Professional Documents
Culture Documents
1 s2.0 S0142061506000755 Main
1 s2.0 S0142061506000755 Main
www.elsevier.com/locate/ijepes
a
Electrical Engineering Department, Faculty of Engineering, Menoufiya University, Shebin El-Kom, Egypt
b
International Steel Rolling Mill Company (ISRM), Sadat, Egypt
Received 18 December 2003; received in revised form 15 December 2005; accepted 18 January 2006
Abstract
This paper presents a shunt active power filter to compensate reactive power and reduce the unwanted harmonics. A shunt active filter
is realized employing three-phase voltage source inverter (VSI) bridge with common DC bus capacitor. The shunt active filter acts as a
current source, which is connected in parallel with a nonlinear load and controlled to generate the required compensation currents. Two
different proposed control methods for determining the reference compensating currents of the three-phase shunt active power filters
based on proportional-integral (PI) controller and artificial neural network (ANN) are presented. Current controller based on modified
hysteresis current controller is used to generate the firing pulses. The proposed system is implemented using a high speed Digital Signal
Processor (DSP). Experimental and simulation results with different nonlinear loads are presented to confirm the validity of the proposed
technique.
Ó 2006 Elsevier Ltd. All rights reserved.
0142-0615/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2006.01.007
538 E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547
This paper presents two different approaches used to the reference currents if1 , if2 and if3 through the control
calculate the reference current of the SAF. The first tech- circuit.
nique used the DC capacitor voltage with PI controller.
Although, this method is simple to implement and achieves 2.2. System modeling
good results, it had drawbacks at unbalance conditions of
the supply voltages. The performance of the SAF is The representation of a three-phase voltages and cur-
achieved by the second technique, which depends on an rents of the VSI in Fig. 1 are as follows: the voltages Vf1,
ANN. The output of the ANN is used to generate the ref- Vf2 and Vf3 supplied by the inverter as a function of the
erence currents of SAF. The basic objective of this control capacitor voltage, Vc, and the state of the switches G1, G3
is to provide a very precise solution to get the reference cur- and G5 are:
rent even under unbalanced conditions of supply voltages. 2 3 2 32 3
A hysteresis controller based on PWM current control is V f1 2 1 1 G1
used, which is widely used due to its fast response. Com- 6 7 Vc6 76 7
6 V f2 7 ¼ 6 1 2 1 76 G3 7 ð1Þ
parisons of simulation results are presented for the two 4 5 6 4 54 5
control strategies. The proposed control system with V f3 1 1 2 G5
ANN is implemented using a DSP. Experimental results
were presented to prove the effectiveness of the design of where G1, G3 and G5 represent three logic variables of the
the control strategy. three legs of the inverter. The inverter conduction state is
represented by these logics. Then the SAF currents can
2. System description and operation be written as:
dif1
2.1. Power circuit description Lf1 ¼ V s1 V f1 ð2Þ
dt
dif2
As shown in Fig. 1, the SAF system consists of a three- Lf2 ¼ V s2 V f2 ð3Þ
phase voltage inverter with current regulation, which is dt
used to inject the compensating current into the power line. dif3
Lf3 ¼ V s3 V f3 ð4Þ
The VSI contains a three-phase isolated gate bipolar tran- dt
sistors (IGBT) with anti-paralleling diodes. The VSI is con-
where if1, if2 and if3 are SAF currents and Vs1, Vs2 and Vs3
nected in Parallel with the three-phase supply through
are the supply voltages. The voltage in the DC capacitor
three inductors Lf1, Lf2 and Lf3. The DC side of the VSI
can be calculated from the SAF currents and switching
is connected to a DC capacitor, C, that carries the input
function as follows:
ripple current of the inverter and the main reactive energy
storage element. The DC capacitor provides a constant DC Z
1
voltage and the real power necessary to cover the losses of Vc ¼ ½G1 if1 þ G3 if2 þ G5 if3 ð5Þ
C
the system. The inductors Lf1, Lf2 and Lf3 perform the
voltage boost operation in combination with the capacitor, The set point of the storing capacitor voltage must be
and at the same time act as the low pass filter for the AC greater than the peak value of the line neutral mains volt-
source current. Then the SAF must be controlled to pro- age in order to be able to shape properly the mains
duce the compensating currents if1, if2 and if3 following currents.
