 7.13. Harmonic current levels in the remaining
generators increase drastically, as much as 0.01 p.u. in the 11th and 13th harmonic
currents, for each generator taken out of service. For a nominal 15 firing angle, harmonic
currents do not exceed the design ratings in any case, although the current harmonic will
be close to the limit in the unit connected mode. It should be noted that the harmonic
design ratings for Benmore generators are higher than those in typical hydro generators
since the 15.6% rated equivalent continuous negative-sequence current harmonic loading
of Benmore generators is much higher than the typical 10%, as shown in Figure 7.14.
On the other hand, the convertor transformer harmonic current loading is at the
largest when all generators are in service, as shown in Figure 7.15.
7.7 CONCLUSIONS
Detailed three-phase harmonic information of a group connected generator-HVDC con-
vertor scheme has been collected and discussed. The use of CHART improves the de-
tection of harmonics with small magnitudes, and provides confirmation of the harmonic
rotational sequences. The high sampling rate of 51.2 kHz enabled accurate derivation of
the actual firing instants which has been used in the simulation to improve the harmonic
prediction. The results have been used to intervalidate the developed Unit Connection in
the harmonic domain model, and close agreements between the measured and simulated
results were achieved, particularly the 11th and 13th current harmonics which are the
largest and most significant. Besides the validation purpose, the measurement results
also provide valuable evidence of the feasibility of the direct connection.
The harmonic current and voltage variation with respect to the number of genera-
tors for the Benmore terminal when being operated in group and unit connected mode
have been examined. It has been found from the measurement and simulation that the
harmonic current levels in the generators increase substantially, as much as 0.01 p.u.
in the 11th and 13th harmonic currents, for each generator taken out of service. For a
nominal 15 firing angle, generator harmonic current levels would be below the specified
maximum design ratings in all cases, but these come very close to the limits in the unit
connected mode. Although the Benmore convertor station has not been operated in the
unit connected mode, the simulation results have provided solid evidence of the viability
of this mode of operation.
Chapter 8
CONCLUSIONS AND FUTURE WORK
8.1 CONCLUSIONS
The direct connection of generators to HV dc convertors is a flexible and attractive option
in which ac harmonic filters, ac switching and protection gear, and a transformation
step are dispensed, leading to a more compact HV dc Power Station with substantial
savings in capital and operating costs. However the absence of ac filters and the close
proximity of generators and convertors lead to strong harmonic interactions in these
schemes. Simplified convertor analysis has been found to be inadequate to determine
such interactions. Dynamic simulation of the direct connected units to the steady state
has been the only acceptable method of harmonic prediction so far. But this is an
inefficient approach considering the long time constants of the synchronous generator.
The primary objectives of the research described in this thesis were to develop
an accurate and efficient harmonic model of direct connected units as a suitable tool
for analysing the harmonic interactions, and to thoroughly investigate the harmonic
behaviour of these schemes.
Due to the lack of accurate generator models in the harmonic domain, a synchro-
nous generator model in the form of a harmonically coupled impedance matrix has been
developed in Chapter 3. The model takes into account the effects of saturation and
the stator-rotor frequency conversion phenomenon due to saliency. The relationship
between the generator terminal voltages and currents in the presence of saliency has
been expressed in sequence components. The analytical expressions are derived from
normally available machine parameters such as direct and quadrature-axis subtransient
reactances. Harmonics produced by generator saturation have been shown to be in-
significant. The generator harmonic model has been validated against time domain
simulations and results from harmonic measurements of the Benmore convertor station
when being operated as a group connected unit. The developed analytical model is thus
a powerful tool for fast and accurate predictions of the harmonic levels on the generator
terminals in the presence of saliency, saturation and distortion.
In the absence of ac harmonic filters at the generator terminal, the ideal sinusoidal
commutation voltage moves inside the generator and the commutating reactance consists
134 CHAPTER 8 CONCLUSIONS AND FUTURE WORK
of the transformer leakage reactance and generator subtransient reactance. Since gener-
ator subtransient reactance varies with rotor angle due to saliency, an iterative solution
for the commutation process is necessary. By choosing the generator terminal voltages
as pseudo-commutation voltages, which can be unbalanced or contain harmonic distor-
tion, the effect of the varying generator subtransient reactance and transformer leakage
reactance can be considered separately in the iterative solution. Given the generator
terminal voltage as a pseudo-commutation voltage, the convertor harmonic currents can
be obtained from the commutation analysis described in Section 4.2, in which the re-
actances are solely the transformer leakage reactances. The generator terminal voltage
harmonics are then updated by injecting the convertor currents into the generator har-
monic impedance. Convolution in the harmonic domain is employed to construct the dc
voltage and ac currents in terms of the switching angles, ac voltage harmonics and dc
current harmonics. Thus the analysis is completely in the harmonic domain.
The harmonic interaction in direct connected units has been described using a set
of nonlinear equations. The electrical variables have been chosen to represent physically
obtainable signals, such as ac terminal voltages and dc current harmonics, to predict the
relationship between these signals. Voltage control, exercised entirely by the generator
excitation in the absence of convertor transformer on-load tap-changers, has also been
incorporated into the formulation. By decomposing the mixture of real and complex
variables into real quantities, the Newton solution can be formulated using real values in
positive harmonics only. This enables the harmonic interaction between generators, the
convertor switching instants and the dc system to be combined into a unified iterative
solution. By exploiting the sparse nature of the Jacobian matrix, the solution has been
solved rapidly.
It has been shown that the convergence of the unified solution is greatly improved
in comparison with sequential solution methods such as Iterative Harmonic Analysis
for such systems. Divergence of Newton's method can occur when the gradient of the
iterating function, represented by the Jacobian of partial derivatives, exceeds a certain
value. Rigorous analysis has showed that the size of the gradient is directly proportional
to the generator subtransient reactance, and inversely proportional to the commutating
reactance. For small values of commutating reactance and large subtransient reactances,
the magnitudes of the Jacobian elements are large, leading to divergence. Nevertheless,
the convergence threshold of the proposed model is small enough and no modification is
needed for harmonic analysis of the direct connected units with typical reactance values.
The developed Unit Connection model also serves to intervalidate the time domain
simulation, until now the only alternative method of harmonic analysis for direct con-
nected units. Close agreement in ac and dc harmonics has been obtained when a small
value of the snubber capacitance value is used in the time domain simulations. The large
artificial snubber capacitance normally used in time domain simulations to aid numerical
stability does have significant influence on the harmonic levels.
8.2 SUGGESTIONS FOR FUTURE WORK 135
By choosing the generator terminal voltage as the firing angle reference, existing
firing control for conventional HV dc schemes can be used for direct connected units with
practically no modification. Provided that the nominal firing angles are kept within
the restriction limits, imposed by the overlap angle, generator terminal power factor,
continuous equivalent negative-sequence harmonic loading, number of generators and
mis-firing constraints, practically no extra capacity for generator terminal power factor
and harmonic current loading are required in direct connected units. The generator
harmonic ratings can be taken at the nominal speed and rated power output.
The effect of rotor angle on the commutation process in the presence of saliency
has been shown to be significant in direct connected units. Substantial improvement in
the accuracy of the harmonic prediction has been obtained by taking into account the
position of the rotor. An accurate harmonic solution in the presence of generator saliency
can only be obtained by solving the non-linearities in a unified solution since all of the
system variables are inter-dependent. However the harmonic current ratings in direct
connected units containing salient-pole generators can be estimated by using a non-
salient generator with its subtransient reactances equal to ~ of the salient generator.
Detailed three-phase harmonic information of the Benmore convertor station when
being operated in the group connected mode was collected and discussed. The test
results have provided valuable evidence of the feasibility of direct connection units, and
have also been used to assess the accuracy of the harmonic domain models and dynamic
simulation techniques. The recorded and simulated harmonic trends are consistent and
reasonably close considering the uncertainties in generator parameters, and the tolerance
of the voltage and current transformers used. Although the Benmore convertor station
has not been operated in the unit connected mode, the simulation results have shown
that the harmonic currents would not exceed the design ratings and thus provided solid
evidence of the viability of this mode of operation at the station.
8.2 SUGGESTIONS FOR FUTURE WORK
8.2.1 Transformer Saturation Modelling
A transformer model in the Harmonic Domain which takes into account the effect of
saturation has been developed [Lisboa 1996]. In order to incorporate this model into the
unified solution, three additional variables describing the dc component of magnetising
currents are required. The harmonic components of magnetising currents or voltages
in each phase can be calculated at each iteration using an FFT, and a curve fitting
technique such as cubic splines can be used to represent saturation curves.
136 CHAPTER 8 CONCLUSIONS AND FUTURE WORK
8.2.2 Subsynchronous Resonances and Interharmonic Studies
The Harmonic Domain can be extended to include frequencies other than harmonic
frequencies to examine possible sub synchronous resonances and interharmonic phenom-
ena. The equations describing machine rotor and mechanical characteristics can be
incorporated into a unified solution to examine the combined electrical and mechanical
interaction. Changes in the machine inductances, particularly at sub synchronous fre-
quencies, can be readily taken into account at each interested frequency. Moreover, the
unified solution method can be used to obtain discrete frequency response characteristics
of direct connected units which can then be used to aid the design of the control system.
In the case of HVdc back-to-back connection for systems that have different power
frequencies, the Harmonic Domain can be extended to take into account the frequencies
resulting from the interaction of all relevant frequencies. As a result, the unified solu-
tion method is a powerful tool for analysing not only the harmonic behaviour but also
subsynchronous and inter harmonic behaviour of HVdc systems.
Appendix A
TEST SYSTEMS DATA
A.I HWANG'S GENERATOR PARAMETERS
The generator parameters are extracted from a paper by Hwang [1969].
Rating 100 MVA
Voltage 13.8 kV
Xd
1.2 pu
I
0.3667
xd
pu
"
0.2
xd
pu
Xq
0.8 pu
"
0.367
Xq
pu
I
Tdo
7.64 seconds
T"
do
0.0633 seconds
T"
qo
0.132 seconds
Ra
0.005 pu
Xo
0.05 pu
A.2 BENMORE'S GENERATORS AND TRANSFORMERS
Convertor Transformers
Type
Rating
Voltage
Leakage
Generators
Rating
Voltage
Delta-Wye and Delta-Delta
187.5 MVA
16-110 kV
11.2 %
112.5 MVA
16.0 kV
138 APPENDIX A TEST SYSTEMS DATA
Xd 1.168 pu
I
0.264
xd
pu
"
0.174
xd
pu
Xq 0.672 pu
"
0.19
Xq
pu
I
Tdo
8.7 seconds
T"
do
0.087 seconds
T"
qo
0.132 seconds
Xo
0.0765 pu
Ra
0.0042 pu
For a twelve-pulse unit connection operation, the generator rating is reduced to
102.3 MVA from 112.5 MVA. The generator harmonic current ratings (based on the
102.3 MVA rating) are given as:
Generator harmonic currents Design ratings
11th 0.0744
13th 0.0595
23rd 0.0193
25th 0.0154
terminal power factor 0.88
Table A.I Generator harmonic ratings for twelve-pulse group connected operation at the Benmore
convertor station.
Generator open circuit saturation curve is given in Figure A.l, which can also be
defined by the set of points which is given in Table A.2.
Field current (A) Terminal line voltage (k V rms)
0 0
100 2.7
200 5.4
300 8.0
400 10.6
500 12.95
600 14.85
700 16.27
800 17.30
900 18.15
1000 18.85
1100 19.40
1200 19.80
Table A.2 Benmore generator open circuit saturation characteristic.
A.2 BENMORE'S GENERATORS AND TRANSFORMERS 139
/
C/.NI\DIAN GI!NF.RI\L I!:LECTRIC co .. LIMITI!:D /' .
,:.: I
:::,::; .
. ,
I:t
:, '1 IJ !
i! '
, ;;r,
@
... . ,D:
JH
.. .:t4 :':f:rfll em:
11
ffi*

