International Journal on Electrical Engineering and Informatics ‐ Volume 3, Number 3, 2011

Intelligent Controllers for Shunt Active Filter to Compensate Current

Harmonics Based on SRF and SCR Control Strategies
Chennai Salim1 and Benchouia Mohamed Toufik2
1
Department of Electrical Engineering, Research Center of Birine, Algeria,
2
Laboratory L.G.E.B., University of Biskra, Algeria
chenaisalimov@yahoo.fr

Abstract: This paper presents a three-phase three-wire shunt active filter based on
intelligent controllers to compensate current harmonics. Shunt active filter is the best
solution to eliminate harmonics drawn from nonlinear load especially for low power
system, the most inverter used is the two-level voltage source. The control strategies
adopted to determine the compensation currents are the synchronous reference frame
(SRF) detection method and the synchronous current reference (SCR) detection method.
To improve the performance of the conventional control scheme based on hysteresis
controller and take advantages of intelligent techniques, fuzzy and ANN current
controllers are proposed in this paper. To maintain the dc voltage capacitor constant
and compensate the inverter losses a proportional integral voltage controller is used.
The numerical simulation is developed and performed using MATLAB-Simulink and
SimPowerSystemd Toolbox from complete structure including control and power
circuits. The obtained results show the simplicity and the effectiveness of the two
proposed intelligent current controllers.

Keywords: Shunt active filter, Intelligent controllers, ANNs, FLC, Current harmonics
compensation.

1. Introduction
With the proliferation of nonlinear power electronics loads, the problem of harmonic is
severity, which influences the power quality of power grid. Passive power filter is a traditional
harmonic restraint method. The passive filtering is a simple way to eliminate the harmonic
currents. However, it does not allow to completely eliminating all of them and has many
drawbacks such as series or parallel resonance with the system impedance [1].
Active filter is used, these last years, to improve power quality on the load side from the grid
current, by injecting compensating currents [2]. The performance of an active filter mainly
depends on the reference current generation strategy, technique control, topology of the filter
inverter, etc…
Principally, the design of any SAF system pass through three essentials criteria: power
inverter topology, current controller and strategy control. Instantaneous power theory [3] is
widely used in different research work, this technique provide good results under different
voltage source conditions but present some drawbacks such us a much calculations number,
necessitate complex mathematical transformation and difficult implementation in practice [4].
Synchronous reference frame (SRF) detection method [5] and Synchronous current reference
(SCR) detection method are two other interesting method; these methods are concise and
requires less computational calculation. In this work, these techniques are adopted.
The controller is the main part of the active power filter operation and has been a subject of
many researches in recent years. Multilevel pulse-width modulated (PWM) techniques have
been proposed for high power or medium-voltage applications such as reactive power
compensation and AC motor drives. The advantages of multilevel techniques are low voltage

Received: April 19th, 2011.  Accepted: August 25th, 2011

372
Chennai Salim, et al.

stress of power semiconductors, lower current or voltage harmonics, and less electromagnetic
interference. To improve the APF performances there’s a great tendency to use intelligent
control techniques, particularly neural network and fuzzy controllers [6],[7]. The advantages of
these controllers are: robustness, no need to accurate mathematical model, can work with
imprecise inputs and accept non-linearity.
In this paper, fuzzy logic and artificial neural network current controllers are proposed to
control the three-phase three-wire shunt active filter with two different control strategies:
synchronous reference frame detection algorithm (SRF) based on fuzzy controller and
synchronous current reference (SCR) detection method based on ANN controller.
The performances of shunt active filter based on these controllers are evaluated using
MATLAB-Simulink and power system Block Set Toolbox under balanced voltage conditions.
The obtained results show that the proposed shunt active filter based on these intelligent
controllers produce a sinusoidal supply current with acceptable low harmonic distortion.

2. Shunt active filter

The basic block diagram including non linear load compensation principle of a shunt active
power filter is shown in Figure 1. It is controlled to draw/supply a compensated current from/to
the utility, such that it cancels harmonic currents of the non-linear load and makes the source
current in phase with the different waveforms. The current drawn from the power system at the
coupling point of the shunt APF will result sinusoidal [8],[9].

