ORI GI NAL ARTI CLE

Optimal tuning of PI controller using PSO optimization

for indirect power control for DFIG based wind turbine
with MPPT
Youcef Bekakra

Djilani Ben Attous
Received: 24 November 2012 / Revised: 1 March 2013
The Society for Reliability Engineering, Quality and Operations Management (SREQOM), India and The Division of Operation and
Maintenance, Lulea University of Technology, Sweden 2013
Abstract In this paper, an articial intelligence method
particle swarm optimization (PSO) algorithm is presented
for determining the optimal PI controller parameters for the
indirect control active and reactive power of doubly fed
induction generator (DFIG) to ensure a maximum power
point tracking of a wind energy conversion system. A
digital simulation is used in conjunction with the PSO
algorithm to determine the optimum parameters of the PI
controller. Integral time absolute error, integral absolute
error and integral square error performance indices are
considered to satisfy the required criteria in output active
and reactive power of a DFIG. From the simulation results
it is observed that the PI controller designed with PSO
yields better results when compared to the traditional
method in terms of performance index.
Keywords Doubly fed induction generator
Wind turbine Field oriented control PI (proportional
integral) Particle swarm optimization (PSO)
Maximum power point tracking (MPPT)
1 Introduction
Wind energy has attracted great attention due to several
advantages such as low pollution (Ackermann 2005).
Among the different alternatives to obtain variable speed
wind turbines, the system based on a doubly-fed induction
generator (DFIG) has become the most popular (Rahimi
and Parniani 2010). The electricity production by wind
power is the most predominant source of renewable energy
in Europe (Hammons 2008). By 2020 it is expected that
wind power generation will supply around 12 % of the
total electricity (Lobos et al. 2009).
The worldwide concern about the environmental pollu-
tion and the possible energy shortage has led to increasing
interest in technologies for generation of renewable electri-
cal energy. Among various renewable energy sources, wind
generation has been the leading source in the power industry
(Qiao et al. 2006). Recently, the DFIG is becoming the main
conguration of wind power generation because of its unique
advantages. Vector control technology is used to control the
generator, and the rotor of DFIG is connected to an AC
excitation of which the frequency, phase, and magnitude can
be adjusted. Therefore, constant operating frequency can be
achieved at variable wind speeds (Guo-qing et al. 2010).
In the last decade, various modern control techniques
such as adaptive control, variable structure control and
intelligent control have been intensively studied for con-
trolling the nonlinear components in power systems.
However, these control techniques have few real applica-
tions probably due to their complicated structures or the
lack of condence in their stability. Therefore, the con-
ventional PI controllers, because of their simple structures,
are still the most commonly used control techniques in
power systems, as can be seen in the control of the wind
turbines equipped with DFIGs. Unfortunately, tuning the PI
controllers is tedious and it might be difcult to tune the PI
gains properly due to the nonlinearity and the high com-
plexity of the system (Qiao et al. 2006).
Y. Bekakra D. B. Attous (&)
Department of Electrical Engineering, Faculty of Sciences and
Technology, University of El Oued, P.O. Box 789,
El Oued, Algeria
e-mail: dbenattous@yahoo.com
Y. Bekakra
e-mail: youcef1984@gmail.com
1 3
Int J Syst Assur Eng Manag
DOI 10.1007/s13198-013-0150-0
Modern heuristic algorithms are considered as effective
tools for nonlinear optimization problems. The algorithms
do not require that the objective function has to be differ-
entiable and continuous. A particle swarm optimization
(PSO) is one of the modern heuristic algorithms and can be
applied to nonlinear and noncontinuous optimization
problems with continuous variables. It has been developed
through simulation of simplied social models (Naka et al.
2003).
Particle swarm optimization is a population based
optimization algorithm which is rst introduced by
Kennedy and Eberhart (1995). It can be obtained high quality
solutions within shorter calculation time and stable con-
vergence characteristics by PSO than other stochastic
methods such as genetic algorithm (GA) (Gozde and
Cengiz Taplamacioglu 2011). The method is based on the
simulation of animal social behaviors such as sh school-
ing, bird ocking, and swarm theory. Since it is population
based and self-adaptive, it has gained an increasing popu-
larity as an efcient alternative to the GAs in solving
optimization problems. Moreover, it is shown to be effec-
tive in optimizing difcult multidimensional discontinuous
problems in a variety of elds. Similar to other population-
based optimization method such as the GA, the PSO
algorithm starts with random initialization of a population
of individuals in the search space. Each particle in the
search space is adjusted by its own ying experience and
the other particles ying experience to nd the global best
solution at each generation. Compared with the GA, PSO
has memorial ability to let the knowledge of good solutions
be retained by all particles, whereas the previous knowl-
edge need not be considered after each evolution in GA.
Furthermore, low computation cost, simplicity of imple-
mentation, and quick convergence ability make PSO pop-
ular in many applications. In addition, to further enhance
the particles learning ability and make it powerful in rea-
soning, some research has been developed to improve the
PSO recently an inertia weight is adopted in PSO to bal-
ance the local and global search ability, while the worst
experience component is included in PSO to give addi-
tional exploration capability (Lin et al. 2009).
Through studying the characteristics of wind turbine, the
paper proposed the maximum power point tracking
(MPPT) control method. Firstly, according to the DFIG
character, the paper adopts the vector transformation con-
trol method of stator oriented magnetic eld to realize the
decoupling control of the active and reactive power using
PI controllers which is tuned by PSO.
In this paper, we investigate the performance of PSO for
optimizing the PI controller gains for indirect control active
and reactive power of the DFIG with MPPT control algo-
rithm and compared with a PI controller which is tuned
manually.
2 Model of turbine
Normally a wind turbine creates mechanical torque on a
rotating shaft, while an electrical generator on the same
rotating shaft is controlled to produce an opposing elec-
tromagnetic torque. The power equation for the wind tur-
bine is as follows (Rahimi and Parniani 2010; Ghedamsi
and Aouzellag 2010):
P
v