111111111111 i: iii i ill!;;::'
. ,- I .1
"
..
,.,.,-, 'I
Figure A.1 Benmore generator open circuit saturation curve.
Appendix B
SALIENT-POLE GENERATOR HARMONIC
IMPEDANCE CALCULATION FROM MEASURED
HARMONIC VOLTAGES AND CURRENTS
This appendix describes the derivation of the self and cross coupling terms of the salient-
pole generator impedance from the terminal voltages and currents measured in a HV dc
group connected operation. The voltage/current relationship of a characteristic har-
monic pair 12n 1 in the presence of saliency can be expressed by the following complex
matrix equation:
(B.1)
Substituting the generator harmonic impedance expressions developed in Chapter 3
yields:
[
Vk ]_ [ R + + - ] [ Ik ]
TT - .k+2(X" -X") j28 R . k+2 (X" X") I . (B.2)
v k+2 J 2 q d e + J -2- d + q k+2
Equation B.2 is decomposed into four real equations which can then be solved for
the four unknown real variables R, and 8. Decomposing equation B.2 yields:
R{Vd
A -B C -D
R{h}
I{Vk} B A D C I{Ik}
(B.3)
R{Vk+2}
_k+
2
C _k+
2
D
A
_k+
2
B
R{h+2}
k k k
I{Vk+2}
k+
2
D
k
_k+
2
C
k
k+
2
B
k
A
I{h+2}
142APPENDIX B SALIENT-POLE GENERATOR HARMONIC IMPEDANCE CALCULATION FROM MEASURED HAl
where
A
B
c
D
Rearranging equation B.3 yields:
n{Vk} n{h} -I{Ik}
I{Vk} I{h} n{h}
n{Vk+2} n{Ik+2}
-kt2I{Ik+2}
I{Vk+2} I{h+2} kt
2n
{h+2}
(B.4)
n{Ik+2} -I{h+2}
A
I{h+2} n{Ik+2} B
_kt
2n
{h} _kt
2I
{h} C
-kt2I{Id
kt
2n
{h} D
(B.5)
Given the harmonic voltages and currents, the variables A, B, C, D can be found by
using equation B.5, subsequently the unknown variables R, 8 can be obtained
from equation B.4, i.e.
R A

=
B _
k k
-
2B
k
8
1 . -1 { 2C }
"2 sm -
(B.6)
The self and cross-coupling terms of the generator impedance for each pair of char-
acteristic harmonics are then obtained using the impedance expressions in equation B.2
Appendix C
CUBIC SPLINE INTERPOLATION
Cubic spline interpolation has been well documented in literature [Gerald and
Wheatly 1985] [Press et al. 1992]. This appendix summarizes the cubic spline inter-
polation formulation with particular application to represent generator or transformer
magnet ising characteristics.
Given a tabulated function Yi = f(Xi), i = 1,2, ... n, the cubic for the ith interval,
which lies between the points (Xi, Yi) and (Xi+1' Yi+1) is in the form
(C.1)
The slope and curvature, i.e. the first and second derivatives, are required so that
the cubics can be fit smoothly from one interval to another. Let Si and Si+1 represent
the second derivatives at the point (Xi, Yi) and (Xi+1' Yi+1) respectively, expressions for
the coefficients in equation C.1 can be obtained as:
Ai
Si+1 - Si
6(Xi+1 - Xi)
Bi
Si
=
-
2
Ci
Yi+1 - Yi
Xi+1 - Xi
Di
Yi
(Xi+1 - Xi)
2Si + Si+1
(C.2)
(C.3)
(C.4)
(C.5)
As the slopes of the two cubics that join at (Xi, Yi) are the same, the relationship
among the cubics is given by:
(C.6)
where hi = Xi+1 -Xi. As equation C.6 represents a relationship among three consecutive
points, it can only be applied to each internal point, from i = 2 to i = n - 1. This gives
n - 2 equations relating the n values of Si' For a unique solution, two further equations
involving Sl and Sn are needed, and can be obtained when conditions at Xl and Xn
144 APPENDIX C CUBIC SPLINE INTERPOLATION
are specified. To some extent these end conditions are arbitrary. The commonly used
choices are:
Take 81 = 0, 8
n
= 0. This is equivalent to assuming that the end cubics approach
linearity at the extremities. This is so-called natural spline and frequently used.
Take 8
1
= 8
2
, 8
n
= 8
n
-
1
. This is equivalent to assuming that the end cubics
approach parabolas at their extremities.
Specify 8
1
and 8
n
so that the first derivatives of end points are equal to the slopes
of the unsaturated and saturated region. In doing so, both the saturated and
unsaturated regions are represented by straight lines.
2 cubic splines

.......... ,',., .................... "'." ..
18
16
14
Q)
OJ 12
Jg
10
Q)
: 8
(J)
6
0.2 0.4 0.6 0.8 1.2
Rotor field current (kA)
6 cubic splines