3-phase S1 S3 S5
AC supply
Lf1 D1 D3 D5
if1
Lf2 if2 C
if3
Lf3 D4 D6 D2
S4 S6 S2
if Control circuit
Vc
Vs
iL
Non linear
Load
Hysteresis
iL1 if1 current
_ controller G1
VS1 _
X i*S1 i*f1
G4
+ +
Vc* iL2
VS2 if2
_ G3
_
Distribution circuit
i*S2 i*f2
Vc X G6
PI + +
iL(t)
W +
X if (t)
Vs (t) ANN
ip (t)
Vs
X
3. Simulation results
W
A number of simulation results with different operating
Z-1 + conditions were developed. In Fig. 5, SAF is connected in
+
X η Z-1 parallel with nonlinear load. Then, the three-phase con-
trolled rectifier with resistive load has to be compensated
by the SAF. In order to limit maximum slope of the rectifier
Fig. 4. The proposed control method for phase-a. current, a smoothing inductor, Lr, has been inserted before
the rectifier. Also, it is prevented the inverter saturation even
in correspondence of rectifier commutations. Since the com-
From Eqs. (9) and (10), the proposed control method for pensation of the filter depends upon the firing angle (a) of
phase-a is designed, as shown in Fig. 4. the rectifier, two operative conditions have been considered.
The neural network learning and training by Widrow- The uncontrolled rectifier is the first case, while the second
Hoff learning rule is used to obtain the weight (W). Then one is the controlled rectifier with firing angle a = 40°. The
the SAF reference current can be calculated from Eq. parameters of system are reported in Appendix.
(10). After calculating the reference current for the SAF,
the hysteresis current controller has been presented to pro- 3.1. Simulation with uncontrolled rectifier (a = 0°)
vide the signals pulses of the VSI that controlled SAF
currents. 3.1.1. Steady state operations
The simulation results in steady state operation are pre-
2.6. Hysteresis current controller sented. Fig. 6 shows the performance of the SAF system
using PI controller. Waveform of the source current with-
Hysteresis current controller with fixed band derives the out SAF is shown in Fig. 6(a). Fig. 6(b) shows the source
switching signals of the inverter from the comparison of the current with SAF superimposed to the supply voltage.
current error to keep the current within the hysteresis band. Fig. 6(c) shows the compensation current. The capacitor
Then the SAF reference currents if1 , if2 , if3 compared with voltage superimposed to its reference is shown in
SAF feedback currents if1, if2, if3 and the error signals are
operated by the hysteresis current controller to generate
the firing pulses which activate the inverter power switches α
in a manner that reduces the current error. Let hb be the
width of the hysteresis band around respective phase 3-phase AC supply
SAF currents, then the equation for the fixed band as is it Lr iL RL
follows:
if1 ¼ imax sinðxtÞ ð11Þ Lf
iup ¼ if1 þ hb ð12Þ Controlled rectifier
if
ilow ¼ if1 hh ð13Þ
Where iup is the upper band, ilow is the lower band. If the
SAF current feedback if1 > ðif1 þ hb Þ then G1 = 1, which
means that the inverter output voltage switches to negative,
in order to keep the current within the hysteresis band. If
C
the SAF current feedback if1 < ðif1 hb Þ then G1 = 1,
which means that the inverter output voltage switches to Fig. 5. SAF connected in parallel with nonlinear loads.
E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547 541
10
Magnitude in Amp.
(a)
5
-5
-10
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
Time (s)
(5 Amp/div- 200 volt/div)
10
Vs (b)
5 is
Magnitude
-5
-10
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
Time (s)
5
Magnitude in Amp.
