International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882

Volume 4, Issue 2, February 2015

Neural Network Controlled Hybrid Active Power Filter with

Distorted Mains for PMSM Drive
Shweta Mishra1, Satya Prakash Dubey2
1
2

(Department of Electrical Engineering, CSVTU University, Bhilai

(Department of Electrical Engineering, CSVTU University, Bhilai

ABSTRACT
The present paper describes dynamic neural network
controlled Hybrid Active Power Filter (HAPF) designed
for harmonic compensation under variable conditions. A
Neural Network based Controller is employed to extract
the fundamental component of current from nonsinusoidal and unbalanced source currents of the
considered supply system. Hybrid filter utilizing the
merits of both shunt passive and active filter for better
compensation performance is applied in this work.
Simulations are carried out using MATLAB/Simulink to
show effectiveness of proposed methodology. The
experimental results show excellent behavior of Neural
Network Controller and performances of HAPF under
various conditions.
Keywords - Artificial neural network (ANN) controller,
Hybrid active power filter, Permanent magnet
synchronous motor (PMSM), Total Harmonic Distortion
(THD)

I.

INTRODUCTION

Nowadays the power quality problems are mostly faced

by the equipments based on the power electronic devices
which are used by the industry. The need of these
devices is not only the high quality energy for proper
working but also the major cause for decreasing of the
power quality. By these conditions of the electric power
the customers are increasingly affected. Therefore, a
power quality problem exists if any of the voltage,
current or frequency deviation from sinusoidal nature
occurs. Power quality problems are common in
industrial, commercial and utility networks as power
electronics appliances are widely used in these fields.
These appliances generate harmonics and reactive
power. Therefore it is very important to compensate the
dominant harmonics and thus Total Harmonic Distortion
(THD) below 5% as specified in IEEE 519 harmonic
standard [1]. On the improvement of power quality,
utilities and researches all over the world have for
decades worked. To the power quality problems there
are many sets of conventional solutions which have

existed for long time. Nowadays the equipment made up

of semiconductor devices is sensitive as well as
polluting. Increase in reactive power demand by the
equivalent load is due to the non linear devices like
power electronic converters which also results in
injection of harmonic currents in distribution grid. As it
is well known that reactive power demand increases the
losses and cause a drop in feeder, which also results in
additional losses, voltage waveform distortion and poor
power quality. So it is necessary to include some sort of
compensation to keep the power quality within limits
proposed by standards.
The shunt active power filter (APF) is a device that is
connected in parallel and cancels the reactive and
harmonic currents from a nonlinear load. The resulting
total current drawn from ac main is sinusoidal. Ideally,
the APF needs to generate the reactive and harmonic
current to compensate the nonlinear loads in the line. In
an APF depicted, current controlled voltage source
inverter is used to generate the compensating current and
is injected into utility power source grid. It cancels the
harmonic components drawn by nonlinear load and
keeps the utility line current sinusoidal. A voltage-source
inverter (VSI) having IGBT switches and an energy
storage capacitor on dc bus are implemented as a shunt
APF. The main aim of the APF is to compensate the
harmonics and reactive power [3].
Active power filter is a dynamic and flexible solution
for mitigation of harmonic current due to their compact
size, no requirement of tuning and stable operation.
Active power filter acts as harmonic current source to
provide the emphatic result to compensate for harmonic
currents as well as reactive power. It has the capability to
inject harmonic current into ac system with the same
amplitude but in opposite phase of the load [2]. As the
HAPF is complex with cost effective parameter control,
the hybrid active power filter has been preferable in
subject of harmonic solution. HAPF gives the
efficacious combination of passive and active filter,
which implies the advantages of both and eliminates the
short-comes of each one. [4].

www.ijsret.org

126

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 2, February 2015

In this paper, a new hybrid filter topology is proposed to

reduce torque ripples, voltage harmonics in PMSM. The
filter topology consists of the IGBT active filter (AF).
Power circuit for APF is proposed as an IGBT based
three-phase voltage source inverter with DC storage
capacitor for the better compensation of non-linear
unbalanced /balanced loads. ANN controller is used to
generate the fundamental from non-ideal voltage source.
The extracted fundamental currents are then subtracted
from the source current to evaluate the reference signal
i.e. harmonic current. The proposed controller has selflearning with high accuracy and simple architecture and
it can be successfully applied for the harmonic filtering
under various power system operating conditions.
Therefore, this paper presents a hybrid power filter using
neural network-controller to control the harmonics under
different non-sinusoidal and unbalanced source/load
conditions for its performance [5].