18
16
14
Q)
OJ 12
Jg
10
Q)
:E 8
(J)
6
0.2
airgap line
0.4 0.6 0.8 1.2
Rotor field current (kA)
Figure C.1 Cubic spline interpolation of the Benmore generator open circuit characteristic. Solid:
manufacturer data, dotted: spline,
Figure C,l shows a natural cubic pline interpolation on the open circuit characteristic
ofthe Benmore generator, Initially only three points from the tabulated data (obtained
from the manufacturer plot) was chosen: the origin, the end and a middle point, With
only two cubic splines, a poor agreement was obtained, whereas an almost exact match
was obtained using 6 cubic splines.
Appendix D
HARMONIC PHASOR REPRESENTATION
D.l COMPLEX FOURIER SERIES
A periodic function f(t) can be represented by the Fourier series oftrigonometric func-
tions:
00
f(t) = ~ + I: [ak cos(kwt) + bk sin(kwt)] , (D.l)
k=l
where
ao
1 f7r
-; -7r x(wt)d(wt)
1 f7r
-; -7r x(wt) cos(kwt)d(wt)
1 f7r
-; -7r x(wt) sin(kwt)d(wt) ,
(D.2)
and w is the fundamental angular frequency and k is the harmonic order.
As the sine and cosine functions can be expressed in terms of complex exponential
terms, i.e.
sin(kwt)
cos(kwt)
e
jkwt
_ e-
jkwt
2j
e
jkwt
+ e-
jkwt
2
(D.3)
the Fourier series in equation D.l can be expressed in the form of a complex exponential
Fourier series as
00
f(t) = I: Ckejkwt, (D.4)
-00
146 APPENDIX D HARMONIC PHASOR REPRESENTATION
where
Co
~ a k - jbk)
2
- ck
ao
2
(D.5)
The complex Fourier series expansion is compatible with the Fast Fourier Trans-
form, a method commonly used to determine the harmonic spectrum of time domain
waveforms. If frequent transformations from the time domain are not required, harmonic
phasors in positive frequency offers several advantages: the redundancy of storing and
processing complex conjugate pairs is avoided, the harmonic phasor is compatible with
existing power systems analysis method such as the load flow, and it is easy to decom-
pose the phasors into real components that is required in an unified Newton's solution
involving a mixture of complex and real number. The Fourier expansion in positive
frequency is to be described in the next section.
D.2 POSITIVE FREQUENCY HARMONIC PHASORS
The Fourier expansion can be represented by a series of phase shifted sine terms by
substituting
ak cos(kwt) + bk sin(kwt) = dk sin(kwt + 'l/Jk), (D.6)
into equation D.1, resulting in
00
f(t) = 2:I{Wkejkwt}, (D.7)
k=O
where
Wk
dke'ifJk
dk
j a ~ + b ~
'l/Jk
t -1 ak
an -.
bk
(D.8)
and W k being the positive frequency peak value harmonic phasors. The Fourier expansion
described by equation D.7 does not contain negative frequency components, and the
D.3 CONVOLUTION OF HARMONIC PHASORS
phasor at harmonic k is given as:
and the dc term becomes
,T, _ ao jrr/2 _ . ao
'l'o--e -]-
2 2'
When peak value phasors are used, the complex power is given by

D.3 CONVOLUTION OF HARMONIC PHASORS
147
(D.9)
(D.lO)
(D.ll)
The multiplication of two domain waveforms is expressed as a discrete convolution
of their Fourier series in the harmonic domain. The convolution of two harmonic phasors
of different frequencies will result in harmonic phasors at the sum and difference har-
monic orders. Discrete convolution can be explained by first multiplying two sinusoidal
waveforms and converted back to phasor form. The time domain multiplication of two
phasor Ak and Bm of harmonic orders k and m is:
IAkl sin(kwt+LAk) X IBml sin(mwt+LBm) = {sin[(k-m)wt +
- sin[(k + m)wt + LAk + LBm + .
(D.l2)
Converting to phasor form
[ej(LAk-LBm+'i) I _ej(LAk+LBm+'i) I ]
2 (k-m) (k+m)
[(IAklejLAk IBmle-jLBmej'i) - (IAk le
jLAk
IBmlejLBmej'i) ]
2 k-m k+m
- (AkBmh+ml (D.l3)
If k is less than m, negative harmonics can be avoided by conjugating the difference
term. Consequently, the convolution of the two sinusoidal waveforms can be expressed
as:
(D.l4)
148 APPENDIX D HARMONIC PHASOR REPRESENTATION
This expression is readily extended to the discrete convolution of the Fourier Series
of two non-sinusoidal periodic waveforms:
n n
L IAkl sin(kwt+LA
k
) L IBml sin(mwt+LBm)
k=O m=O
n n
= L L IAkl sin(kwt+LAk)IBml sin(mwt+LBm). (D.15)
k=Om=O
Rewriting this in terms of phasors as:
n n
FA FB = L L Ak Bm.
(D.16)
k=Om=O
Harmonic phasors of order up to 2n are generated by equation D.16 due to the sum
terms. Substituting equation D.14 for the convolution of two phasors into D.16 and
solving for the lth order component yields:
(A B)z
~ lta(AkBk+I )' + t.AkBk-l- t,AkBH] I> 0, (D.17)
(AB)z
~ [-
2A
OBO + t,AkBk] I o. (D.18)
Due to the conjugate terms, the convolution equations are non-analytic in the com-
plex form, but are differentiable by decomposing into real valued components such as in
rectangular form.
Appendix E
SIMPLIFIED CONVERTOR COMMUTATION
ANALYSIS AND HARMONIC CALCULATION
Assuming balanced and undistorted commutation voltage, ripple free dc current and
balanced commutation reactances, the expression for instantaneous commutation current
of a six-pulse bridge is [Arrillaga 1983]:
ie = ;( [cos a - cos(wt)].
y2X
e
(E.1)
where Ve is the phase-to-phase rms voltage and a is the firing angle. The commutation
ends when commutation current is equal to dc current, i.e. ie = Ide at wt = a + p,:
Ide = ;;e [cos a - cos(a + f.L)],
y2X
e
and thus the commutation angle f.L is calculated as:
-1 ( VZXeIde)
f.L = cos cos a - Ve - a
(E.2)
(E.3)
Substituting equation E.2 into E.1, the currents ofthe incoming and outgoing valves
in a six-pulse bridge during commutation, and subsequently phase a current are obtained
as:
Ide ( cos a - cos wt)
cos a - cos (a + f.L)
I I cos a - cos(wt - 21f/3)
dc-de ()
cos a - cos a + f.L
for a::; wt < a + f.L
21f 21f
for a + - < wt < a + - + I/.
J' 3 - 3""
(E.4)
(E.5)
The rates of changes of incoming and outgoing valve currents are equal since the
commutation reactances are balanced.
150APPENDIX E SIMPLIFIED CONVERTOR COMMUTATION ANALYSIS AND HARMONIC CALCULATION
Star-star transformer primary current
----
=!
0.5
S
Q)
-0
0
.a
'2
OJ
-0.5
ct!

-1
0 90 180 270 0 90 180
Angle (degrees)
0.25
Star-star transformer primary harmonic currents
----
::J 0.2
8
Q) 0.15
-0
.a
0.1
'2
OJ
ct!
0.05

