Electric Power Systems Research 108 (2014) 5058

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Dynamic modeling and optimal control of DFIG wind energy systems

using DFT and NSGA-II
M. Zamanifar a, , B. Fani b , M.E.H. Golshan a , H.R. Karshenas a
a
b

Department of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan, Iran
Department of Electrical Engineering, Najafabad Branch, Islamic Azad University, Najafabad, Isfahan, Iran

a r t i c l e

i n f o

Article history:
Received 12 June 2013
Received in revised form 10 October 2013
Accepted 26 October 2013
Keywords:
Dynamic modeling
DFIG
Optimized control
DFT
NSGA-II

a b s t r a c t
Once a doubly-fed induction generator (DFIG) is subjected to a disturbance by a change in the wind
speed, the stator ux cannot change instantly. Under this condition, rotor back-EMF voltages reect the
effects of stator dynamics on rotor current dynamics, and have an important role on the oscillations of
the rotor current. These oscillations decrease the DFIG system reliability and gear lifetime. Moreover,
by focusing only on small signal analysis, the dynamic damping performance immediately following
such disturbances is often degraded. Additional improvement in performance will be achieved if discrete
Fourier transform (DFT) is used to quantify damping characteristic of the rotor current during changes of
the operating points. This paper introduces an optimization technique based on non-dominated sorting
genetic algorithm-II (NSGA-II) incorporating DFT analysis to achieve better control performance for DFIG
system stability. Considering small signal stability, the main purpose of the control system in the present
paper is to increase the system damping ratio as well as to guarantee enough stability margin. Eigenvalue
analysis and time-domain simulations have been presented to demonstrate that the proposed optimizing method yields better control performance in comparison with one designed using mere eigenvalue
relocation.
2013 Elsevier B.V. All rights reserved.

1. Introduction
In the development of wind turbine (WT) technologies, doublyfed induction generators (DFIGs) are becoming the dominant type
due to their advantages of variable speed operation, four-quadrant
active and reactive power capabilities, independent control of their
active and reactive output powers, high energy efciency, and low
size converters [14]. A diagram of a grid-connected DFIG-based
wind energy generation system is shown in Fig. 1, which is composed of a wind turbine and gear-box, a wound rotor induction
generator, a rotor-side converter (RSC) and a grid-side converter
(GSC). Grid-side converter works at the grid frequency, leading or
lagging in order to produce more or less reactive power while RSC
works at different frequencies, depending on the blades speed [5,6].
Consequently, the speed can be varied while the operating frequency on the stator side remains constant. Rotor-side converter is
used to control the generator speed and reactive power, whereas
the GSC is connected to the grid through a grid-side lter and is
used to control the DC-link voltage.

Corresponding author. Tel.: +98 331 2291111; fax: + 98 331 2291017.

E-mail address: meh.zamanifar@gmail.com (M. Zamanifar).
0378-7796/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.epsr.2013.10.021

Due to the popularity of DFIG systems for wind energy generation, control systems suitable for this application have been
extensively investigated [712]. However, the most popular and
practical control scheme of DFIGs is still eld-oriented control (FOC) based on proportional-integral (PI) controllers, which
enables decoupled control of real and reactive powers [1315].
FOC has been implemented in two ways. One way is to control the DFIG with stator ux orientation, and the other is with
air gap ux orientation. This paper deals with the analysis and
improvement of transient performance in the DFIG modeled with
the stator ux orientation. In this control scheme, the nonlinear cross coupling is eliminated with feed-forward compensation,
after which the machine model becomes linear and PI control
techniques can be applied. Thus, the active and reactive powers
can be controlled by the quadrature and the direct rotor current
components, respectively. Appropriate controller parameters are
needed to achieve better control performance for DFIG system stability. For this purpose, evolution algorithms have been used as
optimization tools in the DFIG controller parameters design procedure [1620]. For instance, in Ref. [17], genetic algorithm has been
applied to optimize the controller parameters of the RSC, and hence,
larger oscillations of the DC-link voltage cannot be avoided. Particle swarm optimization has been also employed to nd the optimal
control parameters in order to achieve optimal control of DFIG

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

b , s , r , t , 2 base, synchronous, rotor, turbine and rotor

slip angular frequency
Vdqs , Vdqr , Vdqg , edq stator, rotor, grid-side lter and back-EMF
dq-axis voltages
idqs , idqr , idqg stator, rotor and grid-side lter dq-axis currents
Ls , Lr , Lm stator and rotor self-inductances and mutual
inductance
stator and rotor resistance
Rs , Rr
Rg , Lg
grid-side lter resistance and inductance
dqs ,
dqr stator and rotor dq-axis uxes
Hr , Ht
inertia constants of generator and turbine
D, Ks , damping and shaft stiffness coefcient, and shaft
twist angle
, R, V , p air density, wind turbine blade radius, wind
speed, and pitch angle