1
2
qS
w
m
3
1
Where q is air density; S
w
is wind turbine blades swept the
area in the wind; m is wind speed.
The output mechanical power of a wind turbine is:
P
m
C
p
P
v

1
2
C
p
qS
w
m
3
2
Where C
p
represents the power coefcient.
C
p
can be described as (Abdin and Xu 2000):
C
p
b; k 0:5 0:0167 b 2
sin
p k 0:1
18:5 0:3b 2
_ _
0:00184k 3
b 2
3
It is a function of the tip speed ratio k and the blade pitch
angle b in a pitch-controlled wind turbine. k is dened as
the ratio of the tip speed of the turbine blades to wind
speed:
k
R X
t
m
4
Where R is blade radius, X
t
is the angular speed of the
turbine.
A gure showing the relation between C
p
, b and k is
shown in Fig. 1. The maximum value of C
p
(C
p max
0:5:)
is achieved for b = 2 degree and for k
opt
= 9.2.
3 Maximum power point tracking (MPPT)
Maximum power variation with rotation speed X of DFIG
is predened for each wind turbine. So for MPPT, the
control system should follow the tracking characteristic
curve (TCC) of the wind turbine (Abdin and Xu 2000).
Each wind turbine has TCC similar to the one shown in
Fig. 2. The actual wind turbine, X is measured and the
corresponding mechanical power of the TCC is used as the
reference power for the power control loop (Eltamaly et al.
2010).
The turbine speed should be changed with wind speed so
that the optimum tip speed ratio k
opt
is maintained. This
strategy called MPPT, (as shown in Fig. 2).
Int J Syst Assur Eng Manag
1 3
In order to make full use of wind energy, in low wind
speed b should be equal to 2 degrees. Figure 2 illustrates
the wind turbine power curve when b is equal to 2 degrees.
To extract the maximum power generated, we must x
the advance report k
opt
is the maximum power coefcient
C
p max
, the measurement of wind speed is difcult, an
estimate of its value can be obtained (Bekakra and Ben
attous 2011):
m
est

R X
t
k
opt
5
The aerodynamic power reference value must be set to the
following value:
P
aer ref