(c)
-5
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
Time (s)
1000
Magnitude in volt
(d)
500
0
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
Time (s)
Fig. 6. Simulation results of proposed SAF using PI controller. (a) Source current without filter. (b) Source current and source voltage with filter.
(c) Compensating current. (d) Capacitor voltage with its reference.
(10 Amp/div- 400 volt/div)
10
Vs (a)
is
Magnitude
-10
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
Time (s)
5
Magnitude in Amp.
(b)
-5
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
Time (s)
Fig. 7. Simulation results of proposed SAF using ANN. (a) Source current with SAF. (b) Compensation current.
542 E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547
Fig. 6(d). It is noticed that the supply current in phase with current. This figure indicates that the supply current in
the supply voltage, and the capacitor voltage follow its ref- phase with supply voltage. In order evaluate the good per-
erence. Fig. 7 shows the performance of the SAF using formance of the control, total harmonic distortion (THD)
ANN algorithm. The source current with the supply volt- is calculated for the source current before and after com-
age is shown in Fig. 7(a). Fig. 7(b) shows the compensation pensating. The calculation shows that THD improved from
10
(a)
Magnitude in Amp.
-5
-10
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Time (s)
(10 Amp/div- 200 volt/div)
20
is Vs (b)
10
Magnitude
-10
-20
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Time (s)
1000
(c)
Magnitude in volt
500
0
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Time (s)
Fig. 8. Simulation results for a step change in load using PI controller. (a) Source current without filter. (b) Source current and source voltage with filter.
(c) Capacitor voltage with its reference.
(20 Amp/div- 400 volt/div)
20
Vs (a)
is
Magnitude
-20
0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8
Time (s)
10
(b)
Magnitude
-10
0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8
Time (s)
Fig. 9. Simulation results for a step change in load using ANN. (a) Source current and source voltage. (b) Weight magnitude.
E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547 543
24.04 % without the SAF to 4.6156 % for using PI control- 3.1.2. Step change in load
ler and 2.9156 % for using ANN control with the SAF. The To observe the regulating process in two control meth-
distortion in supply current with ANN is less than in case ods, the simulation has been studied under the step change
of PI controller method. of the nonlinear load current. Fig. 8 shows the source
10
Magnitude in Amp.
(a)
5
-5
-10
0.4 0.41 0.42 0.43 0.44 0.45 0.46 0.47 0.48 0.49 0.5
Time (s)
(5 Amp/div- 200 volt/div)
20
is Vs (b)
10
Magnitude
-10
-20
0.4 0.41 0.42 0.43 0.44 0.45 0.46 0.47 0.48 0.49 0.5
Time (s)
10
Magnitude in Amp.
(c)
5
-5
-10
0.4 0.41 0.42 0.43 0.44 0.45 0.46 0.47 0.48 0.49 0.5
Time (s)
Fig. 10. Simulation results of proposed SAF using PI controller. (a) Source current without SAF. (b) Source current and source voltage with filter.
(c) Compensating current.
(5 Amp/div-200 volt/div)
10
is Vs (a)
Magnitude
-10
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
Time (s)
10
Magnitude in Amp.
(b)
-10
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
Time (s)
Fig. 11. Simulation results of proposed SAF using ANN. (a) Source current and source voltage. (b) Compensating current.
544 E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547
current without SAF, the source current with SAF super- using PI controller. Fig. 11 shows the same waveforms for
imposed to the supply voltage, and the DC capacitor volt- ANN control besides the weight function W. The final val-
age in case PI controller. Once the time interval of transient ues of THD of supply current before and after compensat-
is finished the DC capacitor can be recovered to be refer- ing are listed in Table 1. The supply current has improved
ence value, and vice versa. Fig. 9 gives the simulation significantly when using the ANN algorithm.