Fig. 2. Phasor diagram of eld-oriented DFIG system.

multiple controllers in [18,19]. Another early algorithm uses bacteria foraging optimization to improve the damping of oscillatory
modes in the DFIG wind turbine [20]. However, single objective and
single operating point conditions have been considered in most
of the algorithms. Therefore, robust damping performance for
changed operating conditions cannot be obtained.
This paper introduces a new procedure for optimal controller
design of DFIG based on both eigenvalue analysis and DFT to
quantify the oscillations damping of DFIG transient response. Nondominated sorting genetic algorithm-II (NSGA-II) is used to obtain
the optimal controller parameters so as to obtain well damping performance as well as sufcient stability margin under variations of
operating points. It is found that the proposed optimizing method
yields a better control performance in comparison with a design
merely based on eigenvalue relocation.
2. DFIG model
The global rotating reference frame of the DFIG system by a d1 and q1 -axis rotating at the angular frequency of s is shown in Fig. 2.
The global reference frame is dened on innite bus bar with the
d1 -axis in the direction of the voltage space vector of Vinf . The local
reference frame of the stator ux is depicted by d2 - and q2 -axis
rotating at dynamic speed , in which the position of the d2 -axis
coincides with the maximum of the stator ux (i.e., ds = s and
qs = 0). Using the motor convention, the following set of equations
modeling the DFIG generator can be derived [21]:

ds

= b (Vds Rs ids +

qs )

(1a)

qs

= b (Vqs Rs iqs

ds )

(1b)

51

dr

= b (Vdr Rr idr + 2

qr )

(1c)

qr

= b (Vqr Rr iqr 2

dr )

(1d)

ds

= Ls ids + Lm idr ,

qs

= Ls iqs + Lm iqr

(1e)

dr

= Lm ids + Lr idr ,

qr

= Lm iqs + Lr iqr

(1f)

Te =

Lm (

qs idr

Ls

ds iqr )

Ps = Vdr idr + Vqr iqr ,

(1g)

Pr = Vdr idr + Vqr iqr

(1h)

Qs = Vds iqs Vqs ids

(1i)

2.1. Rotor modeling

From (1a)(1f), the rotor dynamics are given by:
Rr idr + 2 Lr iqr ed + Vdr
i dr = b
Lr

(2a)

Rr iqr 2 Lr idr eq + Vqr

i qr = b
Lr

(2b)

ed = Lm

eq = Lm

Vds + r

qs

Rs
Ls

ds /Ls

Vqs r

ds

Rs
Ls

qs /Ls

(2c)

(2d)
2

2 /L ), R = R + (L /L ) R and = . The
where Lr = Lr (Lm
s
r
m s
s
r
2
r
conguration of the controllers for the DFIG system is shown in
Fig. 3. As it can be seen, the actual dq current signals idqr are
ref to generate the error sigcompared with their reference signals idqr
nals, which are passed through two PI controllers and these signals

Fig. 1. Schematic diagram of DFIG system.

52

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

Fig. 3. DFIG control loops: rotor current, grid lter current, speed, reactive power and dc voltage control loops.

are then compensated by the corresponding cross-coupling terms

and back-EMF (electromotive force) voltages edq to form the rotor
voltage signals given as follows:
Vdr =

ref
KPidr (idr

idr ) + x5 2 Lr iqr

+ ed

(3a)

ref
iqr ) + x6 + 2 Lr idr + eq
Vqr = KP (iqr

(3b)

ref
idr )
x 5 = KIidr (idr

(3c)

iqr

iqr

ref
iqr )
x 6 = KI (iqr

(3d)

iqr

ref is generated from the

stator-side reactive power Qsref , while iqr
speed control loop [5]. Once in steady state (i.e., sd = 0 and = s ),

(1) leads to:

s

ds (Lm idr

ds )

Ls

Qs,gen =

The dq model of the GSC and lter are:

i dg = b

i qg = b

Lg

(5b)

x 15 = KI

Qs ) + x8

x 7 = KIr (rref r )

Lg

(7a)

(7b)

idg

(8a)

iqg

(8b)

ref
iqg ) x15 Lg idg + Vqs
Vqg = KP (iqg

Vs idr Lm
Ls

(6a)

