Microsoft Word - 13-2833_sETASR_V9_N4_pp4384-4388

Engineering, Technology & Applied Science Research Vol. 9, No. 4, 2019, 4384-4388 4384

www.etasr.com Truong & Bui: Hybrid PSO-Optimized ANFIS-Based Model to Improve Dynamic Voltage Stability

Hybrid PSO-Optimized ANFIS-Based Model to

Improve Dynamic Voltage Stability

Dinh-Nhon Truong

Faculty of Electrical and Electronics Engineering,
Ho Chi Minh City University of Technology and Education,

Ho Chi Minh City, Vietnam

nhontd@hcmute.edu.vn

Van-Tri Bui

Faculty of Electrical and Electronics Engineering,
Ho Chi Minh City Vocational College of Technology,

Ho Chi Minh City, Vietnam

vantrihcm@gmail.com

Abstract—The objective of this paper is to perform a hybrid

design for an Adaptive Neuro-Fuzzy Inference System (ANFIS)

optimized by Particle Swarm Optimization (PSO) to improve the

dynamic voltage stability of a grid-connected wind power system.

An onshore 99.2MW wind farm using Doubly Fed Induction

Generator (DFIG) is studied. To compensate the reactive power

absorbed from the power grid of the wind farm, a Static VAR

Compensator (SVC) is proposed. To demonstrate the

performance of the proposed hybrid PSO–ANFIS controller,
simulations of the voltage response in time-domain are

performed in Matlab to evaluate the effectiveness of the designed

controller. From the results, it can be concluded that the

proposed hybrid PSO-optimized ANFIS-based model can be

applied to enhance the dynamic voltage stability of the studied
grid-connected wind power system.

Keywords-adaptive neuro-fuzzy inference system; particle

swarm optimization; static var compensator; voltage stability

I. INTRODUCTION

Renewable energy is important nowadays and wind
generators or solar panels are used in many countries, while
wind generators are applied more widely than before. An
offshore wind farm can include many wind turbines and wind
driven generators. The output power of the wind generators can
be connected to form a wind farm and they can feed a power
grid using a step-up transformer. In relevance to a wind
generator, DFIG (doubly fed induction generator) is the most
prevalent generator because of its high efficiency compared to
other types of wind generators such as permanent magnet
synchronous generator (PMSG) and induction generator (IG)
[1]. Moreover, DFIG has the capability of controlling both
active and reactive power for better grid integration to the
transmission line [2]. However, connecting the stator windings
directly to the power grid, makes it very sensitive to power grid
faults. Besides, the randomness of wind energy affects the
power quality of the connected power systems. Some power
quality problems such as voltage fluctuations, flicker, harmonic
and voltage deviation of the Baolian wind farm have been
studied in [3]. The measurements of a wind farm located in the
south-east part of Poland were analyzed and compared with the
acceptable values in [4]. In [5], a power quality study of a
large-scale wind farm with and without storage was studied.
Authors showed that with the proposed storage system, voltage

drops, and harmonic voltage and current distortion at the wind
farm output can be improved when a short circuit occurs in the
connection with the power grid. Flexible AC Transmission
Systems (FACTS) devices have been proposed and applied in
power transmission systems and in this situation, Static Var
Compensator (SVC) is used to enhance the voltage stability [6,
7]. In large systems, the SVC is located in the middle of the
transmission line to reduce the oscillation of the system and
thus improve grid stability [8, 9].

In today's control devices, the controller plays an important
role in improving the operability and flexibility of the device.
PID (proportional-integral-derivative) is a linear controller and
is used in many control systems. The main disadvantage of this
controller is that it does not respond well when the input is
large. Recently, many algorithms have been applied to improve
the efficiency of the device, such as a fuzzy logic controller
used in vibration reduction control when linking generator
systems in two large areas [10]. In addition, the fuzzy
controller is used to replace the PID controller using the
voltage deviation as the feedback signal in the voltage control
at the link bus [11]. However, the main disadvantage of fuzzy
logic controllers is that they depend on the programmer's
experience. To improve the accuracy of the algorithms,
combining algorithms together to exploit the advantages of
each algorithm is a good solution that is being widely
researched. In [12], authors proposed an algorithm combining
fuzzy logic and an artificial neural network called ANFIS
(Adaptive-Network-based Fuzzy Inference System). ANFIS is
also a good structure for designing a controller of FACTS
devices [13] because of its simplification and its ease of
designing. ANFIS is used for replacing controllers that are not
obvious or cannot be calculated by definite equivalence.

