Improvement of Power Quality using Advanced

Artificial Neural Network Algorithm

Soumya Ranjan Das Prakash. K. Ray Asit Mohanty

Department of EE Department of EEE Department of EE
IIIT Bhubaneswar, CET, Bhubaneswar, CET, Bhubaneswar,
Bhubaneswar, Bhubaneswar, Bhubaneswar,
[email protected] [email protected] [email protected]

Abstract—In the present scenario of power system network, overcoming the issues elevated in APF, different
active filter are being designed using advanced technology for configurations [4-6] of passive filter and APF is designed.
compensating harmonics and reactive power for better quality Here, SHAF is used to decrease the rating [7] of APF and
of power supply. The present paper implements a Shunt the considered configuration is extremely suitable because
Hybrid Active Filter (SHAF), necessary for reactive power of its reasonable cost and emphasis is given on the
compensation in the nonlinear load and suppresses harmonics performance of SHAF and its controller for handling the
for improving the power quality (PQ) in the distribution harmonics and reactive power in three phase power system.
network. In this work, an adaptive linear neuron (Adaline) In SHAF, the role of controller is very vital and thus in
based variable step size leaky least mean square (VSSLLMS) recent few years, many techniques have been designed for
control approach is applied for extracting harmonics. This controlling the VSI of SHAF. But, during operation it is
competent approach is needed for generating the gating pulses found that the control techniques based on time and
essential for IGBT inverter and increasing the performance frequency domain approaches found ineffective under
compared to the conventional Adaline-LMS controller. The unbalance conditions and may supply degraded results
proposed controller provides faster convergence and also during transient situations.
provides better performance compared to conventional one.
Simulation results show the significant improvement in In recent times, different controllers based on Artificial
Adaline-VSSLLMS approach compared to Adaline-LMS intelligence [8-10] are applied extensively in the control area
approach. The simulations of the proposed controller and systems because of its compliance and flexibility
system is done using MATLAB/ SIMULINK under different characteristics. In this paper, to calculate the reference
load conditions. current, an Adaline based algorithm [11] with variable step
size leaky least mean square (VSSLLMS) [12] is
Keywords—Adaline, LMS algorithm, shunt hybrid active filter implemented in SHAF. The proposed controlling scheme is
(SHAF),variable step size leaky least mean square (VSSLLMS), implemented to produce the gating signals of VSI based
total harmonic distortion (THD). SHAF. The proposed control technique involves weighting
values estimation, maintaining dc bus voltage and
I. INTRODUCTION calculation of reference current. Finally, with the use of
In modern days it is observed that power system pulse width modulation based hysteresis current controller
experience lots of challenges due to the presence of (HCC), the gating pulses are generated by relating the
harmonics produced by power electronics devices which are reference source with the detected source current. The
extensively used in different sectors of power system proposed controller when incorporated in SHAF provides
network [1]. With the use of power electronics devices, significant improvements in the quality of power.
harmonics and reactive powers [2] are generated in the
system network which adversely affects the supply system. II. SYSTEM CONFIGURATION
To deal with these existing issues, usually passive LC filters
are used for compensating harmonics due to their low cost. The configuration of the proposed model is illustrated in
But, certain characteristics. like problems in tuning, fixed Fig. 1 which comprises of series of passive filters and an
compensation, big size and effect of resonance make it APF with VSI design having large capacitor on dc link bus.
difficult to implement these filters in power system network. The ripple filters connected on the output side are useful for
Subsequently, in 1970s [3], voltage and current harmonics mitigating large-frequency ripples. The configuration of
are eliminated by the use of active power filter (APF). SHAF reduces the rating of switching devices. In this
Application of APF in power system network found to context, two different control schemes are performed in
achieve a feasible solution for reducing the distortions in SHAF for generation of reference currents; one is the
currents and voltages. But, because of its incapability in conventional Adaline-LMS controller and the other is
handling the higher voltage grid and also of higher cost, proposed Adaline-VSSLLMS controller. Three phase source
force the researchers to select an alternate solution. For

978-1-5386-9316-2/18/$31.00 ©2018 IEEE

 .