Rg iqg Lg idg Vqg + Vqs

ref
Vdg = KP (idg
idg ) x14 + Lg iqg + Vds

x 14 = KI

pf
KP (Qsref

Rg idg + Lg iqg Vdg + Vds

Considering the grid-side lter current controllers to be

proportional-integral (PI) and under compensation of crosscoupling terms and stator voltages, Vdqs , the dq grid voltage
equations could be stated as:

(5a)

ref
iqr
= KPr (rref r ) x7

2.2. Grid-side lter modeling

Vs2
s Ls

The reactive power control loop is depicted in Fig. 4 and the

corresponding equations are described by:

ref
idr

The reactive power reference value is derived from the active

power reference and the desired value of the power factor [22].
During the entire test, Qsref is set to zero.

(4)

In this equation, Qs is divided into two parts: the magnetizing

part Qs,magn , and the reactive power part exchanged with the grid
Qs,gen . Suppose that = s and Rs = 0, then:
Qs,magn =

(6d)

iqr

where KPidr , KP and KIidr , KI

are the proportional and integrating gains of the rotor current control loops. With the stator
ref is calculated from the desired
ux-orientated reference frame, idr

Qs =

pf

x 8 = KI (Qsref Qs )

idg

ref
(idg
idg )

(8c)

iqg

ref
(iqg
iqg )

(8d)

ref , iref are dq grid-side lter current references. iref is set

where idg
qg
dg
ref is generated from the DC voltage control loop as
to zero, while iqg
follows:
ref
ref
iqg
= KPVdc (Vdc Vdc
) + x17

(9a)

ref
x 17 = KIVdc (Vdc Vdc
)

(9b)

(6b)
(6c)

Fig. 4. Reactive power control loop.

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

53

Fig. 5. (a) Mechanical power versus wind speed and (b) electrical power versus generator speed.

2.3. Drive train modeling

To represent the electrical and mechanical interaction between
DFIG and WT in transient stability studies, the two-mass model as
presented in [14] is considered as:
r =

Te + Tsh
2Hr

(10a)

t =

Tm Tsh
2Ht

(10b)

= b (t r )

(10c)

where Tsh = Ks + D(t r ), Te = Lm ds iqr /Ls

and Tm =
0.5R2 Cp V3 /SB t are the shaft, electromagnetic and mechanical
torques, respectively. The power coefcient Cp is given as:
Cp = 0.22


i =

=

 116
i

0.4p 5 e12.5/i

1
0.035

3
 + 0.08p
p + 1

t R
V

(11a)

1
(11b)

(11c)

pref p
T

pref = KP (Pt Ptref ) + x

(13a)

x = KI (Pt Ptref )

(13b)

In (13), Pt = (Pr + Ps ). During the entire test, Ptref is equal to 1

per unit (pu). When the generator speed exceeds the rated speed
(rrated ), the pitch control is activated and the pitch angle is tuned
so that the turbine power can be restricted to its rated value. Speed
control loop that satises the aforementioned features is depicted
in Fig. 6. It can be seen that the PI controller regulates iqr such that
it provides the reference angular rotor speed, rref . Depending on
the wind speed, rref is set as the rated value of the generator speed
rrated , in pitch angle control mode or at the optimal value derived
from MPPT control mode.
2.4. Stator modeling and interfacing with power grid

As can be seen in Fig. 5, if the wind speed is below the rated

value, the wind turbine control is achieved by driving the generator
speed along the optimum power-speed characteristic curve (zone
AB), which corresponds to the maximum energy captured from
the wind power. In other words, given a particular wind speed,
there is a unique rotational speed required to achieve the goal of
maximum power point tracking (MPPT). In this operating mode,
the wind turbine pitch control is deactivated, and the pitch angle
p is xed to zero. Once the rotor speed exceeds its upper limit,
the pitch controller will begin to increase the pitch angle to reduce
the mechanical power extracted from the wind (zone BC). The
mechanism governing the blade angle can be modeled as a rstorder system with a time constant T as:
p =

For pitch-controlled wind turbines the permissible range will lie

between 0 and +90 or even a few degrees to the negative side [23].
During the entire test in this paper, pref is kept at zero when the
wind speed is below the rated value, while it is determined by the
active power PI controller according to (13) when the wind speed
is higher than the rated value.