0
5 10 15 20 25 30 35
Harmonic order
200
Ci)
Q)
100
Q)
.....
OJ
Q)
0

Q)
0,
-100
c

-200
5 10 15 20 25 30 35
Harmonic order
Figure E.1 Star-star transformer primary current and its harmonic content, commutation period is
13 degrees.
The line current is the same as the dc current in between commutations, i.e.
21f
f
or a + I/. < wt < - + a
r - - 3
(E.6)
The harmonic content can be obtained by integrating the described analytic expres-
sions over one cycle, but this is a complicated process. Alternatively, the waveform can
be contructed from a, f.t (calculated) and Ide, and an FFT is used to obtain the harmonic
information. Therefore, if the commutation voltage, commutation reactance, dc current
and firing angle are given, then the commutation angle and the harmonic content can
be calculated using the conventional formulation.
Figure E.1 shows the star-star transformer phase a primary current constructed
from the dc current and switching instances, and its harmonic content for a case where
a = 10, Xe = 0.05 p.u. and 1 p.u. commutation voltage. The calculated commutation
angle is 13.
Appendix F
JACOBIAN OF PARTIAL DERIVATIVES
The Jacobian matrix is derived for a six-pulse rectifier attached to a salient-pole gen-
erator via star-g/star transformer in this Appendix. The elements of the Jacobian are
the partial derivatives of the mismatch functions with respect to the variables that are
being solved for by Newton's method. The selected variables are V
k
\ V:, Vt,Idk, <Pi, Oi,
and aD. In finding the partial derivative of a mismatch function with respect to one of
these variables, all other variables are held constant. Figure F.1 shows the construction
of the Jacobian for 13 harmonics. The blocks associated with interactions between the
dc current harmonics and the ac voltage harmonics comprise the ac/dc partition, which
is 104 elements square. All other blocks are called switching terms.
The extension to twelve-pulse partial derivatives is relatively straightforward, where
the twelve-pulse ac/dc Jacobian partition being the sum of those of the star and delta
convertors, minus the identity matrix. The switching terms associated with each group
are the same. The derived Jacobian elements have been verified by comparison with the
numerical partial derivatives.
As the variables and functions are a mixture of real and complex numbers, decom-
position in to real values is required. The basic operations that are extensively used to
decompose the complex functions are given in equation F.1 to F.4.
The partial derivatives in real and imaginary components of a function with respect
to a variable are given as:
[
8RFv 8RFv
],
8Fv
8RV 8IV
(F.1)
av
8IFv 8IFv
8RV 8IV
when both Fv and V are complex numbers, and reduced 'to equations F.2 and F.3 when
152 APPENDIX F JACOBIAN OF PARTIAL DERIVATIVES
F<I>
26 ................ . ................. ; .................................................... .
aFV
av
aFV
ax
52 .................................... ; .................................................... .
X= 8,<j>,<x
o

aF
aF
x
Fe

aF
x

av
X= 8,<j>,<x
o
aId
ax
26 52 78 104 129
Vo
V_
V+ Id <I> e do
Figure F.l Partial derivatives in the Jacobian.
one of them is a real number.

V E real
aFv _ [8FV 8FV]
av - 8RV 8RV
Fv E real.
The decomposition of an impedance gives a 2 x 2 matrix, i.e.
[
R{V} 1 = [R{Z} -I{Z} 1 [ R{I} 1
I{V} I{Z} R{Z} I{I}
(F.2)
(F.3)
(F.4)
F.l VOLTAGE CONTROL EQUATION PARTIAL DERIVATIVES 153
F.l VOLTAGE CONTROL EQUATION PARTIAL DERIVATIVES
As the control system frequently deals with filtered signals, only the fundamental voltage
crossings are considered in the determination of the average delay angle.
12
def 1 ~
Faset = aset - 12 L)Bi - 'Yi) = 0,
i=l
Differentiating with respect to the selected variables gives:
oF
aset
ox
oF
aset
--
OBi
0,
1
12
For the phase angle, differentiating I{ FVl ps} = I{V1ps} = 0 gives
O V l ~ s
ox
oI{F
v1ps
}
1
oIV
1ps
oR{Fvlps}
oR"VI ps
oR{Fvlps} oI{F
v1ps
}
_:-'--_0........:.... ~ - 0
OIVI ps - oR"VI ps - .
where the subscript V1 ps denotes the fundamental positive-sequence voltage.
F.2 TERMINAL VOLTAGE PARTIAL DERIVATIVES
The terminal voltage mismatch has been defined as:
Expanding the expression over three phases yields:
[
va j [ E
a
j [zaa zab
Vb - Eb + zba Zbb
V
e
E
e
zea zeb
k k
zae j [Ia j
zbe Ib
zee Ie
k,k k
(F.5)
(F.6)
(F.7)
(F.8)
(F.9)
(F.10)
(F.ll)
[
zaa
+ zba
zea
zab
Zbb
zeb
zae j [ Ia j [ zaa
zbe Ib + zba
zee Ie zea
k,k+2 k+2
zab
Zbb
zeb
zae j [Ia j
zbe Ib
zee Ie
k,k-2 k-2
(F.12)
Thus the voltage mismatch for each phase can be written as:
FVk = V
k
a
- Ek + Zk,klk + +
Z
aa Ia + Za(3 1(3 Za"f TY
+ k,k+2 k+2 k,k+2 k+2 + k,k+2
1
k+2
Z
aa Ia + Za(3 1(3 + Za"f I"f
+ k,k-2 k-2 k,k-2 k-2 k,k-2 k-2
(F.13)
where [a (3 'YJ is a permutation of [a be]. The last six terms in equation F.13 are the
saliency terms which require the calculation of currents k + 2 and k - 2.
F.2.1 With Respect to ac Voltage Variation
Differentiating equation F.13 with respect to one of the phase voltage in both real and
imaginary parts yields
aR{Fvn

= R ul=k-2,k k,l an{vJJ + k,l an{vJJ + k,l an{v,O
{
al
k
Za(3 ale za"f all j
a =1= 6 or k =1= m
k+2 aa al
k
a(3 alt a"f all
1+ 2:1=k-2,k Zk,l an{vj,} + Zk,l an{v,O + Zk,l an{v,O
a=6 and k=m
(F.14)
aR{ Fvn _ ( aa alk a(3 alf a"f )}
- I Zk,l + Zk,l + Zk,l
(F.15)
aI{Fvn

= I l=k-2,k k,l aI{V,O + k,l aI{V,O + k,l
{
2:k+2 al
k
Za(3 alt Za"f all j
a =1= 6 or k =1= m
. zaa al
k
Za(3 ale za"f all
J + ul=k-2,k k,l aI{V,O + k,l aI{V,O + k,l aI{vJ,}
a=6 and k=m
aI{ Fvn _ { (aa alk a(3 alf a"f )}
- I Zk,l + Zk,l + Zk,l .
Subsequently, we need to calculate the partial derivatives of the ac currents with
respect to the ac voltage variation, As ac currents are functions of X =
[V
k
a
, V:, V
k
' I
dk
, i, Oil, and the switching instants are kept constant, only the effect of
phase colt age variation on the sampled commutation currents is required to be consid-
ered. This is because the commutation currents are functions of the terminal voltage,
whereas currents during" normal" conduction periods are functions of dc current har-
monies, which is kept constant.
(F.16)
(F.17)
F.2 TERMINAL VOLTAGE PARTIAL DERIVATIVES 155
As there are 6 commutation periods in one cycle:

{;;{ aoi ,
(F.18)

{;;{ C
aoi
,

(F.19)
where is the ith commutation current lci, sampled over the commutation period.
As there are four commutation period for each cycle of ac current, the coefficient C
aoi
defines whether for the ith commutation the phase a current is a function of the phase
f3 terminal voltage. Referring to Table 4.1, C is given as:
C
aoi
6
a b c
a [-1, 0, -1, -1, 0, -1] [0, 0, 1, 0, 0, 1] [1, 0, 0, 1, 0, 0]
a b [0, 0, 1, 0, 0, 1] [0, -1, -1, 0, -1, -1] [0, 1, 0, 0, 1, 0]
c [1, 0, 0, 1, 0, 0] [0, 1, 0, 0, 1, 0] [-1, -1, 0, -1, -1, 0]
Table F.1 The coefficient matrix G
aoi
which specifies the dependency among commutation current i,
terminal voltage phase, and ac current phase
The commutation current is given by
where
D
Icik ==
-I {tlcikejkBi}
k=l
jkwLeldk - Vebk
jkwLeb
(F.20)
(F.21)
(F.22)
and subscript e and b refer to the phases ending and beginning conduction respectively.
Sampling the commutation current requires a convolution with the relevant sampling
function:
[D+ tlCk]Q9\]i(2i-l), fori=1,2, ... ,6.
k=l
(F.23)
Setting leiD = D, expanding the convolution, and taking the kth component yields
(F.24)
Since the commutation circuit is linear, i.e. no couplings between harmonics, changes
in the commutation current at harmonic l, ieil' with respect to a voltage at different
harmonic m is zero:
8leil
= 0 for l, m such that l =1= m and l > o.
(F.25)
This allows the following partial derivatives to be written:
W < k
- 8n{v:} (2i-1h_m' m_
8l
e
im w* > k
8n{vJ,} (2i-l)m_k' m_
] . (F.26)
+
Similarly,
+
m < k ]
- . (F.27)
m"2 k
The calculations required to evaluate equations F.26 and F.27 can be approximately
halved by using the Cauchy Reimann equations for the partial derivatives of complex
functions. For an analytic function:
8F(z) . 8F(z)
8I{z} = J 8R{z} ,
(F.28)
and for the complex conjugate of an analytic function
8F(z)* .8F(z)*
8I{z} = -J 8R{z}
(F.29)
Consequently, the terms in equation F.27 is obtained by re-arranging the real and
imaginary parts of the corresponding terms in equation F.26. The remaining partial
derivatives, and are obtained from equations F.21 and F.22:
8Ieio
{ e
jkOi
} (F.30) -jI
jkwLeb '
8Ieio
{ e
jkOi
} (F.31)