In this research paper, a Particle Swarm Optimization
(PSO) algorithm is proposed to optimize the parameters of the
ANFIS controller for improving voltage stability. The proposed
approach is based on the combination of PSO and ANFIS
controller. The idea of combining PSO and ANFIS was applied
to examine the electricity market of mainland Spain [14] as
well as for short-term electricity prices prediction [15] with
many advantages.

Corresponding author: Dinh-Nhon Truong

Engineering, Technology & Applied Science Research Vol. 9, No. 4, 2019, 4384-4388 4385

www.etasr.com Truong & Bui: Hybrid PSO-Optimized ANFIS-Based Model to Improve Dynamic Voltage Stability

II. SYSTEM CONFIGURATION

A. Studied System Configuration

The structure of the studied system in this paper is a
practical 99.2MW wind farm in Bac Lieu province, Vietnam
including 62×1.6MW wind turbineS as shown in Figure 1. This
project, panning an area of 1,000ha, transmits power to the
national grid through two parallel 63MVA transformers and
two 110kV transmission lines. The wind farm has an estimated
electricity output of 300GWh per annum. The wind turbine
model is GE 1.6 XLE – 1.6MW supplied by General Electric
using a DFIG-based wind turbine with a single-line diagram
[16]. The stator windings of the DFIG are directly connected to
the low-voltage side of the 0.69/22kV step-up transformer
while the rotor windings of the DFIG are connected to the same
0.69kV side through a rotor-side converter (RSC), a DC link, a
grid-side converter (GSC), and a connection line. For normal
operation of a DFIG, the input AC-side voltages of the RSC
and the GSC can be effectively controlled to achieve
simultaneous output active power and reactive power control
[17].

Fig. 1. Single-line diagram of the studied system.

For normal operation of a wind DFIG, the input AC-side
voltages of the RSC and the GSC can be effectively controlled
to achieve the aims of simultaneous output active power and
reactive power control. The control block diagram of the RSC
of the studied DFIG and the control block diagram of the GSC
of the studied wind DFIG can be seen in [18].

B. SVC Model

For improving voltage stability, a 40 MVAr SVC is
proposed for adjusting the voltage at the connected bus by
compensating the reactive power to the power grid. The
equivalent model of SVC is shown in Figure 2. It consists of a
thyristor switched capacitor (TSC) and thyristor controlled
reactors (TCRS). The control scheme of the studied SVC is
presented in [19]. In the SVC model, if the bus voltage is lower
than the reference value, the value of the equivalent
susceptance (BSVC) of the SVC is positive and on the contrary,
if the bus voltage is higher than the reference value, the BSVC is

negative. The relationship between the firing angle ( )α and the
steady state value of BTCR is given as:

2( ) sin(2 )
( )

π α α
α

π
− +

=L
L

B
X

(1)

where [ ]
2

π
α π= ÷ .

The equivalent susceptance of SVC (BSVC) is:

1
( ) ( )α α= −SVC L

B B
X

(2)

Fig. 2. Single-line diagram of SVC

III. HYBRID PSO–ANFIS CONTROLLER DESIGN

An ANFIS is a class of adaptive multi-layer feedforward
network. It incorporates the self-learning ability of neural
networks with the linguistic expression function of fuzzy
inference. The structure of the ANFIS controller is depicted in
Figure 3. The ANFIS network is composed of five layers.

Fig. 3. Structure of the ANFIS

Each layer contains several nodes which described as the
following node equations:

In Layer 1, the output function
1
iO shows the membership

grade of a fuzzy set A1, A2 , B1, or B2 and it specifies the
degree to the given input e or Δe.

1
( ), 1 2

i i
O A e iµ= = ÷ (3)

Engineering, Technology & Applied Science Research Vol. 9, No. 4, 2019, 4384-4388 4386

www.etasr.com Truong & Bui: Hybrid PSO-Optimized ANFIS-Based Model to Improve Dynamic Voltage Stability

2
( ), 3 4

i i
O B e iµ −= ∆ = ÷ (4)

The membership functions for A and B are usually
described by a generalized bell functions as described bellow:

1
( )

i
i q

A e
e r

µ =
−

(5)

where , ,i i ip q r are the values of the bell function.

