2012 International Conference and Exposition on Electrical and Power Engineering (EPE 2012), 25-27 October, Iasi, Romania

An Adaptive Hysteresis Band Current Control For

Three Phase Shunt Active Power Filter Using Fuzzy
Logic

George Adam, Alina Georgiana Stan (Baciu), Gheorghe Livinţ

Faculty of Electrical Engineering
“Gheorghe Asachi” Technical University of Iaşi
Iaşi, Romania
yojorj@yahoo.com

Abstract— Over the recent years, in order to compensate current current control with fixed hysteresis band has the main
harmonics and reactive power in three power systems, active disadvantages of variable switching frequency, consequently
filtering is used increasingly often. The reference current increasing switching losses, audible noise, sub-harmonics and
extraction method and the PWM technique used in the control of difficulties to design the input filter.
active power filter have an important role, because any
inaccuracy yields to incorrect compensation. This paper presents In addition, in order to keep a high efficiency and safety in
a new adaptive hysteresis band controller for three phase shunt the APF operation, the switching frequency and the DC source
active power filters implemented using the Fuzzy logic. The voltage, which are highly relevant to the current control
simulations were carried using Matlab SimPowerSystems and method used, must be kept as low as possible. To overcome
Fuzzy logic toolboxes under two load configurations, balanced these undesirable problems, adaptive hysteresis band current
and unbalanced. The results are found quite satisfactory in order controllers had been proposed in [1-4].
to keep the switching frequency constant, and to compensate the
current harmonics, unbalance and reactive power in three phase This paper presents a new current control technique, with
electrical systems. the hysteresis band implemented with Fuzzy logic.
The simulations were carried using Matlab
Keywords-active power filters; fuzzy logic; harmonics;
hysteresis band; three phased system
SimPowerSystems and Fuzzy Logic toolboxes to demonstrate
the effectiveness of the proposed adaptive hysteresis band
I. INTRODUCTION using fuzzy logic current control.
With the large number of power electronics based The studied APF in this paper is a three-phase four wire
equipment come the real problem of harmonic distortion. The neutral clamped APF compensating harmonics, unbalance and
consequences of harmonic distortion is deterioration of current reactive power created first by a nonlinear balanced load and
and voltage waveforms, power losses, aging of materials, then by a nonlinear unbalanced load based on the
excessive heating in rotating machines, significant interference Instantaneous Reactive Power Theory (IRPT).
and equipment damage risk.
II. SHUNT ACTIVE POWER FILTER
Over the last years, in order to avoid these undesirable
effects, passive LC filters have been used; they are ineffective Fig. 1 presents the shunt active power filter topology,
due to large size, resonance with the grid, fixed compensation connected in parallel with the source and two types of loads.
characteristics and inability to adapt to network variation. Typically, the shunt APF is a current controlled voltage source
In the last 20 years, the shunt active power filters (APF) inverter which eliminates current harmonics, balance the
were proposed and developed in different topologies to unbalanced loads and compensates the necessary reactive
mitigate current harmonics, balance unbalanced loads and to power required by the nonlinear load, by injecting the
compensate reactive power, without the drawbacks of the compensating current in the point of common coupling (PCC).
conventional passive LC filters. According to (1), the APF needs to produce a current as
The performance of active power filters depends mainly on close as possible to the reference current, which it is usually
the harmonic detection method and current control. Nowadays, obtained by subtracting the fundamental component of the
there are various current control methods and PWM load current from the load current. The compensating current
techniques. One of the most popular techniques is the is equal but opposite to the harmonic components of the load
hysteresis band current control, due to its simplicity of current.
implementation, fast transient response, suitable stability and
high accuracy. However, a conventional hysteresis band
 =  − 1 (1)

978-1-4673-1172-4/12/$31.00 ©2012 IEEE 324

 Loads
Control blocks

Source

Active power filter

Figure 1. Shunt APF principle.