Figure 1. Shunt active filter

3. Control strategies
A. Synchronous reference frame detection method
The first control strategy used in this work to compensate harmonic currents is based on the
synchronous reference frame detection method. The principle of this method is described
below [5]. The three phase currents i a , i b and i c are transformed from three phase ( abc )
reference frame to two phase’s (α − β ) stationary reference frame currents i α and i β using:
⎡ 1 1 ⎤⎡ ⎤
⎢1 − i
⎡i α ⎤ 2⎢ 2 2 ⎥⎥ ⎢ La ⎥
⎢ ⎥= i
⎢ Lb ⎥ (1)
⎣⎢i β ⎦⎥ 3⎢ 3 3⎥
⎢0 − ⎥ ⎢⎣ i Lc ⎥⎦
⎣ 2 2 ⎦

373
Intelligent Controllers for Shunt Active Filter to Compensate Current

Using a PLL (Phase Locked Loop), we can generate cos(θest ) and sin(θest ) from the phase
voltage source v sa , v sb , v sc .

The currents expression i α and i β in (d −q ) reference frame are given by:

⎡i d ⎤ ⎡ sin(θest ) − cos(θest ) ⎤ ⎡i α ⎤
⎢ ⎥=⎢ ⎥⎢ ⎥ (2)
⎣⎢ i q ⎦⎥ ⎣cos(θest ) sin(θest ) ⎦ ⎣⎢i β ⎦⎥

The DC quantities and all other harmonics are transformed to non DC quantities using a low
pass filter:
⎡_ ⎤
⎡i d ⎤ ⎢ i d + id ⎥
⎢ ⎥=⎢ ⎥ (3)
⎣⎢ i q ⎦⎥ ⎢ _ ⎥
+ i q ⎦⎥
⎣⎢ i q
The expression of the reference current i α − ref and i β − ref are given by:

⎡i α − ref ⎤ ⎡ sin(θest ) − cos(θest ) ⎤ −1 ⎡i d ⎤

⎢ ⎥=⎢ ⎥ ⎢ ⎥ (4)
⎢⎣i β − ref ⎥⎦ ⎣cos(θest ) sin(θest ) ⎦ ⎢⎣ i q ⎥⎦

⎡ ⎤
⎡i α − ref ⎤ ⎡ sin(θest ) cos(θest ) ⎤ ⎢ _
⎢ ⎥=⎢ ⎥ ⎢i d + id⎥ (5)
⎥
⎢⎣i β − ref ⎥⎦ ⎣ − cos(θest ) sin(θest ) ⎦ ⎢ i ⎥⎦
⎣ q
The reference currents in the ( abc ) frame is given by:

⎡ ⎤
⎢ 1 0 ⎥
⎡i a − ref ⎤ ⎢ ⎥
⎢ ⎥ 2⎢ 13 ⎥ ⎡⎢i α − ref ⎤⎥
⎢i b − ref ⎥= ⎢− ⎥ (6)
⎢ ⎥ 2 ⎥ ⎢ i β − ref ⎥
3⎢ 2
⎣ i c − ref ⎦ ⎣ ⎦
⎢ 1 3⎥
− ⎥
⎢
⎣ 22 ⎦
Finally, the compensation currents ( i ca , i cb and i cc ) are given by:
i ca = i a − ref − i La
i cb = i b − ref − i Lb (7)
i cc = i c − ref − i Lc

Figure 2 shows the principle scheme of the synchronous reference frame detection method.

374
Chennai Salim, et al.

Figure 2 Synchronous reference frame detection method

B. Synchronous current detection method

The second control strategy is inspired of the synchronous detection reference currents
method [10]. It is based on the measuring of the source voltages (V sa , V sb , V sc ) given by (8) :
v sa (t ) = V sm . sin(ωt )
2π
v sb (t ) = V sm . sin(ωt − 3 ) (8)
4π
v sa (t ) = V sm . sin(ωt − 3 )