II.

PROPOSED TOPOLOGY

Three phase three wire hybrid power filter is used; Shunt

active power filter is used to generate the compensation
current in opposite phase. Power circuit for APF is
proposed as an IGBT based three-phase voltage source
inverter with DC storage capacitor for the better
compensation of loads. In this paper the load is
Permanent Magnet Synchronous Motor (PMSM). Active
power filter has control scheme; neutral network
controller that accounts for reference current generation
.ANN comprises three adaptive linear neurons to extract
the fundamental components of three phase voltages
from non-sinusoidal supply. The capacitors are designed
to limit the dc voltage ripple to a specified value,
typically 1 to 2 %. In this case the capacitor should be
designed for worst case. Since the active filter will
operate in several modes (balanced or unbalanced load),
then the injection of compensation current is done in
order to nullify or mitigate the harmonic currents.
Injection of this compensation current gives improved
power quality. The performance of the active power
filters is dependent to a great extent upon the method
used for the calculation of reference current.

III.

The most commonly used ANN structures for

fundamental and/or harmonic extraction are the
ADALINE and feed-forward MNN. The focus of this
paper is on how to implement the ANN techniques in
power system. Power quality enhancement using shunt
APF is considered for the effective illustration. The
architecture of proposed ADALINE neural network has
two layer (input and output) network having n-inputs
and a single output [6].

Fig. 1: Proposed topology of Shunt Active Power Filter

CONTROL STRATEGY

1. Neural Network Controller

ANNs have the ability to approximate the inputoutput
relations with acceptable accuracy regardless of the
complexity involved. This, in addition to many other
abilities of parallel computation and adaptiveness to
external disturbances, has made ANN a promising
approximation technique in power system applications.

Fig. 2: Internal block of neural network

www.ijsret.org

127

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 2, February 2015
50
40
30
20

Current(A)

10
0
-10
-20
-30
-40
-50
0

IV.

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.1

(b)
40
30
20
10
0

Selected signal: 5 cycles. FFT window (in red): 2 cycles

-10

15

-20

10

-30

5
-40

0
-50
0

-5

0.01

0.02

0.03

0.02

0.04

0.05
Time(sec)

0.06

0.07

0.04
0.06
Time (s)

(c)

0.08

0.09

0.08

0.1

0.1

Fundamental (50Hz) = 7.833 , THD= 1.84%

Mag (% of Fundamental)

0.3
0.25
0.2
0.15
0.1
0.05
0

200

400
600
Frequency (Hz)

800

1000

(d)
10

Torque(Nm)

Where b is bias, w is weight, and i is the input to

the Neural Network. The input to the network is a time
delayed series of the signal whose fundamental
component is to be extracted. The length of this delay
series is 61, which has been decided considering
expected maximum distortion and unbalance in 3-phase
input signal [4]. The input of the ANN system is supply
voltage and current and the output of the system is APF
reference current. The weight adjustment is performed
during the training process of the ADALINE using
Widrow-Hoff delta rule [6]. The ADALINE is a twolayer neural network consisting of an input layer and
output layer. It was invented by Widrow and Hoff when
they were trying to implement their famous training
algorithm, the least mean square (LMS). It can estimate
functions that have linear inputoutput relations, and can
work fairly even with some nonlinear real-world
applications [6].

0.02

50

Current(A)

Fig. 3: Input/output relationship of transfer function

The basic blocks of this network are input signal delay
vector, a purelin transfer function, weight matrix and
bias is shown in figure 3. The input output relationship is
expressed as:
Y=
(1)
n* in + b

0.01

RESULT

The system was subjected to different load conditions

and the Permanent Magnet Synchronous Motor was
operated for a reference speed of 52.3rad/sec under the
following conditions:

-5
0

0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.1

(e)
60

50

Speed(rad/sec)

40

Case I: At no load with rated speed.