This paper uses the instantaneous reactive power theory In this paper the control scheme is based on the usage of
(IRTP), which was proposed by Akagi [5] for three-phase two self-tuning filters [6]. The reference currents iαβ* are
systems with or without neutral wire, in steady state or in calculated by (7):
transients. It consists in the algebraic transformation of the
current and voltage of the system from the abc system to αβ0 ∗ 1 0 1  + ''
! ∗ # = % ! # + % ! # (7)
system using the Clarke transformation, as in (2) and (3): Δ & Δ 

⎡ 1 − 12 − 12 ⎤ where:

  Δ= &&&2 + &&&
 2
2⎢ √3 √3 ⎥  
 = 0 − (2) &&& −
 &&&
3⎢ 2 2⎥ 
0 ⎢√2  %=! #
√2 √2 ⎥  &&& 
 &&&
⎣ 2 2 2 ⎦
& = mean value of imaginary power. It corresponds to the
⎡ 1 − 12 − 12 ⎤ power exchanged between the phases of the load and is
 
2⎢ √3 √3 ⎥  responsible for the existence of undesired current. It must
  = 0 2 −

2⎥    (3)
3⎢ be compensated.
0 ⎢√2 
√2 √2 ⎥  = alternating value of imaginary power. It corresponds to the
⎣ 2 2 2 ⎦ conventional reactive power. It can be compensated by the
APF, depending on the requirements of the system.
According to the IRPT theory, the active, reactive and
 = alternating value of instantaneous real power. Since it
zero-sequence powers are defined as in (4a and 4b) and (5):
does not involve any energy transfer from the source to the
load, it must be compensated [7].
 =  +  (4a)
The reference currents are transformed from three-phase
 =  −  (4b) abc to αβ0 components using inverse Clarke transformation,
0 = 0 0 (5) as in (8):
The currents, voltages and powers in the α-β system can be
⎡ 1 0 √2 ⎤
decomposed in mean and alternating values, corresponding to ∗ 2 ∗
2⎢ ⎥
the fundamental and harmonic components, as in (6): ∗ = 1 √3 √2  ∗
3 ⎢− 2
(8)
2 2⎥ ∗
∗ ⎢ ⎥ 0
 = ̅ +  (6) 1 √3 √2
⎣− 2 − 2 2⎦
where x can be currents, voltages or powers.

325
 The control strategy based on the IRPT theory using self-
tuning filter is shown in Fig. 2.

Figure 2. Control strategy.

III. PROPOSED CONTROL SCHEME Fuzzy HB H(n) Hysteresis pulses

1 if(n-1) Δif(n) calculation controller
The conventional hysteresis band (HB) current control z

technique has proven to be most suitable for controlling APFs. if(n)

The HB current controller decides the switching logic pattern
of the APF, according to: if*(n)
If if < (if* – HB) upper switch is OFF, lower switch is ON. Figure 3. Proposed fuzzy adaptive hysteresis band current control.
If if > (if* + HB) upper switch is ON, lower switch is OFF.
where: if is the measured APF current; From (9) and (10) one can see that at each sampling period
if* is the reference APF current. the HB is updated according to the value of the actual filter
The switching frequency of the hysteresis band current current and current change. Based on this new fuzzy HB
control depends on how fast the current changes from the calculation, the current of the APF follows the reference
upper limit to the lower limit of the hysteresis band or vice current better, resulting in less harmonic distortion in the ac
versa. By increasing the switching frequency, one can get a mains, as we will indicate in the simulation results.
better compensating current waveform at the cost of increased The variable HB made to control the switching frequency
switching losses, audible noise, sub-harmonics and difficulties constant is the output of the fuzzy controller. The inputs of the
to design the input filter. In order to overcome these fuzzy controller are: the filter current if(n) at the nth sample
drawbacks, various adaptive hysteresis band current controls and the current change Δif(n) at the nth instant. Therefore, a
were proposed [1-4]. rule base is needed to satisfy (8).
This paper proposes a new adaptive hysteresis band current Fig. 4 and 5 shows the normalized triangular membership
control for three phase shunt active power filters, where the functions of the inputs and output variables, with the linguistic
hysteresis band is determined by using a fuzzy logic codes: NL (negative large), NS (negative small), Z (zero), PS
controller. (positive small), PL (positive large), VS (very small), S
The design of the adaptive hysteresis band current control (small), M (medium), L (large), VL (very large).
depends only on two main parameters. One of them is the As both inputs, if(n) and Δif(n), have five membership
instantaneous current at a certain sample period (n, n-1), while functions, the number of possible combination is 25, and the
the other is the value of the current change during this period. fuzzy rule base formulated for the APF control is shown in
This way, the control algorithm uses the previous switching table I.
cycles to forward estimate the future compensating current.
The variable hysteresis band for each sampling instant is
presented in Fig. 3, and is as follows:

H(n) = k1if(n) + k2Δif(n) (9)

Δif(n) = if(n) – if(n-1) (10)

where: H(n) represents the hysteresis band at the nth

sampling instant, if(n) represents the APF current at nth
sampling instant, Δif(n) is the current change at the nth instant,
and if(n-1) is the current at the (n-1)th sampling instant. Figure 4. Membership functions for input variables if(n) and Δif(n).
Variable k1 and k2 are calculated by the fuzzy logic controller.

326
 practically sinusoidal and smoothed comparing to Fig. 7, thus
the performance of the active filter with the proposed control
algorithm is found to be excellent. In addition, the 5th
harmonic has a value of only 0.14% and the 7th of only 0.24%.

Current (A)
Figure 5. Membership functions for output variable HB(n).

TABLE I. FUZZY CONTROL RULE TABLE

if NL NS Z PS PL
Δif
NL VS S S M M Figure 8. Source current under nonlinear balanced load with proposed fuzzy
NS S S S M M controller.
Z L L VL L L
PS M M S S S
PL M M S S VS
Fig. 9 shows the simulation results obtained in the
harmonic distortion analysis of the load currents, for each
phase, with nonlinear and unbalanced load.
IV. SIMULATION RESULTS
Fig. 6 shows the simulation results obtained in the
Current (A)
harmonic distortion analysis of the load current, with
nonlinear balanced loads. The total harmonic distortion (THD)
is 26.86%. The highest harmonics are the 5th and the 7th,
representing 20.83% and 12.12% of the fundamental.
Current (A)

Current (A)

Figure 6. Load current under balanced nonlinear load.

By using APF, the THD of the source current is reduced

Current (A)

from 26.86% to 2.24%, thus meeting the limit of the harmonic

standard of IEEE STD. 519-1992, as seen in Fig. 7. The
highest harmonics are still the 5th and the 7th, but now they
represent only 0.17% and 0.29% of the fundamental, which
meets the harmonic standard of IEEE STD. 519-1992.
Figure 9. Load current in phase A,B and C under nonlinear unbalanced load.
Current (A)

The THD in phase A is 41.75% with high 3rd and 7th

harmonics, which have over 20% of the fundamental value. In
phase B the THD is 19.63% with a high 5th harmonic –
15.41% of the fundamental value. Phase C has a THD of
20.10%, with a high 5th harmonic – 15.88%.
The source currents after the compensation can be seen in
Figure 7. Source current under nonlinear balanced load. Fig. 10.

To further improve the performances of the system, a last

test was made with the proposed Fuzzy adaptive hysteresis
band current controller. Fig. 8 shows the source current, with a
THD of only 1.13%. It can be seen that the source current is

327
 Current (A)

Current (A)
Current (A)

Current (A)
Figure 11. Source current in phase A,B and C under nonlinear unbalanced
load with proposed fuzzy controller.
Current (A)

In order to be effective, APF must also eliminate the

neutral current from three-phase unbalanced loads. Fig. 12
shows that even when connecting at 0.6s the unbalanced load
the neutral current is close to 0A.