*
The maximum amplitude of the current supplied by the filter I smax
is given by (9):
* * *
I s max = I smp + I smd (9)
With :
* 2 Pmoy
I smp
= (10)
3 V sm
*
I smd is the component current which maintains the DC voltage across the capacitor C cd .
The reference currents must be sinusoidal and in phase with the voltage source. The desired
current source AC can be calculated by multiplying the maximum amplitude of the current
source of sinusoidal signals unit. These unit signals are given by (11):
i ua (t ) = v sa /V sm
i ub (t ) = v sb /V sm (11)
i uc (t ) = v sc /V sm
* *
i sa
(t ) =I .i
sm ua
(12)
* *
i sb
(t ) =I .i
sm ub
* *
i sc (t ) = I sm .i uc

375
Intelligent Controllers for Shunt Active Filter to Compensate Current

The difference between the current references and the current drawn by the load can
* * *
generate the three compensation current ( i ca
,i cb
et i cc
):
* *
i ca
=i sa
−i
La
* *
i cb
=i sb
−i
Lb
(13)
* *
i cc = i sc − i Lc
Figure 3 shows the principle scheme of the synchronous current reference detection method.

Figure 3. Synchronous current reference detection method

4. DC voltage controller
The regulation loop consists of the comparison of the measured voltage with the reference
voltage, admitting that the function of the system to be controlled is given by [11-18]:
Vs2
(14)
(V dc − ref .C dc .s )

The closed loop transfer function using a PI regulator is given by:

V dc (K p + K i s ).(V s 2 (V dc − ref .C dc .s ))
= (15)
V dc − ref 1 + (K p + K i s ).(V s 2 (V dc − ref .C dc .s ))

The development of this equation gives:

K p .V s 2 Ki
.s +
V dc V dc − ref .C V dc − ref .C
= dc
2
dc
(16)
V dc − ref 2 K p .V s K i .V s 2
s + .s +
V dc − ref .C V dc − ref .C
dc dc

A second order characteristic equation of the closed loop system is deduced:

2 2
s + 2ξωn s + ωn = 0 (17)
Where:

376
Chennai Salim, et al.

K i .V s 2 K p .V s
ωn = ,ξ = (18)
V dc − ref .C dc 2 K i .V dc − ref .C dc

From (11) the proportional and integrator coefficient K p , K i of the controller can be deduced:
2
2.ξ .K i ωn V
.
dc − ref .C
dc
Kp = , Ki = (19)
ωn 2
Vs
The expression of the current I sc to compensate the inverter losses and maintain the
constant dc-link voltage is given by:
I sc = K p .ΔU dc + K i ∫ ΔU dc .dt (20)

To obtain optimal dynamic performance for the system, the value of the damping ration ξ must
be equal a 0.707.

5. Artificial intelligent controllers

A. Fuzzy logic current controller
The main component of an active filter is the current controller. Generally the classical
hysteresis controller is used to generate pulses to the PWM inverter; it is very stable and
generates a minimum noise. Recently, fuzzy logic controllers (FLCs) have been interest a good
alternative in more application. The advantage of fuzzy systems are that they do not need an
accurate mathematical model, they can work with imprecise inputs, can handle non-linearity,
and they are more robust than conventional nonlinear controllers [12],[13].
Fuzzy logic control is the evaluation of a set of simple linguistic rules to determine the
control action. To develop the rules of the fuzzy logic, we need good understand of the process
to be controlled, but it does not require a complicated mathematical model. The desired
switching signals for the filter inverter circuit are determined according to the error in the filter
current. In this case, the fuzzy logic current controller has two inputs, named error e and
change of error de and one output named s. Here the error e and change of error de are the
input variable for the system [18]. To convert it into linguistic variable, we use three fuzzy sets:
N (Negative), ZE (Zero) and P (Positive). Figure 4 shows the membership functions used in
fuzzification.

The fuzzy controller for every phase is characterized for the following:
ƒ Three fuzzy sets for each input,
ƒ Five fuzzy sets for each output,
ƒ Triangular and trapezoidal membership functions,
ƒ Implication using the “min” operator,
ƒ Mamdani fuzzy inference mechanism based on fuzzy implication.
ƒ Defuzzification using the “centroid” method.