100
80

30

20

60
40

10

Voltage(V)

20
0
0

0
-20
-40
-60
-80
-100
0

0.01

0.02

0.03

(a)

0.04

0.05
Time (sec)

0.06

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04

0.1

(f)
Fig. 4: (a) Source Voltage, (b) Load Current, (c) Source
Current, (d) THD, (e) Motor Torque, (f) Speed response
Case II: On Load with rated speed.

www.ijsret.org

128

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882

100

100

80

80

60

60

40

40

20

20

Voltage(V)

Voltage(V)

Volume 4, Issue 2, February 2015

-20

-20

-40

-40

-60

-60
-80

-80
-100
0

0.01

0.02

0.03

0.04

0.05
Time (sec)

0.06

0.07

0.08

0.09

-100
0

0.1

0.01

0.02

0.03

50

50

40

40

30

30

20

20

10

10

Current(A)

Current(A)

0.05
Time (sec)

0.06

0.07

0.08

0.09

0.1

(a)

(a)

-10

-10

-20

-20

-30

-30
-40

-40
-50
0

0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

-50
0

0.1

0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.1

(b)

(b)
50

50
40

40

30

30

20

20

10

10

Current(A)

Current(A)

0.04

0
-10

0
-10

Selected signal: 5 cycles. FFT window (in red): 2 cycles

-20

-20

15
-30

10

-40

Selected signal: 5 cycles. FFT window (in red): 2 cycles

15

-30

10
-40

0
-50
0

0.01

-5
0

0.02

0.03

0.02

(c)

0.04

0.05
Time(seec)

0.06

0.07

0.04
0.06
Time (s)

0.08

0.09

0.08

-50
0

0.1

0.01

-5
0

0.04

(c)

0.05
Time(sec)

0.06

0.07

0.04
0.06
Time (s)

0.08

0.09

0.08

0.1

0.1

Fundamental (50Hz) = 7.865 , THD= 3.28%

1

0.8

Mag (% of Fundamental)

Mag (% of Fundamental)

0.03

0.02

Fundamental (50Hz) = 7.862 , THD= 3.04%

0.6

0.4
0.2

0.02

0.1

0.8
0.6
0.4
0.2
0

200

400
600
Frequency (Hz)

800

1000

200

400
600
Frequency (Hz)

800

1000

(d)

(d)

10

10

Torque(Nm)

Torque(Nm)

-2

-5
0

-4
0
0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.01

0.02

0.03

0.04

0.1

(f)
Fig. 5: (a) Source Voltage, (b) Load Current, (c) Source
Current, (d) THD, (e) Motor Torque, (f) Speed response
Case III: Change in load with rated speed.

www.ijsret.org

(e)

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.1

129

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 2, February 2015
60

10
8

50

6
40

Speed(rad/sec)

Torque(Nm)

30

20

2
0
-2

10

-4
0

-6
-10
0

0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

-8
0

0.1

(f)
Fig. 6: (a) Source Voltage, (b) Load Current, (c) Source
Current, (d) THD, (e) Motor Torque, (f) Speed response

0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.1

(e)
60

50

Speed(rad/sec)

40

Case IV: On Load with change in Speed.

100

30

20

80
60

10

40

Voltage(V)

20

0
0

0.01

0.02

0.03

0.04

0.1

(f)
Fig. 7: (a) Source Voltage, (b) Load Current, (c) Source
Current, (d) THD, (e) Motor Torque, (f) Speed response

-20
-40
-60
-80
-100
0

0.01

0.02

0.03

0.04

0.05
Time (sec)

0.06

0.07

0.08

0.09

0.1

V.

(a)
50
40
30
20

Current(A)

10
0
-10
-20
-30
-40
-50
0

0.01

0.02

0.03

0.04

0.05
Time(sec)

0.06

0.07

0.08

0.09

0.1

CONCLUSION

In this paper, a new Hybrid Active Power Filter topology

has been presented and analyzed. The filter topology
combines the compensation characteristics of the shunt
active power filter and passive filter. The proposed
topology has been shown to be capable of reducing the
torque ripple and current harmonics, and providing
almost sinusoidal voltage to motor terminals, which was
reflected in a smoother line current waveform.