In Layer 2, the product output value
2
iO is formed by

multipling the incoming signals. In this layer, each node output
represents the firing strength of a rule.

2
( ). ( ), 1 2

i i i i
O w A e B e iµ= = ∆ = ÷ (6)

In Layer 3, the outputs of this layer,
3
iO , are called

normalized firing strengths:

1 2

, 1 2i
i i

w
O w i

w w
= = = ÷

+
(7)

In Layer 4, the output
4
iO is the contribution of the i

th
rule

to the overall output:

4
( . . ), 1 2

i i i i i i i
O w z w a e b e c i= = + ∆ + = ÷ (8)

where , ,i i ia b c are consequent parameters.

In Layer 5, the output
5
iO is the final output as the

summation of all incoming signals

5
, 1 2

i i i
O F w z i= = = ÷∑ (9)
In an ANFIS system, neural networks extract automatically

fuzzy rules from numerical data and, through the learning
process, the membership functions are adaptively adjusted. For
improving the training process, PSO is applied to optimize the
parameters of the membership functions of the ANFIS
controller [20]. The objective of this method is to discover the
particle location that outcomes the finest assessment of a
specified fitness function to minimize the training errors of
ANFIS. In PSO, a swarm specifies the number of probable
solutions to a complex problem, where each probable solution
is known as a particle. Every particle set has its opening
parameters in random fashion and is flown throughout the
multi-dimensional search space during the initialization phase
of swarm optimization [21]. As sketched in Figure 4, an
updating mechanism of the PSO technique is presented, where
x(t) and v(t) denote a particle’s position and flight velocity over
a solution space, respectively. The following equations present
the search mechanism:

Velocity update rule:

( ) ( 1) ( )

( )

i i Pbesti i

Gbesti i

v t v t x x t

x x t

ω ρ

= − + −  

+ −  
(10)

where ω is an inertia weight, ρ1, ρ2 are random variables, Gbest
is the best particle among all particles in the swarm, and Pbest is
the personal best position of each particle.

Fig. 4. Updating new position mechanism of PSO

Position update rule:

( ) ( 1) ( ), 1
i i i

x t x t v t t t= − + = + (11)

The results after training of the proposed hybrid PSO-
ANFIS are shown in Figure 5. It can be seen that the training
error between the target and the output of the training data is
very close to zero (Figure 5(a)) and the error mean and error
StD are also very small (Figure 5(b)). It means the proposed
PSO can optimize the parameters of the ANFIS controller.

(a) Training error

(b) Error mean

Fig. 5. Training results of the proposed hybrid PSO-ANFI

IV. SIMULATION RESULTS

The simulation results of the studied Bac Lieu power
system with a wind farm and a proposed SVC with hybrid
PSO-ANFIS controller are presented in Figure 6 in which a
severe three-phase short circuit fault happened at the 220kV
bus in 0.1s. In Figure 6, the black lines are the responses of the
system with SVC and PI controller, the blue lines are the
responses of the system with SVC and ANFIS controller which
is trained by applying the ANFIS Toolbox in Matlab while the
red lines are the ones with SVC and the proposed hybrid PSO-
ANFIS controller. Active power and voltage of the wind farm
are represented in Figures 6(a) and 6(b).

0 50 100 150 200 250 300
-0.02

0.02

0.04
MSE = 5.5726e-05, RMSE = 0.007465

Error

-0.03 -0.02 -0.01 0 0.01 0.02 0.03
0

80
Error Mean = -3.7379e-06, Error St.D. = 0.0074783

Engineering, Technology & Applied Science Research Vol. 9, No. 4, 2019, 4384-4388 4387

www.etasr.com Truong & Bui: Hybrid PSO-Optimized ANFIS-Based Model to Improve Dynamic Voltage Stability

(a) Active power of the wind farm

(b) Voltage of the wind farm bus

(d) Voltage of Dong Hai bus

(e) Voltage of Bac Lieu bus

Fig. 6. The response of the system when a three-phase short circuit fault

happens at 220kV level. Legend:

(a) Voltage of Bac Lieu bus

(b) Voltage of the wind farm bus

Fig. 7. The response of the system when a three-phase short circuit fault

happens at 110kV level. Legend:

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-4

-2

t (s)

P
W
F
(
p
.u
.)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.2

0.4

0.6

0.8

1.2

t (s)

V
W
F
(
p
.u
.)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.2

0.4

0.6

0.8

1.2

t (s)

V
W
F
(
p
.u
.)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0

0.2

0.4

0.6

0.8

1.2

t (s)

V
D
o
n
g
H
a
(p
.u
.)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0

0.2

0.4

0.6

0.8

1.2

t (s)

V
B
a
c
L
ie
u
(
p
.u
.)