377
Intelligent Controllers for Shunt Active Filter to Compensate Current

Figure. 4 Membership function for the inputs and output variables

The linguistic rules for the fuzzy current controller are as follows:
1. If error is Negative and error rate is Negative Then output is Big Negative,
2. If error is Zero and error rate is Negative Then output is Positive,
3. If error is Positive and error rate is Negative Then output is Big Positive,
4. If error is Negative and error rate is Negative Then output in Big Negative,
5. If error is Zero and error rate is Zero Then output is Zero,
6. If error is Positive and error rate is Zero Then output is Big Positive,
7. If error is Negative and error rate is Positive Big Then output is Big Negative,
8. If error is Zero and error rate is Positive Then output is Negative,
9. If error is Positive and error rate is Positive Then output is Big Positive.

The generation process of switching signals is given by Figure 5.

Figure 5. Generation process of switching signals

378
Chennai Salim, et al.

Figure 6 shows the MATLAB-Simulink simulation block diagram of the proposed fuzzy current controller for the th ree-phase shunt active filter.

Figure 6. Shunt active filter based on Fuzzy current controller

379
Intelligent Controllers for Shunt Active Filter to Compensate Current

B. Artificial neural network current controller

Artificial Neural Networks have provided an alternative modeling approach for power
system applications. The MLPN is one of the most popular topologies in use today. This
network consists of a set of input neurons, output neurons and one or more hidden layers of
intermediate neurons. Data flows into the network through the input layer, passes through the
hidden layers and finally flows out of the network through the output layer. The network thus
has a simple interpretation as a form of input-output model, with network weights as free
parameters. The training cycle has two distinct paths [14-15-16], the first one is Forward
propagation (it is the passing of inputs through the neural network structure to its output), the
second one is the error back-propagation (it is the passing of the output error to the input in
order to estimate the individual contribution of each weight in the network to the final output
error). The weights are then modified so as to reduce the output error. To train the neural
network current controller, the Quasi-Newton Levenberg-Marquardt Training algorithm is
used, it is efficient, easy to implement and is not time consuming.
Computations of the algorithm proceed as follows:
1) Initialize the interconnection weights and the biases of the nodes randomly,
2) Calculate the hidden layer outputs as:
h ⎛ ni ⎞
x = f ⎜ ∑ ( x i w ih, j )+b hj w ni
h
+1, j
⎟ (21)
j ⎜ ⎟
⎝ i =1 ⎠
h h
Where x is the output of the hidden node j, x i is the i'th input, w i , j is the weight connecting
j
h h
input node i with hidden node j, b j is the input bias to hidden node j (normally b j =1),
h
w
ni +1, j
is the weight connecting the bias to the hidden node j, ni is the number of input
nodes, and f is sigmoid function defined as:
1
f (x ) = (22)
1+e − x
3) Calculate the output layer outputs as:
⎛ nk ⎞
= f ⎜ ∑ ( x hj w oj ,k ) +bko w onk +1,k ⎟
o
x (23)
k ⎜ ⎟
⎝ j =1 ⎠
o o
Where x is the output of the output node k, w j ,k is the weight connecting the hidden node
k
o o o
j with output node k, b k is the input bias to the output node k (normally b k =1), w nk +1,k is
the weight connecting the bias to the output node k and nk is the number of hidden nodes.
o
4) Calculate δ of each of the output nodes as:
k
o o o T o
δ =x (1 − x )( x −x ) (24)
k k k k k
o
Where δ , is the error (target-output) at the output of the neuron multiplied by the derivative
k
T
of f(x), x is the target output (desired output) of the output node k.
k
h
5) Calculate δ each of the hidden nodes as follows:
j

h h h no o
δ =x (1 − x ) ∑ δo w (25)
j j j k =1 k j ,k

380
Chennai Salim, et al.

h
Where δ is the derivative of f(x) multiplied by the summation of the weights multiplied by
j
the output delta,
6) Adapt the weights of the output layer as:
o o o h o
w (t + 1) = w (t ) + ηδ x + αΔw (t ) (26)
j ,k j ,k k j j ,k
Where 0 < η < 1 is the learning constant, 0 < α < 1 is the momentum constant and,
Δw oj ,k (t ) = w oj ,k (t ) −w oj ,k (t − 1) (27)
7) Adapt the weights of the hidden layer as:
h h h h
w (t + 1) = w (t ) + ηδ x + αΔw (t ) (28)
i,j i,j j i i,j
Where,
h h h
Δw (t ) = w (t ) − w (t − 1) (29)
i,j i,j i,j
8) Repeat steps 1 to 7 until the error e is less than a prescribed small value ε .
no 2
ε = ∑ ⎛⎜ x T − x o ⎞
⎟ (30)
k =1 ⎝ k k ⎠