(b)

REFERENCES

50
40
30
20

[1]

Current(A)

10
0
-10

Selected signal: 5 cycles. FFT window (in red): 2 cycles

-20

15

-30

10

-40

[2]

5
0

-50
0

-5

0.01

0.02

0.03

0.02

(c)

0.04

0.05
Time(sec)

0.06

0.04
0.06
Time (s)

0.07

0.08

0.08

0.09

0.1

0.1

[3]

Mag (% of Fundamental)

Fundamental (50Hz) = 7.82 , THD= 2.22%

0.4

[4]

0.3
0.2
0.1
0

[5]
0

200

400
600
Frequency (Hz)

800

1000

(d)

www.ijsret.org

Roger C.Dugan, Mark F. McGranaghan, Surya

Santoso and H.Wayne Beaty, Electrical power
system quality, Mcgraw-Hill.
H. Akagi, Trends in active power line
conditioners, IEEE Trans. Power Electronics, vol.
9, no. 3, pp. 263-268,May 1994
J. Yongtao Dai, Wenjin Dai, Harmonic and
Reactive Power Compensation with Artificial Neural
Network Technology , IEEE, 2008.
Laxmi Devi Sahu, S.P Dubey, ANN based Hybrid
Active Power Filter for Harmonics Elimination with
Distorted Mains,IJPEDS, Vol. 2, No. 3, pp.
241~248, September 2012.
Ali Ahmed Adam, Kayhan Gulez, Torque ripples
and EMI Noise Minimization in PMSM Using
Hybrid Filter, SICE annual conference, Sept.1720,Kagawa University,2007.

130

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 2, February 2015
[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

Mohammed Qasim, Vinod Khadkikar, Application

of Artificial Neural Networks for Shunt Active Power
Filter Control, IEEE TRANSACTIONS ON
INDUSTRIAL INFORMATICS, VOL. 10, NO. 3,
AUGUST 2014.
A. A. James and E. Rosenfeld, Bernard widrow, in
Talking Nets: An Oral History of Neural Networks.
Cambridge, MA, USA: MIT Press, ch. 3, pp. 45
69,1998.
Satya
Prakash
Dubey,Pukhraj
Singh
and
H.V.Manjunath, DSP based neural network
controlled hybrid active power filter,International
Journal
of
Emerging
Electric
Power
System,Vol.4,No.2,pp.1-16,2005.
N.Gupta,S.P.Dubey and S.P.Singh, Neural
Network-based shunt active filter with direct current
control
for
power
quality
conditioning,
International Journal of Power Electronics,
InderScience
Publishers
Vol.6,No.3,pp.597620,2011.
Kayhan Gulez, Ali Ahmed Adam, and Halit Pastaci,
Torque Ripple and EMI Noise Minimization in
PMSM Using Active Filter Topology and FieldOriented Control, IEEE TRANSACTIONS ON
INDUSTRIAL ELECTRONICS, VOL. 55, NO. 1,
JANUARY 2008.
K. Sangsun, and P. N. Enjeti, "A new hybrid active
power filter (APF) topology", IEEE Transactions on
Power Electronics, vol.17, no. 1, pp. 48-54,2002.
Kiran
Boby,
Prof.Acy
M
Kottalil,
N.P.Ananthamoorthy, Simulation of PMSM Vector
Control System with Fuzzy Self-Adjusting PID
Controller Using MATLAB, International Journal
of Scientific and Research Publications, Volume 3,
Issue 3,ISSN 2250-3153, March 2013.
H.Akagi, New Trends in active power line
conditioning, IEEE Trans. Indu.Applica., vol.32,
no.6, Nov./Dec. 1996.
Dr.T.Govindaraj,
Dhivya.N.M,
Simulation
Modelling on Artificial Neural Network Based
Voltage
Source
Inverter
Fed
PMSM,
INTERNATIONAL JOURNAL OF INNOVATIVE
RESEARCH IN ELECTRICAL, ELECTRONICS,
INSTRUMENTATION
AND
CONTROL
ENGINEERING Vol. 2, Issue 1, January 2014.

www.ijsret.org

131