results in case of ANN algorithm. The supply current
and voltage are in phase during the load step change, as 3.3. Simulation results with unbalance in voltage
shown in Fig. 9(a). Fig. 9(b) shows the weight magnitude
will return to its value as the step change period finished. 3.3.1. PI controller
Fig. 12 shows the simulation results under the unbalance
3.2. Simulation with firing angle a = 40° three-phase supply voltages. The supply voltage and cur-
rent without SAF shown in Fig. 12(a). Fig. 12(b) shows
The simulation results in case of steady state operation the SAF current superimposed to its reference current.
with static load when a = 40° are presented. Fig. 10 shows The capacitor voltage and its reference is shown in
the simulation waveforms of the source current without Fig. 12(c). In order evaluate the good performance of the
SAF, the source current with SAF superimposed to the control, total harmonic distortion (THD) is calculated for
supply voltage, the SAF current, and the capacitor voltage the source current before and after compensating. The cal-
culation shows that THD improved from 25.7857% with-
Table 1 out SAF to 15.4395% for using SAF with PI controller.
Source current total harmonic distortion: THD%
Without SAF With SAF With SAF (ANN) 3.3.2. ANN algorithm
(PI controller)
Fig. 13, shows the simulation results when the supply
a = 0° THD% = 24.04 THD% = 4.6156 THD% = 2.9156 voltages are unbalance. The results have been obtained
a = 40° THD% = 25.1115 THD% = 5.58 THD% = 2.9407
under the same pervious condition of the PI controller.
(5 Amp/div-200 volt/div)
Vs (a)
is
Magnitude
10
(b)
If*
Magnitude in Amp
0
If
-5
-10
0.3 0.32 0.34 0.36 0.38 0.4
Time (s)
1000
(c)
Magnitude in volt
500
0
0.3 0.32 0.34 0.36 0.38 0.4
Time (s)
Fig. 12. Simulation results under unbalance three-phase supply voltage. (a) The supply current and voltage with SAF. (b) The SAF current composed to
its reference current. (c) The capacitor voltage composed to its reference.
E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547 545
Through the figures and calculation the THD of source voltage isolators sense the unit vector of supply voltages
current with SAF, the THD is reduced from 15.4395% and DC capacitor voltage. Also, the current isolators using
(value obtained by means of PI controller) to 6.918% (value LEM hall-effect current sensors sense the compensating
obtained by this control algorithm). current. The control scheme is implemented in real time
using a 32-bit digital signal processor (DSP-TMS320C31).
4. Experimental results The control algorithm generates three-phase reference sup-
ply currents, which are the output of DSP. The three-phase
Active power filter prototype for experimental purpose reference supply currents are fed to hysteresis PWM cur-
has been set up and tested. The prototype system is shown rent controller. The resulting pulses are gating the IGBT’s
in Fig. 14. The power circuit is constructed IGBT’s. The of the SAF.
(5 Amp/div-200 volt/div)
Vs
is
Magnitude
3-phase AC
supply
G1 G3 G5
Vc
G4 G6 G2
G1 G3 G5
Base drive
is1
is2 PWM
is3 Hysteresis current controller
i*s1 i*s2 i*s3
Vc
A/D Vs1
D/A converter Vs2
converte
Vs3
PC
Data-bus
computer
Reference
Control algorithm
generator Non-linear
DSP board (ds1102) load
Fig. 15. Experimental results of proposed SAF. (a) Source current and voltage without filter. (b) Source current, its reference and source voltage.
(c) Compensation current.
5. Conclusion
References
A control strategy for high performance three-phase [1] Dixon JW, Venegas G, Moran LA. A series active power filter based
shunt active filter with unity input power factor has been on a sinusoidal current controlled voltage-source inverter. IEEE T
proposed. Two control strategies are considered, the first Ind Electron 1997;44(5):612–7.
is PI controller and the second is based on ANN. The con- [2] Jeong S, Woo M. DSP-based active power filter with predictive
current control. IEEE T Ind Electron 1997;44(3):329–36.
trol scheme using three independent hysteresis current con-
[3] Buso S, Malesoni L, Mattavelli P. Comparison of current control
trollers has been implemented. The operation and techniques for active filter applications. IEEE T Ind Electron 1998;
modeling of the SAF have been described. 45(5):722–9.