To complete the model of the grid-connected DFIG, two more

equations are required, namely, the stator modeling and the interfacing with the power grid. The equivalent grid impedance consists
of Le and Re . Its dynamics are obtained by:
i d = b

i q = b

Re id + Le iq Vds + Vdinf

(14a)

Le
Re iq Le id Vqs + Vqinf

(14b)

Le

where id = ids + idg , iq = iqs + iqg , as shown in Fig. 1 and Vdinf , Vqinf are
dq components of the innite bus voltages. Also
  is the difference
between the innite bus voltage angle,  s = b s dt, and the stator

ux angle is  = b dt (see Fig. 2). As a result,

 = b (s )

(12)

In the stator ux orientation,

Vqs Rs iqs ds = 0, ids = ( ds Lm idr )/Ls

Fig. 6. Speed control loop.

(15)

ds = b (Vds Rs ids ),
and iqs = Lm iqr /Ls .

54

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

Then, from (14a), the d-axis stator ux dynamic equation is

described by:

ds

= b

[(Rs + Re )(Lm idr

ds ) Re Ls idg

+ (Ls iqg Lm iqr )Le + Le ((Lm i dr Ls i dg )/b ) + Ls Vdinf ]

Moreover, from (14b) it turns out:

=

4. Optimization model

Ls b Vqinf Re Ls b iqg + (Rs + Re )Lm b iqr + Le (Lm i qr Ls i qg )

(Ls + Le )b

ds

+ Le b (Ls idg Lm idr )

(17)

Eqs. (15)(17) constitute the basis for interfacing DFIG with the
power grid.

The DC-link model can be deduced from the balance of real

power at the converter DC-link node as given by [14]:
Pr Pg
CVdc

4.1. Eigenvalue-based objective functions

The model of DFIG WT can be represented by a set of ordinary
differential equations as:
x = f (x, u)

(19)

where x and u are the vectors of DFIG WT system state and input
variables, respectively. They are dened as x = [ ds , , idr , iqr , x5 ,
x6 , x7 , x8 , r , , t , idg , iqg , x14 , x15 , Vdc , x17 , x , p ]T and u =
T

2.5. DC-link capacitor modeling

V dc = b

(16)

Ls + Le

ref , V , V , ref , Q ref , iref ] . Linearizing (19) around an operu = [Vdc

inf
r
s
dg
ation point, the system state matrix A can be calculated. To increase
the DFIG system damping ratio as well as to guarantee sufcient
stability margin, the parameters of the PI controller may be selected
to minimize the following eigenvalue-based objective functions.

(18)

where Vdc is the DC-link voltage, and Pg (= Vdg idg + Vqg iqg ) and Pr are
the powers supplied to the GSC and the rotor circuit, respectively.
It is noted that for a phase-locked loop (PLL) controller the corresponding bandwidth is relatively large and can be ignored in the
model [24].

3. Behavior analysis of DFIGs during changes of operating

points
The stator modes have the lowest damping ratios but their oscillation frequencies are high, and hence, usually their dynamics is
neglected in the literature. However, during the variations of the
operating points, these modes could have signicant effects on the
DFIG transient behavior and they should be considered during the
optimal design. According to the principle of constant ux linkages,
the stator ux cannot change instantaneously when the generator
is subjected to a disturbance by a change in the wind speed. This
leads to a DC component in the stator ux, which appears as an
oscillatory term when transferred into the synchronous reference
frame with a corresponding natural frequency near the network
frequency. Transient response of the stator ux can be divided into
steady state and oscillatory parts, which guarantees that no discontinuities appear in the stator ux when the operating point
changes [25]. According to (4), the oscillatory ux induces an oscillatory component in the d-axis rotor current (idr ), which can be
relatively large with a low damping ratio. However, by focusing
only on the small signal condition, the damping characteristic of idr
following a change in the operating point will be degraded. Additional improvement in DFIG dynamic performance is achievable if
DFT analysis is also utilized to quantify damping characteristic of
the rotor current. This paper focuses on the optimal tuning of the
DFIG controller parameters, which can improve the system damping performance immediately following a large disturbance. Three
sub-objective functions are applied for the optimal tuning of all
the controller parameters. One is to use an optimization technique
based on DFT analysis, which improves the rotor current dynamic
following a change in the operating point. Moreover, the other
functions are to increase the system damping ratio as well as to
guarantee enough stability margin, typically found in traditional
small signal stability analysis.