PI ANFIS PSO-ANFIS

0 0.25 0.5 0.75 1 1.25 1.5
0

0.2

0.4

0.6

0.8

1.2

t (s)

V
B
a
c
L
ie
u
(
p
.u
.)

0 0.25 0.5 0.75 1 1.25 1.5
0

0.2

0.4

0.6

0.8

1.2

t (s)

V
W
F
(
p
.u
.)

0 0.25 0.5 0.75 1 1.25 1.5
0.2

0.4

0.6

0.8

1.2

t (s)

V
D
o
n
g
H
a
(p
.u
.)

PI ANFIS PSO-ANFIS

Engineering, Technology & Applied Science Research Vol. 9, No. 4, 2019, 4384-4388 4388

www.etasr.com Truong & Bui: Hybrid PSO-Optimized ANFIS-Based Model to Improve Dynamic Voltage Stability

The voltage of the three buses i.e. Tra Noc, Dong Ha and
Bac Lieu are shown in FigureS 6(c)-6(e) respectively. From
these figures, it is easily to see that by applying the hybrid
PSO-ANFIS controller for the SVC device, the output values
of these parameters are more stable and more effective. The
voltage of each node is improved and the number of
oscillations and the overshoot after a three-phase short circuit
fault occurred, are also reduced. For more details, Figure 7
exhibits the simulation results of the studied system when a
three-phase short circuit fault happened at the Bac Lieu bus
working at 110kV level and lasting 5 cycles. By obseving the
response of the voltage at Bac Lieu bus shown in Figure 7(a) it
can be seen that voltage drop to zero occurred during the fault.
However, with the response plotted in Figures 7(b) and 7(c) the
voltage magnitude of the wind farm bus and the voltage at
Dong Ha bus only dropped to 0.2p.u. Thanks to the operation
of the SVC and its designed controller, a large amount of
reactive power was supplied in order to improve the voltage
level of these buses.

V. CONCLUSIONS

In this paper, a hybrid PSO-ANFIS controller for SVC was
designed and applied in a grid-connected wind power system.
SVC can support fast response to the system to balance reactive
power in the grid which helps to improve dynamic voltage
stability. The results showed that the proposed controller can be
used to improve the voltage quality and reduce the number and
amplitude of oscillations in hard operating conditions such as a
three-phase short circuit fault occurrence. It can be concluded
from the time-domain simulation results on Matlab that the
hybrid PSO-ANFIS designed controller has more advantages
compared to the ANFIS controller and can enhance power
quality of the studied system under severe operating conditions.

REFERENCES

[1] R. Pena, J. C. Clare, G. M. Asher, “Doubly fed induction generator using
back-to-back PWM converters and its application to variable-speed

wind-energy generation”, Electric Power Applications, Vol. 143, No. 3,
pp. 231-241, 1996

[2] L. Wang, L. Y. Chen, “Reduction of power fluctuations of a large-scale
grid-connected offshore wind farm using a variable frequency

transformer”, Transactions on Sustainable Energy, Vol. 2, No. 3, pp.
226-234, 2011

[3] Q. Xia, Z. Wang, F. Liu, Y. Li, Y. Peng, Z. Xu, “Study on Power
Quality Issues of Wind Farm”, 36th Chinese Control Conference,
Dalian, China, July 26-28, 2017

[4] M. Latka, M. Nowak, “Analysis of Electrical Power Quality Parameters
in the Power Grid with Attached Wind Farm”, Progress in Applied
Electrical Engineering, Koscielisko, Poland, June 25-30, 2017

[5] G. A. Ramos, M. A. Riosm, D. F. Gomez, H. Palacios, L. A. Posada,
“Power Quality Study of a Large-Scale Wind Farm with Battery Energy
Storage System”, Industry Applications Society Annual Meeting,

Cincinnati, USA, October 1-5, 2017

[6] A. Jain, P. P. Singh, S. N. Singh, “Control Strategies for Output Power
Smoothening of DFIG with SVC in Wind Conversion System”, Region