1
2
C
p max
q S
w
m
3
est
6
From Fig. 2 we can see there is one specic angular fre-
quency at which the output power of a wind turbine is
maximum occurs at the point where C
p
is maximized.
Connected all the maximum power point of each power
curve, the optimal power curve (MPPT curve) is obtained.
4 DFIG model
The general electrical state model of the induction machine
obtained using Park transformation is given by the follow-
ing equations (Bekakra and Ben attous 2011; Machmoum
and Poitiers 2009; Senthil Kumar and Gokulakrishnan
2011):
Stator and rotor voltages:
V
sd
R
s
i
sd

d
dt
/
sd
x
s
/
sq
V
sq
R
s
i
sq

d
dt
/
sq
x
s
/
sd
V
rd
R
r
i
rd

d
dt
/
rd
x
s
x /
rq
V
rq
R
r
i
rq

d
dt
/
rq
x
s
x /
rd
_

_
7
Stator and rotor uxes:
/
sd
L
s
i
sd
M i
rd
/
sq
L
s
i
sq
M i
rq
/
rd
L
r
i
rd
M i
sd
/
rq
L
r
i
rq
M i
sq
_

_
8
The electromagnetic torque is given as:
C
e
pMi
rd
i
sq
i
rq
i
sd
9
and its associated motion equation is:
C
e
C
r
J
dX
dt
10
The state variable vector is then:
X i
sd
i
sq
i
rd
i
rq
_
T
The state model can then be written as:
X
:
A X B U 11
Where:
A Must be an n-by-n matrix, where n is the number of
states
B Must be an n-by-m matrix, where m is the number of
inputs
with:
X
:

d
dt
i
sd
d
dt
i
sq
d
dt
i
rd
d
dt
i
rq
_ _
T
U V
sd
V
sq
V
rd
V
rq
_
T
0 2 4 6 8 10 12 14 16 18
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
X: 9.2
Y: 0.5
Lamda
C
p
B = 2
B = 3
B = 4
Fig. 1 Aerodynamic power coefcient variation C
p
against tip speed
ratio k and pitch angle b
0 100 200 300 400 500 600 700 800
0
500
1000
1500
2000
2500
3000
3500
4000
Turbine rotational speed (rpm)
T
u
r
b
i
n
e

m
e
c
h
a
n
i
c
a
l
e

p
o
w
e
r

(
W
)
MPPT
4 m/s
5 m/s
6 m/s
7 m/s
Fig. 2 Turbine powers various speed characteristics for different
wind speeds, with indication of the maximum power tracking curve
Int J Syst Assur Eng Manag
1 3
A
a
1
ax x
s
a
3
a
5
x
ax x
s
a
1
a
5
x a
3
a
4
a
6
x a
2

x
r
x
s
a
6
x a
4
x
r
x
s
a
2
_

_
_

_
;
B
b
1
0 b
3
0
0 b
1
0 b
3
b
3
0 b
2
0
0 b
3
0 b
2
_

_
_

_
where:
a
1 r
r
; a
1

R
s
rL
s
; a
2

R
r
rL
r
; a
3

R
r
M
rL
s
L
r
;
a
4

R
s
M
rL
s
L
r
; a
5

M
rL
s
; a
6

M
rL
r
b
1

1
rL
s
; b
2

1
rL
r
; b
3

M
rL
s
L
r
; r 1
M
2
L
s
L
r
5 Field oriented control of DFIG
In this section, the doubly fed induction machine (DFIM)
model can be described by the following state equations in
the synchronous reference frame whose axis d is aligned
with the stator ux vector /
s
, (/
sd
= /
s
and /
sq
= 0).
The control of the DFIG must allow a control inde-
pendent of the active and reactive powers by the rotor
voltages generated by an inverter. By neglecting resis-
tances of the stator phases the stator voltage will be
expressed by (Machmoum and Poitiers 2009):
V
sd
0 and V
sq
V
s
x
s
/
s
12
We lead to a decoupled power control; where, the
transversal component i
rq
of the rotor current controls the
active power. The reactive power is imposed by the direct
component i
rd
.
P
s
V
s
M
L
s
i
rq
13
Q
s

V
2
s
x
s
L
s
V
s
M
L
s
i
rd
14
The arrangement of the equations gives the expressions of
the voltages according to the rotor currents:
V
rd
R
r
i
rd
rL
r
di
rd
dt
gx
s
rL
r
i
rq
V
rq
R
r
i
rq
rL
r
di
rq
dt
g
M
L
s
V
s
gx
s
rL
r
i
rd
_