-jn jkwLeb '
8Ieim
1
(F.32)

=
jkwLeb
It is assumed that phase r5 is beginning conduction; phase r5 ending conduction is
accounted for in the C matrix. This completes the derivation of the partial derivative
of terminal voltage mismatch with respect to terminal voltage variation.
F .2.2 With Respect to de Ripple Current Variation
There are two effects that need to be considered. The ac currents during "normal"
conduction periods are functions of dc current harmonics, and the commutation currents
are functions of the dc ripples. Differentiate the voltage mismatch with respect to the
dc current harmonics yields:
The phase currents are composed of commutation currents and the dc current sam-
pled on the ac side:
6
I
a
= 2:)Ei lei lSi W2i-l] + Id lSi w
a
. (F.37)
i=l
Ei, listed in Table F.2, is a coefficient matrix that specifies how the commutation
currents contribute to each of the phase currents, and w
a
is the phase a dc current
compound sampling function.
EC;<
t
ex
a b c
1 1 0 -1
2 0 -1 1
i 3 -1 1 0
4 1 0 -1
5 0 -1 1
6 -1 1 0
Table F. 2 Coefficient matrix Ef defining the contribution of the commutation currents to each phase
current. i is the commutation number.
Differentiating the kth component of equation F.37 with respect to n{ ldm} gives:
(F.38)
The first term is similar to the differential with respect to voltage already derived
in the previous section, as a variation in ldm affects only lcim and leiO. Expanding the
convolution in the first term, selecting the kth component, and differentiating yields
Expanding the convolution in the second term of equation F.38 gives the kth com-
ponent of the sampled dc current:
k > 0, (F.40)
where ldk is the phase ex sample of the dc current. This equation can be differentiated
to give:
(F.41)
Applying the above analysis for variations in the imaginary part of Idm gives
and
(F.43)
These equations are then substituted into an equation analogous to equation F.38
and hence into equations F.34 and F.36. The remaining derivatives, and
are obtained from equations F.21 and F.22:
8Icio
. {LeejkOi}
(F.44) =
-JI L '
8R{Idm}
eb
8I
c
io
. {LeejkOi}
(F.45)
8I{Idm}
-JR L '
eb
8Icim Le
(F.46)
-
8R{Idm} Leb
F .2.3 With Respect to end of Commutation Variation
A variation in the end of commutation angle, (Ph, affects the (2h-l)th and 2hth sampling
functions. This affects all harmonics of the ac side sampled commutation currents, and
dc current. Combining these two effects into the phase current variation, and injecting
into the ac system yields the variation in the terminal voltage mismatch:
R { (zaa 8It: Za(3 8If Za,,{ 8IZ) }
Lt k,l 8h + k,l 8h + k,18h '
l=k-2,k
(F.47)
{
( aa 8It: a(38If a"{ 8IZ ) }
- I Zk,l 8h + Zk,l 8h + Zk,l 8h .
(F.48)
Differentiating the kth component of equation F.37 with respect to 1w
(F.49)
A variation in h affects only the sampling of the hth commutation current in the
first term of equation F.49. Expanding only this convolution of the summation yields:
k > O.
(F.50)
Differentiating this gives
{
nh 8w* nh (8W* ) *
L..t chI 8'" + L..t chI 8'"
.
""' I (2h-l)l_k ""' I (2h-l)l+k
l=k 'Ph 1=0 'Ph
~ 8W(2h-l)k_l}
L..t chI 8'" k > 0,
1=0 'Ph
(F.51)
where
(F.52)
The compound sampling function in the second term of equation F.49 is affected by
variation in four of the six end of commutation angles. This is because the transfer of dc
current to the ac side is defined by two conduction periods per cycle; the beginning and
end of each conduction period corresponding to an end of commutation angle. The effect
of a variation in h therefore depends upon whether it corresponds to the beginning or
end of a positive or negative conduction period for phase a. This information is already
collated as the coefficient matrix Ef.
A similar analysis to the above for the second term of equation F.49 results in
{
nh 8,T,* nh ( 8'T' * ) *
""' ~ 2hl-k ""' ~ 2h/+k
L..t Idl 8'" + L..t Idl 8'"
l=k 'Ph 1=0 'Ph
~ I 8W2hk-l}
L..t dl 8'" k > 0,
1=0 'Ph
(F.53)
Since the compound sampling function is a sum of constituent sampling functions,
it has been replaced in the partial derivative by the only term which is a function of h'
This is then substituted back into equation F.49 to give the required partial derivative.
F .2.4 With Respect to Firing Instant Variation
The transfer of dc current to the ac side of the convertor is defined entirely with reference
to the end of commutation angles. The effect of a variation in the firing instants on
the phase currents can therefore be obtained by analysing just the sampling of the
commutation currents. The partial derivative of the ac voltage mismatch with respect
to a firing angle variation is:
Since the ac side sampled dc current is not a function of e
h
a { By } k
arh
= a W(2i-l)} k
aeh
a {Bh W (2h-l)} k
ae
h
(F.54)
(F.55)
Noting that a change in eh affects Icho and W(2h-l)' Applying the product rule to
the expansion of the convolution leads to
This completes the analysis of the terminal voltage mismatch equation partial deriv-
atives as there is no dependence upon ao, the only remaining variable.
F.3 DIRECT CURRENT PARTIAL DERIVATIVES
The direct current partial derivatives are obtained by a process analogous to that used in
obtaining the terminal voltage partial derivatives. The dc voltage partial derivatives are
obtained and then applied to the dc system admittance to calculate the direct current
differential. The direct current mismatch equation is
(F.57)
Finding the partial derivatives of this equation is therefore mainly concerned with
Vdk, which is a function of all the convertor variables except aD.
Differentiating equation F.57 with respect to terminal voltage yields
8R{F
n
}
{ 8V
dk
}
(F.58)
8R{V;;}
R - Y
dk
8R{V;;} ,
8R{F
n
}
{ 8V
dk
}
(F. 59)
8I{V;;}
R - Y
dk
8I{V;;} ,
8I{Fn}
{ 8V
dk
}
(F.60)
8R{V;;}
I -Ydk8R{V;;} ,
8I{F
1
d
{ 8V
dk
}
(F.61)
8I{V;;}
I -Y
dk
8I{V;;} .
Vdk is given by
k > O. (F.62)
The only terms in this equation which are a function of V;; are the twelve pre-
convolved dc voltage phasors Vdil' There are three equations which define the twelve
pre-convolved de voltage phasors:
(F.63)
for a 'normal' conduction interval,
(F.64)
F.3 DIRECT CURRENT PARTIAL DERIVATIVES 163
for a commutation on the positive dc rail, and
(F.65)
for a commutation on the negative rail. Differentiating these three equations with respect
to Vr?;, gives 36 possible values for ~ ~ according to the sample number i and the phase
m
of O. The required partial derivatives are summarized in table F.3 by reference to the
partial derivatives of equations F.63, F.64, and F.65, where:
aVdil
an{v;+}
aVdil
an{Vl-}
for a 'normal' conduction interval,
aVdil
an{Vl
e
}
aVdil
an{Vl
b
}
aVdil
an{VlO}
= 1
-1
-1
for a commutation on the positive dc rail, and
aVdil Lb
=
an{Vl
e
} Lb +Le
aVdil Le
an{Vl
b
} Lb+L
e
aVdil
1
an{VlO}
(F.66)
(F.67)
(F.68)
for a commutation on the negative rail. The Cauchy Reimann equations can be used to
give the partial derivative with respect to the imaginary part of the voltage variation as
j times that listed above.