Fig.1. Circuit topology of the proposed system using Adaline based VSSLLMS Algorithm

voltages and currents are mixed with sinusoidal components III. CONTROL ALGORITHIMS
and harmonics.By utilizing Adaline based controllers, the
sinusoidal components are extracted from non-sinusoidal A. Adaline Based LMS Algorithm
supply. In the VSSLMS algorithm, step-size is significantly
adjusted for the period of conjunction. Three Adalines are In this controlling technique, the reference current are
implemented in each phase and are utilised for harmonics calculated by means of the estimated weighted values and
extraction and reactive components extraction from the variable parameters by controlling the bus voltage across the
sensed load currents. This algorithm shows fast and dynamic capacitor.
response compared to Adaline-LMS controller and other
The reference current for three phase system is given as
traditional controllers. The configuration of Adaline-
VSSLLMS controller is shown in Fig. 2. isa* = ( wn+ + isn ) vsa (1)
The development in design of SHAPF comprises designs
of passive and APF and different types of controllers.
Hence, it is very important to choose the parameters of both isb* = ( wn+ + isn ) vsb (2)
passive and active filters operating in an efficient manner.
isc* = ( wn+ + isn ) vsc
Capacitor across VSI in SHAF regulates the dc link voltage
(3)
and the value of capacitor may stimulus to the operation of
SHAF.
Where, wn+ is the weighted variable and is given as

wn(k+1) = wn(k) + μilk − wn(k)vs(k)  vs(k)

+
w =
(w +
na + wnb+ + wnc+ ) (4)
n
3

From (1), (2) and (3) vsa , vsb and vsc are considered as
unit voltage template and isn is the output current from PI
Fig. 2. Configuration of Adaline-VSSLLMSAlgorithm controller.
B. Adaline based VSSLLMS Algorithm During compensation, the power system model is operated
In this algorithm, the Adaline weights (as in eq. (4)), are using SHAF with incorporation of Adaline-LMS algorithm
usually reorganized by employing the least mean square and Adaline-VSSLLMS algorithm. From simulation results,
(LMS) algorithm. Weight are found out from the it is evident that the improvement in the currents due to non-
components of the compensating current which is given as linear load is found to be more satisfactory using Adaline-
LMS algorithm when compared to the system without
iFa (t ) = iLa (t ) − iXa (t ) = iLa (t ) − [ w1a , w2a ] va (t ) connecting any hybrid filter. The improvement in THD
(5) value is found to be 4.60%. Fig. 4(a) and 4(b) respectively
Where, w1a and w2a are Adaline weight vectors. show the simulated results of load current, compensation
current and the dc link voltage of the SHAF and the
Based on the proposed algorithm, the Adaline weight corresponding THD value is shown in Fig. 4(c).
vectors are specified as Time Series Plot:Load current
100
w1a (n + 1) = (1 − 2Δ(n)λ1a (n) ) w1a (n) + Δ(n)ea (n) x1a (n)

i La
0
(6)
w2 a (n + 1) = (1 − 2Δ(n)λ2a (n) ) w1a (n) + Δ(n)ea (n) x2 a (n) -100
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

200 Time Series Plot:Load current

Where, (n) is treated as variable step-size parameter
and for LMS algorithm, the parameter λ (k ) = 0 .

i Lb
0

-200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
IV. SIMULATED RESULTS AND ANALYSIS 200 Time Series Plot:Load current
The prime focus of the paper is to examine the

i Lc
0
performance of the proposed control technique using SHAF. -200
A simulation model for the proposed control scheme has 0 0.02 0.04 0.06 0.08 0.1
Time (seconds)
0.12 0.14 0.16 0.18 0.2

been performed with MATLAB/Simulink using parameters (a)

shown in Appendix in Table II. The proposed control Time Series Plot:compensation current
100
technique is accomplished under different working 50
conditions. The performance of SHAF is tested with
icomp

0
VSSLLMS control technique which can be easily -50
implemented. The proposed model is run at a sampling -100
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
period of 20s. Time Series Plot:DC link voltage
2000
Time Series Plot: LOAD CURRENT PHASE A
1 1000
Vdc
i La

0 0

-1 -1000
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time Series Plot:LOAD CURRENT PHASE B Time (seconds)
1
(b)
i Lb

-1
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time Series Plot:LOAD CURRENT PHASE C
1
i Lc

-1
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time (seconds)
(a)