1
Min i

(20)

J2 (X) = Max i

(21)

J1 (X) =

where  i and i are the damping factor and the real part of the
ith eigenvalue, respectively, and X is the vector of PI controller
parameters. The objective function J1 makes sure that the minimum
damped eigenvalue is heavily damped and the system small signal
stability is ensured. The optimization aims to minimize J2 in order
to shift all the eigenvalues as far to the left of the left hand side of
the complex plane as possible. These objective functions consider
only an operating point, and hence, sufcient stability margin for
changed operating conditions cannot be obtained.
4.2. Application of DFT for damping enhancement
Additional improvement in performance will be achieved if DFT
is utilized for estimating the damping factor of the rotor current
during changes of operating points. It has been investigated in Ref.
[26] that well-damped oscillation signals have signicantly lower
amplitude in the frequency domain. Moreover, to achieve accurate damping factor estimates, the long data windows should be
selected. The window length in this study is 100 cycles (1.67 s in
60 Hz). As mentioned in the previous section, once the operating
point changes, an oscillatory component in the d-axis rotor current
will be induced at a frequency of 60 Hz. It follows that minimization
of the fundamental components of the d-axis rotor current will
result in a well-damped DFIG system response in the time domain.
Thus, the parameters of PI controller may be selected to minimize
the following DFT based objective function.
J3 (X) = |Idr [1]|

(22)

The above objective function along with J1 and J2 enhance the

small signal stability at different operating points. Fig. 7 shows the
optimization procedure of the tuning of DFIG control system.
5. Optimization using NSGA-II
Pareto-based tness assignment was rst proposed by Goldberg [27]. The idea is to assign equal probability of reproduction
to all non-dominated individuals in the population. The method

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

55

Fig. 7. Ofine PI gains tuning procedure.

consists of assigning rank 1 to the non-dominated individuals and

removing them from contention, then nding a new set of nondominated individuals, ranked 2, and so forth. NSGA-II differs from
a simple genetic algorithm only in the way the selection operation
is performed. The superiority of NSGA-II lies in the way multiple
objectives are reduced to a single tness measured by the creation of number of fronts, sorted according to non-domination. The
procedure of the optimization is designed as follows:
5.1. Step one: Initialization
iqr

iqr

pf

pf

In Section 2, there are 14 parameters, KPidr , KP , KIidr , KI , KP , KI ,

idg

iqg

idg

iqg

KPr , KIr , KP , KP , KI , KI , KPVdc and KIVdc for the controllers of WT

system in MPPT control zone and two more parameters, namely,

KP and KI in pitch angle control zone. It is supposed that KPidr =

iqr
KP ,

iqr
KI ,

KIidr

idg
KP

iqg
KP ,

idg
KI

iqg
KI .

=
=
=
The process is started by
randomly generating N individuals of the initial population around
the original values, which could be obtained by the poles placement
methodology [17]. Then, the lower and upper bounds, Kmin and
Kmax of the parameters should be specied to dene the range of
the searching space.

5.2. Step two: Evaluation

With the sub-objective functions dened in Section 4, the multiobjective optimization problem can be now formulated as:
J = Min{J1 , J2 , J3 }

(23)

Considering the small signal stability, the main purpose of (23)

is to increase the system damping ratio as well as to guarantee
enough stability margin.
5.3. Step three: The best compromise solution
In order to choose the optimal controller parameter among the
Pareto optimal set, a Fuzzy-based approach is employed in the
present paper. The kth objective function of a solution in a Pareto
optimal set Jk is represented by a membership function
k dened
as:

k =

Jk Jk,min
Jk,max Jk

J
J

k,max k,min
0

Jk,min < Jk Jk,max

(24)

Jk Jk,max

where Jk,max and Jk,min are the maximum and minimum values of
the kth objective function.
For each solution i, the membership function is calculated as:

= m k=1
n k
i

i=1

k=1

ik

(25)