_
15
With:
T
r

L
r
R
r
; T
s

L
s
R
s
; g
x
s
x
x
s
The inverter connected to the rotor of the DFIG must
provide the necessary complement frequency in order to
maintain constant the stator frequency despite the variation
of the mechanical speed.
The system studied in the present paper is constituted of
a DFIG directly connected through the stator windings to
the network, and supplied through the rotor by a static
frequency converter as presented in Fig. 3.
6 Designing of PI controller using PSO
The PSO as an optimization tool provides a population-
based search procedure in which individuals called parti-
cles change their position (state) with time. In a PSO sys-
tem, particles y around in a multidimensional search
space. During ight, each particle adjusts its position
according to its own experience (This value is called P
best
),
and according to the experience of a neighboring particle
(This value is called G
best
), made use of the best position
encountered by itself and its neighbor (Lalitha et al. 2010)
(as shown in Fig. 4).
This modication can be represented by the concept of
velocity. The velocity of each agent can be modied by the
following equation:
v
k1
w v
k
c
1
rand P
best
x
k
c
2
rand
G
best
x
k

16
Using the above equation, a certain velocity, which
gradually gets close to P
best
and G
best
can be calculated.
The current position (searching point in the solution space)
can be modied by the following equation:
x
k1
x
k
v
k1
; k 1; 2; . . .; n 17
Where x
k
is current searching point, x
k?1
is modied
searching point, v
k
is current velocity, v
k?1
is modied
velocity. P
best
is the best solution observed by current
particle and G
best
is the best solution of all particles, w is an
inertia weight, c
1
and c
2
are two positive constants, rand is
a random generated number with a range of [0,1].
The following inertia weight is used (Lalitha et al.
2010):
wk w
max

w
max
w
min
k
max
_ _
k 18
where k
max
, k is the maximum number of iterations and the
current number of iterations, respectively. Where, w
min
and
w
max
are the minimum and maximum weights respectively.
Appropriate value ranges for c
1
and c
2
are 12, but 2 is the
most appropriate in many cases. Appropriate values for
w
min
and w
max
are 0.4 and 0.9 (Eberhart and Shi 2000)
respectively.
Int J Syst Assur Eng Manag
1 3
The PSO based approach to nd the global maximum
value of objective function as shown in Fig. 5.
The PI controller is a good controller in the eld of
machine control, but the problem is the mathematical
model of the plant must be known. In order to solve
problems in the overall system, several methods have been
introduced to tuning PI controller. Our proposed method
uses the PSO to optimize the active and reactive power
PI controller parameters, the PSO is utilized off line to
determine the controller parameters (K
p
and K
i
) (based on
quadrature rotor current error i
rq
linked to active power P
s
and direct rotor current i
rd
linked to reactive power Q
s
) of
the DFIG as shown in Fig. 6. The performance of the DFIG
varies according to PI controller gains and is judged by the
value of integral time absolute error (ITAE). The perfor-
mance index sum (ITAE) is chosen as objective function.
The purpose of stochastic algorithms is to minimize the
objective function. All particles of the population are
decoded for K
p
and K
i
.
Fig. 3 DFIG variable speed
wind energy conversion MPPT
control
V
k
X
k
V
k+1
X
k+1
V
Gbest
Gbest
k
Pbest
k
V
Pbest
X
Y
Fig. 4 Concept of a searching point by PSO
Start
Generate initial populations
Run the DFIG power control model
Calculate parameters
[K
i
, K
p
] of PI controller
Calculate the fitness function
Calculate the P
best
of each particle
and G
best
of population
Update the velocity, position, G
best
,
P
best
of particles
Stop
Yes
No
Print optimal parameters
[K
i
, K
p
]
Maximum iteration
number reached ?
Fig. 5 The owchart of the PSOPI control system
Int J Syst Assur Eng Manag
1 3
ITAE criterion is widely adopted to evaluate the
dynamic performance of the control system (Allaoua et al.
2009). The index ITAE is expressed in Eq. (19), as follows:
ITAE
_
1
0
t et j j dt 19
In this paper a time domain criterion is used for
evaluating the PI controller. The performance criteria used
for comparison between using a PI controller which is
tuned by PSO technique and manually. PI controller
includes integration absolute error (IAE) and integrated
of squared error (ISE).
The IAE and ISE performance criterion formulas are as
follows (Allaoua et al. 2009):
IAE
_
1
0
et j j dt 20
ISE
_
1
0
e
2
t dt 21
7 Simulation results
The DFIG used in this work is a 4 kW, whose nominal
parameters are indicated in Appendix.
The parameters of PSO algorithm are shown in Table 1.
The selecting of the parameters of PSO was as follows:
After several experiments, we obtained the following
The augmentation of swarm size leads to complicate
the calculation and a prolongation in computing time.
In our case swarm size = 15 is a good selection.
In our case, the number of maximum iteration = 20 is
satisfying for obtaining good results as shown in Fig. 7.
In many scientic papers c
1
= c
2
= 2, w
max
= 0.9 and
w
min
= 0.4 [for example in Ref. (Lalitha et al. 2010;
Eberhart and Shi 2000)].
The velocity, current position and tness value of each
optimal particle during the simulation are presented in
Table 2, after this table the best tness value is 3.943e ? 005
appeared in iteration number 9, and the optimal gains are
DFIG
Grid
FOC
+
Park
-1
Estimation
r