(c)
Fig. 4 (a) Performance of SHAF using ADALINE-LMS Algorithm in three
phase load currentunder balanced load (b) compensation current and dc link
voltage (c) THD value of load current
Further, the analysis is carried out using the proposed
advanced technique, i.e, Adaline-VSSLLMS algorithm.
Using the proposed control scheme, improvement in THD
values and waveforms are noticed as compared to that in
Adaline-LMS algorithm. At t=0.02, the SHAF is able to
(b) compensate the harmonics in the currents as it injects filter
currents for compensating the imbalance in load current.
Fig.3. (a) Distortions in three phase load currents without using SHAFunder
balanced load (b) THD analysis
Fig. 5(a) and 5(b) respectively show the simulated results of
load current, compensation current and dc link voltage of the
Before compensation, without the use of SHAF, the SHAF under balanced load. The corresponding THD value
power system is made to perform with non-linear loads. is shown in Fig. 5(c).
From the simulation, it is analyzed that without connecting The current compensation became sinusoidal even with
any hybrid filter, distortions in the currents are found during variation in magnitude of the load current. Results obtained
different operating conditions. The simulated results are from simulations indicate that the proposed method
shown in Fig. 3(a) and the THD value is shown in Fig. 3(b). effectually eliminates unbalanced currents. Moreover, using
the proposed control scheme, the dc link voltage maintains
consistency during the operation of SHAF.
Time Series Plot:Load current
2
i La

-2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time Series Plot:Load current
2
i Lb

-2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time Series Plot:Load current
2
(b)
i Lc

0
Fig.6. (a) Distortions in three phase load currents without using SHAF
-2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 underunbalanced load (b) THD analysis
Time (seconds)

(a) It is evident from simulated results that the load current,

ILa, ILb and ILc are composed of harmonic components and
Time Series Plot:compensation current
1 the load current waveform is fully unbalanced. The
corresponding THD value is obtained to be 39.53% and
i comp

0 shown in Fig. 6(b). The proposed system is then performed

by connecting SEHAF using Adaline -LMS algorithm under
-1
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 unbalanced conditions and simulated results are illustrated
Time Series Plot:DC link voltage in Fig. 7(a) and 7(b) and the corresponding THD value in
150
Fig. 7(c) is found to be 4.95 %. Subsequently, the proposed
100
system is operated under load unbalanced conditions using
Vdc

50

0
Adaline-VSSLLMS algorithm. The simulated results show
-50 significant improvement in THD value of 2.85% as
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time (seconds) compared to that in the case of Adaline-LMS Algorithm.
The simulated results are presented in Fig. 8(a) and 8(b)
(b)
with its THD value is shown in Fig. 8(c). A comparative
result is shown in Table I and Fig. 9.
Time Series Plot:Load current
200
i La

-200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time Series Plot:Load current
500
i Lb

-500
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time Series Plot:Load current
500
i Lc

(c) -500
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

Fig. 5 (a) Performance of SHAF using ADALINE-VSSLLMS Algorithm in Time (seconds)

three phase load current under balanced load (b) compensation current and (a)
dc link voltage (c) THD value of load current
Time Series Plot:compensation current
Time Series Plot:Load current 200
200
0
i comp
i La

-200 -200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

Time Series Plot:Load current -400

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
500
Time Series Plot:DC link voltage
i Lb

0 400
-500 200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
V dc

Time Series Plot:Load current 0

200
-200
i Lc

0
-400
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Time (seconds)
Time (seconds)
(b)
(a)
 TABLE I
THD Analysis
Controllers % THD Analysis

Balanced loads Unbalanced loads

Without controllers 28.38 39.53

With Adaline- LMS 4.60 4.95

controller

With Adaline- 2.39 2.85

VSSLLMS controller
(c)
Fig. 7 (a) Performance of SHAF using Adaline-LMS Algorithm in three
phase load current under unbalanced load (b) compensation current and dc 50
link voltage (c) THD value of Load current Without
40
30 controller
Time Series Plot:Load current
20
20 Adaline-LMS
iLa

0
10
-20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
0 Adaline
Time Series Plot:Load current
20 Balanced load Unbalnced VSSLLMS
load
iLb

-20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Fig.9. Comparative results showing THDs analysis under different load
Time Series Plot:Load current conditions.
20

V. CONCLUSION
iLc

-20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
The improvement of PQ in three phase system using a
Time (seconds) hybrid power filter consisting of a shunt active power filter
and passive filters is analysed in the present work. This
(a)
paper presents a comparative analysis using Adaline linked
Time Series Plot:compensation current LMS algorithm and Adaline based VSSLLMS algorithm
500
with normal supply system connecting with balanced and
unbalanced load. The proposed system reduces the power
icom p