An experimental SAF has been carried out on DSP to [4] Cheng P, Bhattacharya S, Divan D. Line harmonics reduction
explore the advantages and practical implementation with in high-power systems using square-wave inverters-based domi-
nant harmonic active filter. IEEE T Power Electr 1999;14(2):265–
the proposed control strategy. The experimental and simu-
72.
lation results showed that the proposed system has a good [5] Huang S, Wu J. A control algorithm for three-phase three-wired
performance in the harmonic compensation and improving active power filters under nonideal mains voltages. IEEE T Power
the input power factor. Electr 1999;14(4):753–60.
E.E. EL-Kholy et al. / Electrical Power and Energy Systems 28 (2006) 537–547 547
[6] Verdelho P, Marques G. Four-wire current-regulated PWM voltage [15] Akagi H, Kanazawa Y, Nabae A. Instantaneous reactive power
converter. IEEE T Ind Electron 1998;45(5):761–70. compensators comprising switching devices without energy storage
[7] Kamran F, Habetler T. Combined deatbeat control of a series- components. IEEE T Ind Appl 1984;IA-20:625–30.
parallel converter combination used as a universal power filter. IEEE [16] Saetieo S, Devaraj R, Tomey D. The design and implementation of a
T Power Electr 1998;13(1):160–8. three-phase active power filter based on sliding mode control. IEEE T
[8] Dastfan A, Gosbell V, Platt D. Control of a new active power filter Ind Appl 1995;31:993–1000.
using 3-d vector control. IEEE T Power Electr 1999;14(4):5–12. [17] Singh B, Al-Haddad K, Chandra A. Active power filter with sliding
[9] Chandra A, Singh B, Al-Haddad K. An improved control algorithm mode control. In: Proc Inst Elect Eng, Generation Transm Distrib;
of shunt active filter for voltage regulation, harmonic elimination, 1997, vol. 144. p. 564–8.
power factor correction, and balancing of nonlinear loads. IEEE T [18] Rastogi M, Mohan N, Edris A. Hybrid-active power filtering of
Power Electr 2000;15(6):495–503. harmonic currents in power systems. IEEE T Power Del 1995;10(3):
[10] Buso S, Malesani L, Mattavelli P, Veronese R. Design and fully digital 1994–2000.
control of parallel active filters for thyristor rectifiers to comply with [19] Bhattacharya S, Veliman A, Divan A, Lorenz R. Flux based active
IEC-1000-3-2 standards. IEEE T Ind Appl 1998;34(3): 508–17. power filter controller. In: Proc IEEE-IAS Annual Meeting Record;
[11] Fujita H, Akagi H. The unified power quality conditioner: the 1995. p. 2483–91.
integration of series-and shunt-active filters. IEEE T Power Electr [20] Jou H. Performance compression of the three-phase active power
1998;13(2):577–84. filter algorithms. In: Proc. Inst. Elect. Eng., Generation Transm.
[12] Salo M, Tussa H. A vector controlled current-source PWM rectifier Distrib; 1995, vol. 142. p. 646–52.
with a novel current damping method. IEEE T Power Electr 2000; [21] Dixon J, Garcia J, Moran I. Control system for three-phase
15(6):464–70. active power filter which simultaneously compensates power
[13] Barbosa P, Santisteban J, Watanabe E. Shunt-series active power factor and unbalanced loads. IEEE T Ind Electron 1995;42(6):636–
filter for rectifiers ac and dc sides. IEEE Proc Elec Power Appl 1998; 41.
145(6):577–84. [22] Saetieo S, Devaraj R, Tomey D. A new control approach to three-
[14] Peng F. Harmonic source and filtering approaches. IEEE T Ind Appl phase active power filter for harmonics and reactive power compen-
2001;7(4):18–25. sation. IEEE T Power Sys 1998;3(5):133–8.