10 Humanitarian Technology Conference, Agra, India, December 21-23,
2016

[7] E. A. Awad, E. A. Badran, F. H. Youssef, “Mitigation of switching
overvoltages in microgrids based on SVC and supercapacitor”, IET
Generation, Transmission & Distribution, Vol. 12, No. 2, pp. 355–362,

2018

[8] Y. Chang, Z. Xu, G. Chen, J. Xie, “A Novel SVC Supplementary
Controller Based on Wide Area Signals”, Power Engineering Society

General Meeting, Montreal, Canada, June 18–22, 2006

[9] A. Jalilvand, M. D. Keshavarzi, “Adaptive SVC Damping Controller
Design, Using Residue Method in a Multi-Machine System”, 6th

International Conference on Electrical Engineering/Electronics,
Computer, Telecommunications and Information Technology, Pattaya,

Thailand, May 6–9, 2009

[10] L. O. Mak, Y. X. Ni, C. M. Shen, “STATCOM with fuzzy controllers
for interconnected power systems”, Electric Power Systems Research,

Vol. 55, No. 2, pp. 87–95, 2000

[11] I. Mansour, D. O. Abdeslam, P. Wira, J. Merckle, “Fuzzy Logic Control
of an SVC to Improve the Transient Stability of AC Power Systems”,
35th Annual Conference of IEEE Industrial Electronics, Porto, Portugal,

November 3–5, 2009

[12] J. S. R. Jang, “ANFIS: Adaptive-network-based fuzzy inference
system”, Transactions on Systems, Man, and Cybernetics, Vol. 23, No.

3, pp. 665–685, 1993

[13] A. Albakkar, O. P. Malik, “Adaptive Neuro-Fuzzy FACTS Controller
for Transient Stability Enhancement”, 16th National Power System

Conference, Hyderabad, India, December 15-17, 2010

[14] C. G. Martos, J. Rodriguez, M. J. Sanchez, “Mixed models for short-run
forecasting of electricity prices: Application for the Spanish market”,

Transactions on Power Systems, Vol. 22, No. 2, pp. 544–552, 2007

[15] H. M. I. Pousinho, V. M. F. Mendes, J. P. S. Catalao, “A hybrid PSO-
ANFIS approach for short-term wind power prediction in Portugal”,

Energy Conversion and Management, Vol. 52, No. 1, pp. 397-402, 2011

[16] The WindPower, 1.6xle, available at: https://www.thewindpower.net/
turbine_en_670_ge-energy_1.6xle.php, 2019

[17] J. P. S. Catalao, H. M. I. Pousinho, V. M. F. Mendes, “Short-term
electricity prices forecasting in a competitive market by a hybrid

intelligent approach”, Energy Conversion and Management, Vol. 52,
No. 2, pp. 1061–1065, 2011

[18] D. N. Truong, “STATCOM Based Fuzzy Logic Damping Controller For
Improving Dynamic Stability Of A Grid Connected Wind Power
System”, International Conference On System Science And Engineering,

Puli, Taiwan, July 7-9, 2016

[19] V. T. Bui, D. N. Truong, “Voltage Stability Enhancement of Bac Lieu
Wind Power by ANFIS Controlled Static Var Compensator”, 4th

International Conference on Green Technology and Sustainable
Development, Ho Chi Minh City, Vietnam, November 23-24, 2018

[20] M. A. Shoorehdeli, M. Teshnehlab, A. K. Sedigh, M. A. Khanesar,
“Identification using ANFIS with intelligent hybrid stable learning
algorithm approaches and stability analysis of training methods”,

Applied Soft Computing, Vol. 9, No. 2, pp. 833–850, 2009

[21] S. P. Singh, S. C. Sharma, “A novel energy efficient clustering algorithm
for wireless sensor networks”, Engineering, Technology & Applied

Science Research, Vol. 7, No. 4, pp. 1775-1780, 2017

AUTHORS PROFILE

Dinh-Nhon Truong, PhD is a senior lecturer at Ho Chi Minh City University

of Technology and Education, Vietnam. He received his PhD from the
Department of Electrical Engineering, National Cheng Kung University,

Tainan, Taiwan. His research interests are grid-connected wind power
systems, dynamic stability, and FACTS devices.

Van-Tri Bui, ME, is currently pursuing his PhD at the Ho Chi Minh City
University of Technology, Vietnam. His main research interest is the

application of FACTS devices to improve dynamic stability of a grid-
connected wind power systems.