The partial derivatives in equations F.66, F.67, and F.68 are then substituted into
sample (i) DC Voltage derivatives
b e 0
+
- eqn
1 A C B F.67
2 A B F.66
3 C B A F.68
4 A C F.66
5 B A C F.67
6 B C F.66
7 A C B F.68
8 B A F.66
9 C B A F.67
10 C A F.66
11 B A C F.68
12 C B F.66
Table F. 3 Construction of dc voltage partial derivatives.
the partial derivatives of equation F.62, which are given below:
1 . U dim *
12 [( !::IV; ) * {
"2
J
~ I { V ~ } Wim+k +
- ,k > 0,
m < k 1
m"? k
8Vdim ,T, < k
- 8I{VJ,} 'l'ik-m m_
8Vdi;p ,T,* > k
8I{V
m
} 'l'im-k m_
(F.69)
1 ,k > O.
(F.70)
This completes the derivation of the partial derivative of dc current mismatch equa-
tion with respect to terminal voltage.
F.3.2 With Respect to dc Ripple Current Variation
Apart from being the most significant term in the dc current mismatch equation, the
dc ripple affects the commutation currents, and also causes a voltage drop through the
commutating reactance. These last two effects mean that the dc voltage is a function
of the de current ripple. It is therefore necessary to obtain partial derivatives of the dc
voltage harmonics in a similar manner to that undertaken already for the derivative with
respect to terminal voltage. Differentiating the dc current mismatch equation gives:
aR{F
Ik
}
{ y av"
k -j. m
R - dk &R{Idm}
(F.71)
aR{I
dm
} 1 y; &Vdk
k=m
- dk&R{Idm}
aR{F
Ik
}
{ aVdk}
(F.72)
aI{I
dm
}
R - Y
dk
aI{ I
dm
} ,
aI{FIk}
{ y &V" k -j. m
I - dk &I{Idm}
(F.73)
aI{Idm}
. y; &Vdk
k=m
J - dk &I{Idm}
aI{Flk}
{ aVdk}
(F.74)
aR{I
dm
}
I -Y
dk
aR{I
dm
} .
Differentiating equation F.62 with respect to R{Idm} yields
aVdk 1 t [ (BVd;m )' {
&Vdim \If
m"'5:k
1 ,k > O.
- &R{Idm} ik-m
8R{Idm} = 2
J
i=l 8R{Idm} \lfim+k +
&Vdim \If *
m'2 k
&R{Idm} i m-k
(F.75)
Similarly,
a 12 [( D r {
&Vdim \If
m"'5:k
1 ,k > O.
Vdk 1 . Vdim *
8I{I
dm
} = 2
J
~ 8I{Idm} \lfim+k +
- &I{Idm} ik-m
&Vdim \If*
m'2 k
&I{Idm} i m-k
(F.76)
The partial derivatives ~ r t : } are obtained from equations F.63, F.64, and F.65:
(F.77)
during normal conduction, and
(F.78)
during any commutation. The imaginary partial derivatives are then obtained by the
Cauchy Reimann equations. The correct phase subscripts can be obtained from Ta-
ble F.3. This completes the linearized dependence of the dc mismatch upon dc ripple
variation.
F.3.3 With Respect to end of Commutation Variation
The effect of variation in the end of commutation angle h is limited solely to the
sampling of relevant dc voltage sections. This is best explained with reference to the
following equation for the dc voltage:
(F.79)
In this equation, only two of the twelve Wi are functions of h' Table F.4 shows
that W2h and W2h-1 are functions of h. Equation F.62 can therefore be differentiated
to yield:
{
EN! 2hk-1 ~ (8W 2hl-k ) *
L..t V
d
(2h-l)I 8'" - L..t V
d
(2h-1)1 8'"
~ ~ ~ ~
~ [ (
8W
2hl-k) *] * ~ 8W2hk-1
~ V
d
(2h-1)1 8h - ~ V
d
(2h)1 8h
~ (8W2hl-k)* ~ [ (8W2hl-k)*]*}
+ f::t V
d
(2h)1 8h + V
d
(2h)I 8h (F.80)
sample (i) ai bi
1 (h
1
2
1
8
2
3 8
2 2
4
2
8
3
5 8
3 3
6
3
8
4
7 84
4
8
4
8
5
9 8
5 5
10
5
8
6
11 8
6 6
12
6
8
1
Table F.4 Limits of convertor states for use in sampling functions.
Differentiating the dc ripple mismatch equation F.57 yields:
8R{FJk} = R {_y; 8V
dk }
8h dk 8h '
(F.81)
and
8I{FJk} = I {_y; 8Vdk}
8h dk 8h .
(F.82)
Substituting equation F.80 into F.81 and F.82 gives the required partial derivative
of dc ripple mismatch with respect to end of commutation.
F .3.4 With Respect to Firing Instant Variation
The partial derivatives of dc ripple mismatch with respect to firing angle are obtained
in an exactly similar manner. The result is:
(F.83)
and
(F.84)
where
FA END OF COMMUTATION MISMATCH EQUATION
PARTIAL DERIVATIVES
The end of commutation mismatch is the current in the phase that is commutating off,
at the end of the commutation. This current should be zero, and is given by:
for a commutation on the positive rail, where
D -I {I: leikejkOi}
k=l
ldk - leik
jkwLeldk - V
k
eb
jkwLeb
(F.86)
(F.87)
(F.88)
(F.89)
The end of commutation mismatch equation is therefore a function of all the vari-
ables except the average delay angle ao. A commutation on the negative rail is accounted
for by substituting -ldm in the above equations for ldm'
FA.1 With Respect to ac Voltage Variation
Differentiating equation F.86 with respect to an arbitrary voltage phase and harmonic
yields:
aFc/Ji _ {aD aleim jmc/Ji}
- I - - e ,
(F.90)
and
aFc/Ji _ {aD aleim jmc/Ji}
- n - - e .
(F.91)
The partial derivatives in these equations are obtained from equations F.87 to F.89
FA END OF COMMUTATION MISMATCH EQUATION PARTIAL DERIVATIVES 169
as
8D
1 ejkBi
(F.92)
8R{V:;} jmw(Le + L
b
) ,
8l
eim
-1
(F.93)
8R{V:;} jmw(Le + Lb)'
8D
J e
jkBi
(F.94)
8I{V:;} jmw(Le + L
b
) ,
8l
eim
-J
(F.95)
8I{V:;} jmw(Le + Lb)'
It is assumed in the above analysis that V:; corresponds to a phase ending con-
duction. Multiplying by -1 gives the required partial derivative for the case that V:;
corresponds to a phase beginning conduction.
The analysis is similar to that for ac phase voltage variation. Differentiating equa-
tion F.86 yields:
(F.96)
and
(F.97)
The partial derivatives in these equations are obtained from equations F.87 to F.89
as
8D
Le ejkBi
(F.9S)
8R{ldm} Le + Lb '
8l
eim
-Le'
(F.99)
8R{ldm} Le + Lb'
8D
jLe jkBi
(F.100)
8I{ldm}
e ,
Le +Lb
8l
eim
-jLe
(F.I01)
8I{ldm} Le + Lb'
This analysis assumes that the commutation is on the positive rail. A similar analy-
sis holds for a commutation on the negative rail, but with - Idm substituted into equa-
tion F.86.
FA.3 With Respect to end of Commutation Variation
This partial derivative gives the effect on the 'residual' commutating-off current at the
end of the commutation, if the end of commutation is moved. It is obtained simply by
differentiating equation F.86 with respect to CPi to give
[)F ' { nh '}
= I """" jkF,I., eJki
[)L L...J nk .
f ~ k=l
(F.I02)
FAA With Respect to Firing Instant Variation
The dc offset to the commutation, D, is a function of the firing instant, and so the only
effect of (}i on Fi is through D. Differentiating the expression for D, equation F.87,
gives the required partial derivative.
[)F ' {nh '}
= I - """"J'kI e
JkOi
[)(), L...J ck .
~ k=l
(F.I03)
F.5 FIRING INSTANT MISMATCH EQUATION PARTIAL DERIVATIVES
F.5 FIRING INSTANT MISMATCH EQUATION PARTIAL
DERIVATIVES
171
For the implemented constant current controller, the firing instant mismatch is not a
function of the ac terminal voltage. This is because the firing order is obtained solely
from monitoring the dc current. This also means that the firing mismatch is not a
function of the end of commutation instants. The firing instant mismatch is a function
only of the dc ripple, the firing angle, and the average firing order. The firing mismatch
equation is:
where
. ~ G (p l)I
X ~ k - 1 + jkwT + jkwTJ dk
The partial derivatives are readily obtained as:
where
aF
Oi
acxo
aF
Oi
a()i
1,
8CXik _ G (p 1)
8Idk - 1 + jkwT + jkwTJ .
(F.104)
(F.105)
(F.1D6)
(F.1D7)
(F.1D8)
(F.1D9)
(F.110)
F.6 AVERAGE DELAY ANGLE PARTIAL DERIVATIVES