rd
V
U
dc
rq
V
*
s
P
rd
i
rq
i
*
rq
i
rq
i
*
0
s
Q =
*
rd
i
+
-
rd
i
s s
s
V
L

+
PI
PSO
Algorithm
PI
Win
Turbine
s
s
L
MV

s
s
L
MV

Gear
MPPT
0 10 20 30 40 50 60
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
Turbine rotational speed (rpm)
M
echanical power (W
)
8 m/s
7 m/s
6 m/s
6 m/s
*
s
P
r

-
+
-
, K K
p
i
s
P
+
-
s
P
*
s
P
PWM
Fig. 6 Indirect control of DFIG with PSO
Int J Syst Assur Eng Manag
1 3
K
i
= 1914.8 and K
p
= 1967.4, which are shown clearly in
Figs. 7 and 8.
Where the initial parameters of PSO are:
current position rand dim; n upbnd lwbnd
lwbnd velocity rand dim; n
where rand Random numbers, dim = 2 Dimension of
swarm (Ki and Kp), n = 15 Size of the swarm, upbnd:
1980 The Upper bound for the initial of the swarm, lwbnd:
1900 The Lower bound for the initial of the swarm.
The tness function variations and the variation of
optimal K
i
et K
p
gain during the simulation are presented in
Figs. 7 and 8 respectively.
In order to evaluate the MPPT control strategy, we
proposed a step change in wind speed is simulated in
Fig. 9, the wind speed starts at 5 m/s, at 3 s, the wind speed
suddenly changing at 6 m/s, as 6 s, the wind speed is
7 m/s.
The Fig. 10 presents the turbine speed. Figure 11 pre-
sents the power coefcient. Figure 12 presents the stator
active power without and with PSO resulting of the MPPT.
Figure 13 shows the stator reactive power without and with
PSO versus time. Figure 14 stator current without and with
PSO and these zoom. Figure 15 presents a comparison of
PI controllers without and with PSO for stator current at
startup. Figures 16 and 17 present spectrum of phase stator
current harmonics without and with PSO respectively.
Table 1 Parameters of PSO algorithm
Swarm size 15
Number of maximum iteration 20
c
1
= c
2
2
w
max
0.9
w
min
0.4
0 2 4 6 8 10 12 14 16 18 20
3.943
3.9435
3.944
3.9445
3.945
3.9455
3.946
3.9465
3.947
3.9475
3.948
x 10
5
X: 9
Y: 3.943e+005
Iteration
F
i
t
n
e
s
s