Figure 10. Source current in phase A,B and C under nonlinear unbalanced
load.

In phase A the THD is now 2.66%, and the magnitude of

the 3rd harmonic is now only 1.79% of the fundamental. In
Current (A)

phase B the THD is 2.11% and in phase C the THD is 2.28%,

thus meeting the harmonic standard of IEEE STD. 519-1992.
To further improve the performances of the system, a last
test was made with the proposed Fuzzy adaptive hysteresis
band current controller. Fig. 11 shows the source current, with
a THD of only 1.13%. It can be seen that the source current is
practically sinusoidal and smoothed comparing to Fig. 7, thus
the performance of the active filter with the proposed control
Time (s)
algorithm is found to be excellent. In addition, the 5th
Figure 12. Neutral current elimination with proposed fuzzy controller.
harmonic has a value of only 0.14% and the 7th of only 0.24%.
Fig. 13 shows that when connecting the APF at 0.3s the
reactive power decreases close to zero, even when the
Current (A)

unbalanced load is connected at 0.6s, proven that the APF is a

very effective tool to compensate reactive power with the
proposed adaptive hysteresis Fuzzy controller.

328
 The simulation results verify the proposed method, with a
Power (W)

quick response time, switching frequency nearly constant,

reactive power close to zero and good quality of filtering.

ACKNOWLEDGMENT
Active Power This paper was realized with the support of POSDRU
CUANTUMDOC “DOCTORAL STUDIES FOR
EUROPEAN PERFORMANCES IN RESEARCH AND
INOVATION” ID79407 project funded by the European
Reactive Power Social Found and Romanian Government.
REFERENCES

[1] M. Kale, E. Ozdemir, “An adaptive hysteresis band current controller for
shunt active power filter”, Electric Power Systems Research 73, 2005,
pp. 113-119.
[2] S. Kerruche, F. Krim, “Three-phase Active Filter Based on Fuzzy Logic
Time (s)
Controller”, International Journal of Sciences and Techniques of
Figure 13. Reactive power compensation. Automatic control & computer engineering, vol. 3, no. 1, 2009, pp. 942-
955.
[3] B. Mazari, F. Mekri, “Fuzzy Hysteresis Control and Parameter
V. CONCLUSIONS Optimization of a Shunt Active Power Filter”, Journal of Information
Science and Engineering 21, 2005, pp. 1139-1156.
Fixed hysteresis band current control techniques has been
[4] P. Rathika, Dr. D. Devaraj, “Fuzzy Logic – Based Approach for
proven to be most suitable method for controlling active Adaptive Hysteresis Band and Dc Voltage Control in Shunt Active
power filters line current, at the cost of variable switching Filter”, International Journal of Computer and Electrical Engineering,
frequency. In order to overcome the drawbacks of variable vol. 2, no. 3, 2010, pp. 1793-1863.
switching frequency, various adaptive hysteresis band current [5] Akagi, H.; Kanazawa, Y.; Nabae, A. 1983. „Generalized Theory of the
Instantaneous Reactive Power in Three-Phase Circuits”, in Proc. IPEC-
control were proposed over the years. Tokyo’83 Int. Conf. Power Electronics, Tokyo, pp. 1375-1386.
This paper proposes a new fuzzy adaptive hysteresis band [6] Abdusalam, M.; Poure, P.; Karimi, S.; and Saadate, S. 2009. “New
current control for shunt active power filter. The simulations digital reference current generation for shunt active power filter under
of the proposed system were carried in Matlab distorted voltage conditions” in Electric Power Systems Research 79, pp.
SimPowerSystems environment. 759-765.
[7] Afonso, J. A.l Freitas, N.J.S.; Martins, J. S. 2003. „p-q Theory Power
Components Calculations”, IEEE International Symposium on Industrial
Electronic, Rio de Janeiro, Brasil.

329