To be able to produce the correct output data, the network was trained with an improved
algorithm, during the learning process the error function was minimized with an increasing
number of training epochs.
The input pattern of the current controller network is the error values ( Ei a , Ei b , Ei c )
between the measured filter currents (i ha , i hb , i hc ) and the compensating reference currents
* * *
(i ca , i cb , i cc ) whereas the outputs values are the switching states Sa , Sb and Sc for every
phase. The hidden layer contains 20 neurons with a sigmoid activation function, whereas the
output layer contains three neurons with a linear activation function. The network was trained
using Levenberg-Marquardt back propagation algorithm, about 10000 training examples
obtained by simulation [17].

Figure 7 shows the MATLAB-Simulink simulation block diagram of the proposed ANN
controller for the three-phase shunt active filter.

381
Intelligent Controllers for Shunt Active Filter to Compensate Current

Figure.7 Shunt active filter using ANN current controller

382
Chennai Salim, et al.

6. Simulation results and discussion

The purpose of the simulation is to show the effectiveness of the shunt active filter using intelligent controllers (fuzzy logic and ANN current
controllers) to reducing the harmonic currents produced on the load side under ideal voltages conditions. The model parameters used for simulation are:
Voltage source Vs=220V, Frequency Fs=50Hz, Resistor Rs=0.1mΩ, Inductance Ls=0.0002mH, Resistor Rch =48.6Ω, Inductance Lch=40mH, DC
Voltage Udc-ref=700V, Capacitance Cdc=3000μF, Resistor Rc=0.27m Ω, Inductance Lc=0.8mH.

Figure 8 shows the voltage source and line current before compensation

400

300

200
Voltagesource(V)

100

-100

-200

-300

-400
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Time(s)

383
Intelligent Controllers for Shunt Active Filter to Compensate Current

15

10

5
Currentsourceisa(A)

-5

-10

-15
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time(s)

Figure 8. Voltage source and line current before compensation

384
Chennai Salim, et al.

0.9

0.8
ental

0.7
paredtothefundam

0.6

0.5

0.4
agcom

0.3
M

0.2

0.1

0
0 5 10 15 20 25 30 35 40 45 50
Harmonic order

Figure. 9 Source current spectrum before compensation

385
Intelligent Controllers for Shunt Active Filter to Compensate Current

The line current and its spectrum after compensation using fuzzy logic controller are represented in Figure 10 and Figure 11.

50

40

30
Currentsourceisa(A)

20

10

-10

-20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time (s)

Figure. 10 Source current using Fuzzy Logic controller

386
Chennai Salim, et al.

0.9

0.8
ental

0.7
paredtothefundam

0.6

0.5

0.4
agcom

0.3
M

0.2

0.1

0
0 5 10 15 20 25 30 35 40 45 50
Harmonic order

Figure. 11 Source current spectrum THD 2.28% using Fuzzy Logic controller

387
Intelligent Controllers for Shunt Active Filter to Compensate Current

Figure 12 shows the current and voltage source after compensation, figure 13 present the dc capacitor voltage when a step change in load current is
introduced between t1 = 0.25 s and t2 = 0.45 s.

400

300

200
Vsa(V)andIsa(A)

100

-100

-200

-300

-400
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Time (s)

Figure. 12 Current and source voltage using Fuzzy Logic controller

388
Chennai Salim, et al.

720

700

680
Cvoltage(V)

660

640
D

620

600

580
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Time(s)

Figure. 13 DC capacitor voltage

389
Intelligent Controllers for Shunt Active Filter to Compensate Current

The line current and its spectrum after compensation using artificial neural network current controller are represented in Figure 14 and Figure 15.