f
u
n
c
t
i
o
n
X: 1
Y: 3.948e+005
X: 5
Y: 3.945e+005
X: 20
Y: 3.943e+005
Fig. 7 The tness function variation during simulation
Table 2 The velocity, current position and tness value of each
optimal particle
Iteration
No.
Optimal parameters Fitness value
1 Velocity(1,1) = 0.1904 3.948e ? 005
K
p
= Current_position(1,1) = 1975.5
Velocity(2,1) = 0.0257
K
i
= Current_position(2,1) = 1975.8
5 Velocity(1,1) = 22.1064 3.945e ? 005
Current_position(1,1) = 1994.4 [upbnd
K
p
= upbnd = 1980
Velocity(2,1) = 26.5410
K
i
= Current_position(2,1) = 1963.2
9 Velocity(1,1) = -42.0899 3.943e ? 005
K
p
= Current_position(1,1) = 1967.4
Velocity(2,1) = 6.3232
K
i
= Current_position(2,1) = 1914.8
20 Velocity(1,1) = -42.0899 3.943e ? 005
K
p
= Current_position(1,1) = 1967.4
Velocity(2,1) = 6.3232
K
i
= Current_position(2,1) = 1914.8
0 2 4 6 8 10 12 14 16 18 20
1900
1910
1920
1930
1940
1950
1960
1970
1980
Iteration
K
i
,

K
p

g
a
i
n
Ki
Kp
Fig. 8 The variation of optimal K
i
et K
p
gain during simulation
0 1 2 3 4 5 6 7 8 9
0
1
2
3
4
5
6
7
8
Time (s)
W
i
n
d

s
p
e
e
d

(
m
/
s
)
Fig. 9 Wind speed proles
Int J Syst Assur Eng Manag
1 3
We observe after Figs. 9 and 10 when wind speed v is
5 m/s the optimal turbine speed X
t
of DFIG is 87.51 rad/s,
when v is 6 m/s, X
t
is 100 rad/s and when v is 7 m/s, X
t
is
112 rad/s, after each adjustment, the stable turbine speed
totally with the theoretical value. During this adjusting
process, realize the maximum wind energy tracking con-
trol. As can be seen from the gures, the stator active
power is controlled according to the MPPT strategy, and
the reactive power is maintained to zero, to guarantee a
unity power factor at the stator side. The power coefcient
C
p
is kept around its optimum (C
p
= 0.5), as shown in
Fig. 11.
A comparison is done here with the results obtained
from conventional PI controllers, which also aims at active
and reactive power ripple minimization. The results of the
comparison are that the active and reactive powers ripple is
reduced considerably with the help of PSO as shown in
Figs. 12 and 13. In addition, the stator active in the startup
is reduced in case of with PSO, as shown in Fig. 12b,
compared with a case of without PSO, as shown in
Fig. 12a.
Table 3 lists the performances of stator active power of
the two controllers (the gains of conventional PI controller
are calculated from the pole compensation method), from
these values obtained it is clearly visible that the error
magnitude obtained in different criteria for conventional
method is big as compared to the proposed tuning method
based on PSO algorithm, which is shown clearly in
Fig. 12.
Similarly, Stator current with PSO tuned PI controller is
smooth as compared with that of conventional PI control-
ler, or the over-current in the stator circuit is reduced in
start up when we are using the PSO as shown in Fig. 15.
From Figs. 16 and 17, it is clear that the stator current in
conventional PI has a high THD (with THD = 17.67 %) as
compared to the stator current in case when we are used
PSO technique (with THD = 15.70 %).
0 1 2 3 4 5 6 7 8 9
0
20
40
60
80
100
120
Time (s)
T
u
r
b
i
n
e

s
p
e
e
d

(
r
a
d
/
s
)
Fig. 10 Speed turbine according the MPPT with PSO
0 1 2 3 4 5 6 7 8 9
0
0.1
0.2
0.3
0.4
0.5
Time (s)
C
p
Fig. 11 Power coefcient C
p
variation with PSO
0 1 2 3 4 5 6 7 8 9
-4000
-2000
0
2000
4000
6000
8000
Time (s) Time (s)
S
t
a
t
o
r

a
c
t
i
v
e

p
o
w
e
r

(
W
)
Without PSO
reference
measured
0 1 2 3 4 5 6 7 8 9
-4000
-2000
0
2000
4000
6000
8000
S
t
a
t
o
r

a
c
t
i
v
e

p
o
w
e
r

(
W
)
With PSO
reference
measured
a b
Fig. 12 Stator active power injected in the grid according the MPPT: a without and b with PSO
Int J Syst Assur Eng Manag
1 3
0 1 2 3 4 5 6 7 8 9
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
Time (s)
a b
Time (s)
S
t
a
t
o
r