The average delay angle, ao, is obtained by requiring that the average dc voltage, when
applied to the dc system, should result in the current order. Thus ao is a control variable,
required for the case of constant current control. The mismatch equation is:
(F.l11)
where Vdc represents a dc voltage source, and
(F.112)
Thus Fao is a function of all the convertor variables, with the exception of ao itself.
Analysis of the partial derivatives of equation F.111 is similar to that for the partial
derivatives of the dc ripple mismatch.
Differentiating the mismatch equation F.l11 with respect to an arbitrary phase and
harmonic of ac voltage yields
and
{
aI{VjJ = I - Y
dO
aI{VjJ .
Differentiating equation F.112 yields
12
aVdO _! . '" ( aVdim )
an{vJJ - 2
J
an{vjJ m ,
12
aVdO 1 . '" (aVdim *)
aI{VjJ = ;jJ aI{VjJ Wim ,
(F.113)
(F.114)
(F.115)
(F.116)
which when substituted back into equations F.113 and F.114 give the required partial
derivatives. The remaining partial derivatives, etc, have already been obtained
in equations F.63, F.64, and F.65, and by reference to Table F.3.
F.6 AVERAGE DELAY ANGLE PARTIAL DERIVATIVES 173
Variation in the dc ripple affects the dc voltage samples in a similar manner to that of
a variation in the terminal voltage above. Differentiating the mismatch equation F.lll
with respect to an arbitrary phase and harmonic of dc current ripple yields
(F.117)
(F.118)
Differentiating equation F.112 yields
12
8VdO 1 (8Vdim *)
8R{I } = 2
J
L.t 8R{I } Wi m ,
dm i=1 dm
(F.119)
and
12
8VdO 1 . (8Vdim *)
8I{I } = 2
J
L.t 8I{I }Wim ,
dm i=1 dm
(F.120)
which when substituted back into equations F.117 and F.118 give the required partial
derivatives. The remaining partial derivatives, etc, have already been obtained
in equations F.77 and F.78.
F .6.3 With Respect to end of Commutation Variation
The effect of a variation in the end of commutation is to modify the sampling of the
dc voltage sections in equation F.112. Differentiating equation F.lll with respect to CPh
yields
8F
ao
= I {_y; 8V
dO
}
8CPh dO 8CPh .
(F.12l)
This requires the partial derivative Differentiating equation F.112 yields
(F.122)
where use has been made of Table FA to determine the only two sampling functions
that are affected by CPh.
F.6,4 With Respect to Firing Instant Variation
The effect of a change in the firing angle on the average delay angle mismatch equation
is similar to that for a change in the end of commutation angle. The sampling of the
dc voltage sections is modified, and this changes the average dc voltage. The analysis
carried out above for the end of commutation variation is also valid in this case, with
only the two affected sampling functions, as determined from Table F.4 being different.
The result is that
(F.123)
where
{
nh 8\[1* nh 8\[1*}
8VdO 1 . (2h-2)1 (2h-1)1
ae ="2,1 I:: V
d
(2h-2)I 88 + I:: V
d
(2h-1)1 88 .
h 1=1 h 1=1 h
(F.124)
This completes the derivation of the partial derivatives required for the Jacobian
matrix, as the average delay angle mismatch equation is not a function of the average
delay angle.
Appendix G
PUBLISHED PAPERS
N.Q. Dinh and J. Arrillaga, Direct Connections of generators and HVDC converters,
Proc. Third New Zealand Conference of Postgraduate Students in Engineering and Tech-
nology, July, 1996, New Zealand, pp. 313-18.
N.Q. Dinh and J. Arrillaga, Generator Modelling in the Harmonic Domain, Proc. In-
ternational Power Engineering Conference, May, 1997, Singapore, pp. 804-08.
N.Q. Dinh, J. Arrillaga, A.R. Wood, S. Chen, W. Enright, B.C. Smith and Z. Du, Har-
monic Evaluation of Benmore Converter Station When Operated as a Group Connected
Unit, IEEE Transactions on Power Delivery, Vol 12, No 4, Oct, 1997, pp. 1730-35.
N.Q. Dinh, J. Arrillaga and B.C. Smith, A Steady-state model of direct connected
generator-HVDC convertor units in the Harmonic Domain, accepted for publication in
the Proc. lEE Generation, Transmission and Distribution.
REFERENCES
11/14-09, C.J.W.G. (1993), 'Hvdc unit connected generators', ELECTRA, Vol. 149,
August.
11/14-09, C.J.W.G. (1997), Guide for Preliminary Design and specification of Hydro
Stations with HVDC Unit Connected generators.
36.05.02/14.03.03, C.J.T.F. (1996), 'Ac system modelling for ac filter design - an
overview of impedance modelling', ELECTRA, No. 164, February, pp.133-151.
ARRILLAGA, J. (1983), High Voltage Direct Current Transmission, Vol. 6, Peter Pere-
grinus Ltd., London, UK.
ARRILLAGA, J., WATSON, N., EGGLESTON, J. AND CALLAGHAN, C. (1987), 'Com-
parison of steady state and dynamic models for the calculation of ac/dc system
harmonics', lEE Proceedings Pt. C, Vol. 134, No.1, January, pp.31-37.
ARRILLAGA, J., YONGE, L., CRIMP, C. AND VILLABLANCA, M. (1993a), 'Harmonic
elimination by de ripple reinjection in generator-converter units operating at vari-
able speeds', lEE Proceedings-C, Vol. 140, No.1, August, pp. 57-64.
ARRILLAGA, J., CAMACHO, J., MACDoNALD, S. AND ARNOLD, C. (1993b), 'Operating
characteristics of unit and group connected generator-hvdc converter schemes', lEE
Proceedings-C, Vol. 140, No.6, November, pp. 503-8.
BLONDEL, A. (1923), 'The two-reaction method for study of oscillatory phenomena in
coupled alternators', Revue generale de I'electricite, Vol. 13, February, pp. 235-51.
B ROYDEN , C. (1965), Mathematics of Computation, Vol. 19. pp. 557-93.
BURDEN, R.L., FAIRES, J.D. AND REYNOLDS, A. (1981), Numerical Analysis, PWS
Publishers, Boston, Massachusetts, second ed.
CALLAGHAN, C. AND ARRILLAGA, J. (1990), 'Convergence criteria for iterative har-
monic analysis and its application to static convertors', ICHPS IV, October,
pp.38-43.
CALVERLEY, T.E., OTTAWAY, C.H. AND TUFNELL, D.H.A. (1973), 'Concept of a unit
generator converter transmission system', lEE Conference on High Voltage dc and
ac Transmission.
178 REFERENCES
CHEN, S., WOOD, A.R. AND ARRILLAGA, J. (1996), 'Hvdc convertor core saturation
instability; a frequency domain analysis', lEE Proc. Pt C, Vol. 143, No.1, January,
pp.75-81.
CHEUKSUN, W., MOHAN, N., WRIGHT, S. AND MORTENSEN, X. (1989), 'Feasibility
study of ac and de side active filters for hvdc converter terminals', IEEE Trans.
on Power Delivery, Vol. 4, pp. 2067-75.
CLARKE, E. (1950), Circuit Analysis of AC Power Systems, Vol. 2, John Wiley & Sons,
London.
(1996), Continuous Harmonic Analysis in Real Time: Technical Reference, CHART
Instrument Limited, New Zealand.
DOHERTY, R. AND NICKLE, C. (1926), 'Synchronous machines 1: An extension of
blondel's two-reaction theory', AlEE Transactions, Vol. 45, pp.927-42.
DOMMEL, H. (1969), 'Digital computer solution of electromagnetic transients in sin-
gle and multiphase networks', IEEE Trans. on Power Apparatus and Systems,
Vol. PAS-88, No.4, April, pp. 388-399.
DOMMEL, H., CHIU, B. AND MEYER, W. (1980), 'Analysing transients in a.c./d.c.
systems with the bpa electromagnetic transients program', IEEE Conference Pub-
lication 205, on Thyristor and Variable Static Equipment for a. c. and d. c. Trans-
mission, pp. 109 - 113.