25

20

15

10
Courant sourceisa(A)

-5

-10

-15

-20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Temps (s)

Figure. 14 Source current using ANN controller

390
Chennai Salim, et al.

0.9

0.8
ental

0.7
plitudepar rapport aufondam

0.6

0.5

0.4
Am

0.3

0.2

0.1

0
0 5 10 15 20 25 30 35 40 45 50
Rang des harmoniques

Figure. 15 Source current spectrum THD 1.49% using ANN controller

By visualizing Figures 10 and 14, we can conclude the successful operation of the proposed shunt active filter based on two the intelligent current
controllers. Before the application of shunt active power filter, the source current is equal to non-linear load current; it is highly distorted and rich on
harmonics. The SAF start the compensation process instantly when it is connected to the non-linear load. The system passes through transient period of
0.01s necessary to attain the steady state of the source current. The dc voltage is maintained constant and equal to the reference value Udc-ref=700V
using PI voltage controller.

391
Intelligent Controllers for Shunt Active Filter to Compensate Current

The dc voltage presented in figure (13) pass through a transitional period of 0.04 s before
stabilization and reaches its reference Udc-ref=700V with moderate peak voltage
approximately equal to 8 V when a step change in load current is introduced between t1 = 0.25
s and t2 = 0.45 s. After compensation the THD is significantly reduced from 28.16% to 1.49%
for ANN and to 2.28% for fuzzy controller. Figure 12 demonstrate that the current source after
active filter application is sinusoidal and in phase with the voltage source. The performance of
the proposed shunt active filter based on ANN controller in terms of harmonics elimination and
reactive power compensation is better than FLC current controller. The THD values obtained
with the two controllers respect widely the IEEE standards Norms (THD ≤5%).

Conclusion
In this paper, a three-phase three-wire shunt active filter based on intelligent controllers
using two control strategies is proposed to compensate current harmonics. The first controller
based on FLC use the synchronous reference frame detection strategy. The second controller
based on ANNs use the synchronous current detection strategy. The dc-link voltage is regulated
using a proportional integral voltage controller with fast dynamic response in case of load
current variation. The Simulation results obtained show the robustness and the effectiveness of
the two proposed intelligent controllers. After compensation the source current is balanced,
sinusoidal and in phase with line voltage source. The harmonic spectrum shows that the THD
is very acceptable and respect IEEE standard Norms.

References
[1] Jidong Wang, “Simulation of three-phase three-wire shunt active power filter”,
International Journal of Sciences and techniques of automatic Control & Computer
engineering, Volume 3,N°1, pp. 942-955, (2009).
[2] Udom. Khruathep, Suittichai Premrudeepreechacharn, Yuttana Kumsuwan
: “Implementation of shunt active power filter using source voltage and source current
detection”, IEEE, pp.2364-2351, (2008).
[3] Reyes H. Herrera, Patricio Salemeron, Hoyosung Kim, “Instantaneous Reactive Power
Theory Applied to Active Power Filter Compensation: Different Approaches,
Assessment, and Experimental Results”, IEEE, Trans. on Industrial Electronics, pp. 184-
196, (2008)
[4] Chandra A, Singh B, Al-Haddad K, “An improved control algorithm of shunt active filter
for voltage regulation, harmonic elimination, power factor correction, and balancing
nonlinear loads”, IEEE T Power Electr, pp. 495-503, 2000.
[5] S. Bhattachrya, D. Divan, “Synchronous frame based controller implementation for a
hybrid series active filter systems”, IEEE, pp. 2531-1995, (1995).
[6] A. Hamadi, K. El-Haddad , S.Rahmani, H.Kankan, “Comparison of fuzzy logic and
proportional integral controller of voltage source active filter compensating current
harmonics and power factor”, IEEE, International conference on industrial technology
(ICIT), pp. 645-650, (2004).
[7] C. N. Bhende, S.Mishra, “TS-fuzzy-controlled active power filter for load compensation”,
IEEE, Transactions on Power Deliv,21, pp. 1459-1465, (2006).
[8] M. Sarra, K. Djazia, A. Chaoui and F. Krim; “Three-phase active power filter with
integrator proportional control”, 3rd International Conference on Electrical Engineering,
pp. 506-511, (2009).
[9] B. Sing, K. Haddad, A. Chandra, “A new control approach to three-phase active filter for
harmonics and reactive power compensation”, IEEE,Trans. Power Syst.13(1), pp. 133-
138, (1998).
[10] J. Mazumdar, R. G. Harley, G. K. Venayagamoorty, “Synchronous reference frame based
active filter current reference generation using neural networks”, IEEE, pp. 4404-4409,
(2008).