r
e
a
c
t
i
v
e

p
o
w
e
r

(
V
A
R
)
Without PSO
reference
measured
0 1 2 3 4 5 6 7 8 9
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
S
t
a
t
o
r

r
e
a
c
t
i
v
e

p
o
w
e
r

(
V
A
R
)
With PSO
reference
measured
Fig. 13 Stator reactive power: a without and b with PSO
0 1 2 3 4 5 6 7 8 9
-15
-10
-5
0
5
10
15
20
Time (s)
a b
Time (s)
S
t
a
t
o
r

c
u
r
r
e
n
t
s

(
A
)
Without PSO
4 4.02 4.04 4.06
-10
0
10
0 1 2 3 4 5 6 7 8 9
-15
-10
-5
0
5
10
15
20
S
t
a
t
o
r

c
u
r
r
e
n
t
s

(
A
)
With PSO
4 4.02 4.04 4.06
-10
0
10
Fig. 14 Stator current: a without and b with PSO
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-15
-10
-5
0
5
10
15
20
Time (s)
S
t
a
t
o
r

c
u
r
r
e
n
t

(
A
)
With PSO
Without PSO
Fig. 15 Comparison of PI controllers without and with PSO for stator
current at startup
0 20 40 60 80 100 120 140 160 180 200
0
0.2
0.4
0.6
0.8
1
Order of Harmonic (harmonic number)
M
a
g
n
i
t
u
d
e

(
%

o
f

F
u
n
d
a
m
e
n
t
a
l
)
Without PSO , THD = 17.67%
Fig. 16 Spectrum of phase stator current harmonics without PSO
Int J Syst Assur Eng Manag
1 3
8 Conclusion
In this work, we have presented a complete wind energy
conversion system made with a DFIG. This system is
constituted of a DFIG with a stator connected directly to
the grid while the rotor is connected through inverter
PWM. The aim of controlling the rotor side converter is to
extract a maximum power from the wind by using a eld
oriented control and an optimal speed reference which is
estimated from the wind speed. The PSO algorithm is then
used to nd the optimal gains of the PI controllers for the
active and reactive power in order to minimize the ripple in
the stator current. The performance index for various error
criteria for the proposed controller using PSO algorithm is
proved to be less than the controller tuned manually. It is
clear from the results that there is a reduction of ripple in
active and reactive powers as well as stator current when
the proposed PSO method is used.
Appendix
Appendix A: System parameters
Rated values: 4 kW, 220/380 V, 15/8.6 A.
Rated parameters: R
s
= 1.2 X, R
r
= 1.8 X, L
s
= 0.1554
H, L
r
= 0. 1568 H, M = 0.15 H, p = 2.
Wind turbine parameters are: R(blade radius) = 3 m,
G (Gearbox) = 5.4.
Air density: q = 1.22 kg/m
3
.
Appendix B: Nomenclature
v Wind speed
q Air density
R Blade radius
P
m
Mechanical power of wind speed
C
p
Power coefcient
S
w
Swept area
k Tip speed ratio
X
t
Angular speed of the turbine
C
e
Electromagnetic torque
C
r
Load torque
J Moment of inertia
b Bitch angle
V
sd,q
Stator d-q frame voltage
V
rd,q
Rotor d-q frame voltage
i
sd,q
Stator d-q frame current
i
rd,q
Rotor d-q frame current
/sd,q
Stator d-q frame ux
/rd,q
Rotor d-q frame ux
R
s
, R
r
Stator and rotor resistance
L
s
, L
s
Stator and rotor inductance
L
s
Mutual inductance
r Leakage factor
p Number of pole pairs
T
s
, T
r
Statoric and rotoric time-constant
x
s
, x Stator and rotor d-q reference axes speed
g Slip coefcient
References
Abdin ES, Xu W (2000) Control design and dynamic performance
analysis of wind turbine-induction generator unit. IEEE Trans
Energy Convers 15(1):9196
Ackermann (2005) Wind power in power systems. Wiley, Chichester
Allaoua B, Gasbaoui B, Mebarki B (2009) Setting up PID DC Motor
speed control alteration parameters using particle swarm opti-
mization strategy. Leonardo Electron J Pract Technol 14:1932
Bekakra Y, Ben attous D (2011) Active and reactive power control of
a DFIG with MPPT for variable speed wind energy conversion
using sliding mode control. World Acad Sci Eng Technol
(WASET) 60:15431549
Eberhart RC, Shi Y (2000) Comparing inertial weights and constric-
tion factor in particle swarm optimization. In: Proceedings of the
international congress on evolutionary computation, San Diego,
pp 8488
Eltamaly AM, Alolah AI, Abdel-Rahman Mansour H (2010) Mod-
ied DFIG control strategy for wind energy applications. In:
IEEE 2010 international symposium on power electronics,
electrical drives, automation and motion, pp 659653
0 20 40 60 80 100 120 140 160 180 200
0
0.2
0.4
0.6
0.8
1
Order of Harmonic (harmonic number)
M
a
g
n
i
t
u
d
e