EGGLESTON, J. (1985), Harmonic Modelling of Transmission Systems Containing Syn-
chronous Machines and Static Convertors, PhD thesis, University of Canterbury,
New Zealand.
EGGLESTON, J.F., ARRILLAGA, J. AND SEMLYEN, A. (1988), 'Analysis ofthe harmonic
distortion resulting from the interaction between synchronous machine and hvdc
convertors', IEEE Trans. on Power Delivery, Vol. 3, No.1, January, pp. 325-332.
GERALD, C. AND WHEATLY, P.O. (1985), Applied numerical analysis, Addison-Wesley,
third ed.
GOLE, A., FERNANDO,!., IRWIN, G. AND NAYAK, O. (1997), 'Modelling of power
electronic apparatus: Additional interpolation issues', International Conference
on Power Systems Transients Proceedings, June, pp.23-28.
HARLEY, R.G., LIMEBEER, D.J.N. AND CHIRRICOZZI, E. (1980), 'Comparative study
of saturation methods in synchronous machine models', lEE Proc., Pt. B, Vol. 127,
No.1, January, pp.1-7.
REFERENCES 179
HWANG, H. (1969), 'Unbalanced operation of three-phase machines with damper cir-
cuits', IEEE Trans. on Power Apparatus and Systems, Vol. PAS-88, No. 11, No-
vember, pp.1585-93.
KILGORE, L. (1935), 'Effects of saturation on machine reactances', AlEE Trans., Vol. 54,
May, pp. 545-50. Disc., v.56, 903-4, July, 1937.
KIMBARK, E.W. (1968), Power system stability: Synchronous machines, Dover Publi-
cations, Inc., New York, New York, NY, USA.
KIMBARK, E.W. (1971), Direct Current Transmission, Vol. 1, John Wiley and Sons,
Inc.
KRISHNAYYA, P. (1973), 'Stresses on generators and transformers of block and double-
block connections proposed for hvdc power station infeed.' , lEE conf. on 'High
voltage de and/or ac power transmission', Publication no. 107, pp. 279-286.
LARSON, E., BAKER, D. ANI;) MACIVER, J. (1989), 'Low order harmonic interaction
on ac/dc systems', IEEE Transactions on Power Delivery, Vol. 4, No.1, January,
pp.493-501.
LEE, C. (1961), 'Saturation harmonics of polyphase induction machines', AlEE Trans.,
Vol. Part III, No. 80, October, pp. 597-603.
LISBOA, M.L.V. (1996), Three-phase three-limb transformer models in the Harmonic
Domain, PhD thesis, University of Canterbury, New Zealand.
MACDONALD, S., ENRIGHT, W., ARRILLAGA, J. AND O'BRIEN, M. (1995), 'Harmonic
measurements from a group connected generator hvdc converter scheme', IEEE
Transactions on Power Delivery, Vol. 10, No.4, October, pp.1937-43.
MCPHERSON, G. AND LARAMORE, R.D. (1990), An Introduction to Electrical Machines
and Transformers, Wiley.
MEDINA, A. AND ARRILLAGA, J. (1994), 'Analysis of transformer-generator interac-
tion in the harmonic domain', lEE Proc.-Gener. Transm. Distrib., Vol. 141, No.1,
January, pp. 38-46.
NAIDU, M. AND MATHUR, R. (1989), 'Evaluation of unit connected, variable speed,
hydropower station for hvdc power transmission', IEEE Transactions on Power
Systems, Vol. 4, No.2, May, pp.668-676.
NARAIN, H.G. (1996), 'High-voltage dc transmission: A power electronics workhorse',
IEEE Spectrum, April, pp. 63-72.
PARK, R. (1929), 'Two-reaction theory of synchronous machines-p.1, generalized method
of analysis', AlEE Trans., Vol. 48, July, pp. 716-30.
180 REFERENCES
PRESS, W., TEUKOLSKY, S., VETTERLING, W. AND FLANNERY, B. (1992), Numerical
Recipes in FORTRAN, Cambridge University Press, New York, New York, second
ed.
REEVE, J. AND BARON, J. (1971), 'Harmonic interaction between hvdc convertors and
ac power systems', IEEE Transactions on Power Apparatus and Systems, Vol. 90,
No.6, pp.2785-2793.
SANKAR, S. (1991), Dynamic Simulation of AC/DC systems with reference to convertor
control and unit connection, PhD thesis, University of Canterbury, New Zealand.
SAY, M.G. (1983), Alternating Current Machine, Longman Scientific & Technical, UK,
fifth ed.
SEMLYEN, A., ACHA, E. AND ARRILLAGA, J. (1988), 'Newton-type algorithms for the
harmonic phasor analysis of non-linear power circuits in periodical steady state
with special reference to magnetic non-linearities', IEEE Trans. on Power Delivery,
Vol. 3, No.3, July, pp.1090-1097.
SMITH, B.C., WATSON, N.R., WOOD, A.R. AND ARRILLAGA, J. (1995), 'Steady state
model ofthe ac/dc convertor in the harmonic domain', IEEE Proc.-Gener. Transm.
Distrib., Vol. 142, No.2, March, pp.109-118.
SMITH, B.C., WATSON, N.R., WOOD, A.R. AND ARRILLAGA, J. (1996), 'Newton solu-
tion for the steady-state interation of ac/dc systems', IEEE Proc.-Gener. Transm.
Distrib., Vol. 143, No.2, March, pp.200-210.
UHLMAN, E. (1975), Power Transmission by Direct Current, Springer-Verlag.
VALCARCEL, M. AND MAYORDOMO, J. (1993), 'Harmonic power flow for unbalanced
systems', IEEE PES winter meeting 93 WM 061-2 PWRD.
WOOD, A.R. AND ARRILLAGA, J. (1995), 'Hvdc convertor waveform distortion - a
frequency domain analysis', lEE proceedings Pt C, Vol. 142, No.1, January, pp. 88-
96.
WOODFORD, D., GOLE, A. AND MENZIES, R. (1983), 'Digital simulation of de links
and ac machines', IEEE Trans. on Power Apparatus and Systems, Vol.PAS-102,
No.6, June, pp.1616-1623.
XIA, D. AND HEYDT, G. (1982), 'Harmonic power flow studies, part i and ii', IEEE
Trans on Power Apparatus and Systems, Vol. PAS 101, No.6, pp.1257-70.
Xu, W.W., MARTI, J.R. AND DOMMEL, H.W. (1990), 'Multiphase harmonic load flow
solution technique', IEEE PES winter meeting 90WM 098-4 PWRS.
REFERENCES 181
Xu, W.W., DOMMEL, H.W. AND MARTI, J.R. (1991), 'A synchronous machine model
for three-phase harmonic analysis and emtp initialization', IEEE Trans. on Power
Systems, Vol. 6, No.4, November, pp.1530-1538.
Xu, W.W., DRAKOS, J.E., MANSOUR, Y. AND CHANG, A. (1994), 'A three-phase
converter model for harmonic analysis of hvdc systems', IEEE Trans. on Power
Delivery, Vol. 9, No.3, July, pp.1724-1730.
YACAMINI, R. AND DE OLIVEIRA, J.C. (1980), 'Harmonics in multiple convertor sys-
tems: a generalised approach.', lEE Proceedings Pt.B, Vol. 127, No.2, March,
pp.96-106.
YACAMINI, R. AND DE OLIVEIRA, J.C. (1986), 'Comprehensive calculation of convertor
harmonics with system impedances and control representation', lEE Proceedings
Pt. B, Vol. 133, No.2, March, pp. 95-102.
ZAGRADISNIK, 1. AND HRIBERNIK, B. (1988), 'Influence of anisotropy of magnetic mate-
rialOli the saturation. harmonics in the three-phase induction motor', IEEE Trans.
on Magetics, Vol. 24, No.1, January, pp.491-494.
ZHANG, W., ASPLUND, G., ABERG, A., JONSSON, U. AND LOOF, O. (1993), 'Active
dc filter for hvdc system - a test installation in the konti-skan dc link at lindome
converter station', IEEE Trans on Power Delivery, Vol. 8, No.3, July.
ZLATEV, Z. (1991), Computational Methods for General Sparse Matrices, Kluwer Aca-
demic Publishers, Netherlands.
ZOLLENKOPF, K. (1970), 'Bi-factorisation basic computational algorithm and program-
ming techniques', Conference on large sets of sparse linear equations.

Dinh Thesis

Uploaded by

Copyright:

Available Formats

Dinh Thesis

Uploaded by

Document Information

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

Dinh Thesis

Uploaded by

Copyright:

Available Formats

Harmonic Domain Modelling of

Direct Connected Generator and

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.