392
Chennai Salim, et al.

[11] A. Chaoui,
C F.Krimm, J. P. Gauberrt, L. Rambaullt “DPC controolled three-phaase active filterr
for power
p quality improvement””, Elsevier, Eleectrical Power and Energy Systems
S 30, pp.
476--485, (2008).
[12] Kerrrouche Soumiaa, Krim Fateh, “Three-phasee active powerr filter based on o fuzzy logicc
conttroller”, International Journnal of Sciencees and techniques of automaatic Control &
Com mputer engineering, Volume 3,N°1, 3 pp. 9422-955, (2009).
[13] Wennjin Day, Baoffu Wang, Youuhui Xi, “Novel fuzzy logicc controller forr active powerr
filterr”, International Conference on Computational Intelligennce for Measurrement Systemss
and Applications, (2009).
(
[14] Zouiidi, F. Fnaiech h, K. Haddadd, “Neural Nettwork three-phhase three-wiree shunt activee
filterr”, IEEE, pp. 5-10,
5 (2006).
[15] Abddelaziz Zouidi, Farhat Fnaiecch, Kamal Al Haddad, H “Neurral network coontrolled three--
phasse three-wire sh hunt active powwer filter,” IEEEE, ISIE, (20066).
[16] Jiangg You-hua, Chen Yong-Weei, “Neural nettwork control techniques off hybrid activee
powwer filter,” Intternational Coonference on artificial Inteelligence and computationall
intellligence, pp. 266-30, (2009).
[17] Chennnai Salim, Benchouia
B M-T T, Goléa.A, Zouzou
Z S. E “Application
“ d réseaux dee
des
neurrones dans la commande
c d’uun filtre actif parallèle
p en utilisant Matlab-Simulink”, 6thh
Interrnational Confference on Eleectrical Engineeering’2010, Batna, Algeria, 11-13 Octoberr
(20110).
[18] Chennnai Salim, Beenchouia M-T, Goléa. A, Zoouzou S. E “Fuuzzy logic currrent controllerr
for shunt active filter
f to comppensate harmonics currents based on AN NN dc voltagee
reguulator”, International Connference on Electrical Enngineering, ellectronics andd
autoomatics’2010, Bejaia,
B Algeriaa, 2-3 November (2010).

Chenn nai Salim Waas born in Biskra, Algeria, on February 03, 1969. Hee
obtain
ned his engineeering degree inn Electrotechniics from Biskraa University inn
1992. He was recruitted in 1993 as senior engineeer in power eleectronics in thee
Nucleaar Research Ceenter of Birine, Algeria. Sincce 2000, he hass been workingg
as reseearcher in the Electrical
E Engiineering Deparrtment. He obtaained his M.Scc
degreee in electrical engineering
e in 2009 from Médea Universitty, Algeria. Hee
is currrently workingg towards his PhD
P degree in Electrical Enggineering from m
Biskraa University, Algeria.
A His ressearch interestss are electricall drives, powerr
electroniccs, energy quallity, power systtems, and intellligent control.

Bench houia Moham med Toufik Waas born in Biskkra, Algeria. He H received hiss
Engineer degree inn Electrotechnnics and Maagister degree in electricall
engineeering from Biskra
B University in 1991 and 1998 resspectively. Hee
receiv
ved his Ph.D. inn electrical enngineering from m Biskra University in 2006..
Since 2001 he has held a teaching and research position
p in the Department off
Electrical Engineering of Biskra University,
U Allgeria. His research interestss
are in electrical drivees, power electtronics and pow
wer systems.

393