(
%

o
f

F
u
n
d
a
m
e
n
t
a
l
)
With PSO , THD = 15.70%
Fig. 17 Spectrum of phase stator current harmonics with PSO
Table 3 Comparison between PI tuned manually and PI tuned by
PSO
Controller
generator
response
PI tuned manually
K
i
= 710.64;
K
p
= 2,000
PI tuned by PSO
K
i
= 1914.8;
K
p
= 1967.4
IAE 2,933 2,741
ITAE 1.047e ? 004 1.035e ? 004
ISE 3.139e ? 006 2.248e ? 006
Int J Syst Assur Eng Manag
1 3
Ghedamsi K, Aouzellag D (2010) Improvement of the performances
for wind energy conversions systems. Int J Electr Power Energy
Syst 32(9):936945
Gozde H, Cengiz Taplamacioglu M (2011) Automatic generation
control application with craziness based particle swarm optimi-
zation in a thermal power system. Int J Electr Power Energy Syst
33:816
Guo-qing W, Hong-jun N, Guo-xiang W, Jing-ling Z, Wei-nan Z,
Jing-feng M, Yang C (2010) On maximum power point tracking
control strategy for variable speed constant frequency wind
power generation. J Chongqing Univ (Engl Ed) 9(1):2128
Article ID: 1671-8224(2010)01-0021-08
Hammons TJ (2008) Integrating renewable energy sources into
European grids. Int J Electr Power Energy Syst 30:462475
Kennedy J, Eberhart R (1995) Particle swarm optimization. In:
Proceedings of IEEE international conference on neural net-
works, Perth, pp. 19421948
Lalitha MP, Reddy VCV, Usha V (2010) Optimal DG placement for
minimum real power loss in radial distribution systems using
PSO. J Theor Appl Inf Technol 13(2):107116
Lin FJ, Teng LT, Lin JW, Chen SY (2009) Recurrent functional-link-
based fuzzy-neural-network-controlled induction-generator sys-
tem using improved particle swarm optimization. IEEE Trans
Ind Electron 56(5):15571577
Lobos T, Rezmer J, Janik P, Amar H, Alonso M, Alvarez C (2009)
Application of wavelets and Prony method for disturbance
detection in xed speed wind farms. Int J Electr Power Energy
Syst 31:429436
Machmoum M, Poitiers F (2009) Sliding mode control of a variable
speed wind energy conversion system with DFIG. In: Interna-
tional conference and exhibition on ecologic vehicles and
renewable energies, MONACO, March 2629
Naka S, Genji T, Yura T, Fukuyama Y (2003) A hybrid particle
swarm optimization for distribution state estimation. IEEE Trans
Power Syst 18(1):6068
Qiao W, Venayagamoorthy GK, Harley RG (2006) Design of optimal
PI controllers for doubly fed induction generators driven by wind
turbines using particle swarm optimization. In: IEEE 2006
international joint conference on neural networks (IJCNN 06),
Georgia Institute of Technology, Atlanta, pp. 19821987
Rahimi M, Parniani M (2010) Dynamic behavior analysis of doubly-
fed induction generator wind turbinesthe inuence of rotor and
speed controller parameters. Int J Electr Power Energy Syst
32(5):464477
Senthil Kumar N, Gokulakrishnan J (2011) Impact of FACTS
controllers on the stability of power systems connected with
doubly fed induction generators. Int J Electr Power Energy Syst
33:11721184
Int J Syst Assur Eng Manag
1 3