0 rating of the active part. At the same time, it can improve the
compensation performance and damp out the parallel
resonance as compared to that with only the passive filters.
-500
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 The proposed algorithm makes the system stable and
Time Series Plot:DC link voltage harmonics free by maintaining the dc link bus voltage across
60
the VSI. Simulated results demonstrate the robust
40
performance of the proposed controller in power system.
V dc

20 VI. APPENDIX
TABLE II
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time (seconds) Parameters Value

(b) Line voltage and Frequency 415V,50Hz

Ls= 0.15mH, Lac= 1.5mH, CDC = 1500ߤF,
RL=60,L=10ߤF; (R1 =2 , R2 =4 , R3
Line and Load impedance =6 )
Cf5= 50 ߤF Lf5 =8.10mH; Cf7= 20 ߤF Lf7
Tuned Passive Filter = 8.27mH; Cf11= 20 ߤF Lf11=8.270mH
Ripple Filter CRF = 50 ߤF , LRF =0.68mH
Active Filter parameters CD = 1500ߤF VD =750V
Filter coupling inductance 2.5mH
KP=0.032,KI=0.00004,forSeries PF;
(c) Controller gain KP=24, KI=1.2, for Shunt PF
Fig. 8 (a) Performance of SHAF using Adaline-LMS Algorithm in three Hysteresis Band Limit 0.5A
phase load current under unbalanced load (b) compensation current and dc
link voltage (c) THD value of Load current Sampling time 20s
 REFERENCES Transactions on Industrial Electronics., vol. 63, no. 8, pp. 4849-4859,
2016
[1] O. P. Mahela,A. G. Shaik, “ Topological aspects of power quality
improvement techniques: A comprehensive overview,” Renewable [7] N. He, D. Xu, and L. Huang, “The application of particle swarm
and Sustainable Energy Reviews., vol. 58, pp. 1129-42, May 2016 optimization to passive and hybrid active power filter design,” IEEE
transactions on industrial electronics., vol. 56, no. 8, 2841-2851, 2009
[2] J. S. Subjak, J. S. Mcquilkin, “Harmonics-causes, effects,
measurements, and analysis: an update,” IEEE transactions on [8] M. Qasim and V. Khadkikar, “Application of artificial neural
industry applications., vol. 26, no. 6, pp. 1034-42, 1990 networks for shunt active power filter control,” IEEE Transactions on
industrial informatics., vol. 10, no. 3,pp. 1765-1774, 2014
[3] Z. Shuai, A.Luo,W. Zhu, R. Fan, K. Zhou, “Study on a Novel Hybrid
Active Power Filter Applied to a High-Voltage Grid,” IEEE [9] K. W. Kow, Y. W. Wong and R. K.Rajkumar, "A review on
Transactions on Power Delivery.,vol. 24, no. 4, pp. 2344-2352, 2009 performance of artificial intelligence and conventional method in
mitigating PV grid-tied related power quality events,” Renewable and
[4] Ferreira, S. Costa, R. B. Gonzatti, R. R. Pereira, C. H. da Silva, L. E. Sustainable Energy Reviews., vol. 56, pp. 334-346.2016
B. da Silva, and G. Lambert-Torres, “Finite control set model
[10] V. Mahajan, P. Agarwal, and H. Om Gupta, “An artificial intelligence
predictive control for dynamic reactive power compensation with
based controller for multilevel harmonic filter,” International Journal
hybrid active power filters,” IEEE Transactions on Industrial
of Electrical Power & Energy Systems.,vol. 58, pp. 170-180, 2014
Electronics., vol. 65, no. 3, pp. 2608-2617, 2018
[5] J. Ye, A. Shen, Z. Zhang, J. Xu, and F. Wu, “Systematic design of the [11] M. Mangaraj, A. K. Panda, T. Penthia, “Investigating the performance
hybrid damping method for three-phase inverters with high-order of DSTATCOM using ADALINE based LMS algorithm,” IEEE 6th
filters,” IEEE Transactions on Power Electronics., vol. 33, no. 6, pp. International Conference InPower Systems (ICPS)., pp. 1-5, Mar.
4944-4956, 2018 2016
[6] A. Javadi, A. Hamadi, L. Woodward, and K. Al-Haddad, [12] P. Garanayak, G. Panda, “Harmonic Elimination and Reactive Power
“Experimental investigation on a hybrid series active power Compensation with a Novel Control Algorithm based Active Power
compensator to improve power quality of typical households,” IEEE filter,” Journal of Power Electronics., vol. 15, no. 6, pp. 1619-27,
2015