where n is the number of objective functions and m is the number

of solutions. The solution having the maximum value of
i is the
best-compromised solution.
6. Simulation and results
In this section, some simulations are presented to conrm the
improvement in control performance of the DFIG system with the
proposed optimized parameters. The studies are done on the single
machine innite bus power system shown in Fig. 3, with the system
parameters given in Table 1.
6.1. Eigenvalue analysis
Eigenvalue analysis of the DFIG wind turbine system is discussed in the following, where the focus is on participation factor,
frequency, and damping ratio analysis. Using the system parameters presented in Table 1 and controller gains listed in Appendix
A, the system modes and corresponding variables with the modal
oscillation frequency, damping ratio and highest participation factors are obtained as in Table 2. It can be seen that the system is
stable, since all the eigenvalues have negative real parts. In particular, there are four modes of oscillation. The modes 12,13 are
associated with the mechanical variables x7 , t , p . These modes
are weakly damped. The second slowest modes 7,8 are electromechanical modes associated with the rotor variables r and . The
third oscillating modes 3,4 are electrical modes. Stator variables
ds and  have the highest contributions in these modes. The stator
modes have the lowest damping ratios, but their oscillation frequencies are high, and hence, usually their dynamics is neglected.
But, as mentioned earlier in Section 3, once the operating point
changes, the rotor back-EMF voltages reect the effects of stator
dynamics on rotor current dynamics, and have an important role
on oscillations of the rotor current. As can be seen in Table 2, the
mode 1 associated with idr is a non-oscillatory mode.
6.2. Time-domain simulations
In this paper, three methods for setting the DFIG parameters are
considered.
Method one: The PI controller parameters are tuned through pole
placement method [17].
Method two: The parameters are optimized through eigenvaluebased objective functions.
Method three: The controller parameters are obtained through
the proposed method.
The controller parameters are presented in Appendices AC
for methods one to three, respectively. The coordinated tuning
of the parameters is performed by NSGA-II at a base wind speed
of 6 m/s. Furthermore, to improve the system damping performance, the response of DFIG system under variation of 15% in wind

56

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

Table 1
Parameters of the 1.76-MVA, 575-V, 60-Hz DFIG WT.
Vb = 575 V, Sb = 1.76 MVA, fb = 60 Hz, b = 377 rad/s, s = 1 pu
Ht = 4.3 s, Hr = 0.75 s, ks = 0.6 pu/elec.rad, D = 1.2 pu,  = 1.225 kg/m3 , R = 34.93 m, rrated = 1.2 pu
Rs = 0.00706 pu, Ls = 3.07 pu, Lm = 2.9 pu, Rr = 0.005 pu, Lr = 3.056 pu
C = 12.72 pu, Vdc = 1200 V
Rg = 0.003 pu, Lg = 0.3 pu
Re = 0.05 pu, Le = 0.3 pu

P.U. system
Wind turbine
DFIG
Converter
Grid lter impedance
Network impedance

Table 2
Eigenvalues and participation factors (V = 15 m/s).
System mode

Oscillation frequency

Damping ratio

State variables with highest participation factors (%)

1 = 1193
2 = 1210
3,4 = 3.1 j368
5 = 602
6 = 63.5
7,8 = 2.83 j12.96
9 = 1259
10 = 1.94
11 = 1.97
12,13 = 0.38 j0.4
14 = 0.08
15 = 0.21
16 = 0.5
17 = 0.49
18 = 1
19 = 1

58.5

2.06

0.06

1
1
0.01
1
1
0.21
1
1
1
0.69
1
1
1
1
1
1

Rotor electrical
Filter electrical
Stator electrical
Rotor electrical
DC-link
Rotor electromechanical
Filter electrical
q-Axis rotor current control
d-Axis rotor current control
Speed control and turbine mechanical and pitch angle control
Active power control and pitch angle control
Speed control and pitch angle control
DC-link control
Reactive power control
q-Axis lter current control
d-Axis lter current control

speed is considered for evaluating the DFT-based objective function

included in method three. In the following, dynamic simulations are
carried out to observe the response of the DFIG system under a disturbance by a change in the wind speed. The system is subjected to
a disturbance by a change in the wind speed at t = 0.2 s, when V = 9
m/s. It amounts to 12% variation in the wind speed.
The dynamic of slip speed is simulated and the result is shown
in Fig. 8(a). In the following analysis, light blue, dark blue and
orange lines are related to methods one to three, respectively. It
has been investigated in Ref. [22] that the nature of instability in
DFIG systems is basically different from the rotor angle instability of

idr = 94%
iqg = 92%
ds = 50%,  = 49%
iqr = 99%
Vdc = 94%
r = 43%, = 49%
idg = 99%
x6 = 97%
x5 = 99%
x7 = 25%, t = 38%, p = 24%
x = 76%, p = 20%
x7 = 38%, p = 54%
x17 = 99%
x8 = 99%
x15 = 99%
x14 = 99%

synchronous generators. Normally, the DFIG speed range is about

30% around the synchronous speed. If the generator speed violates the above limits, the converter cannot handle the slip power
and the generator may become unstable. As it can be observed in
Fig. 8(a), although the generator slip speed is within the allowable
range in all methods, the lowest growth of the generator speed
occurs in method three. Also, Fig. 8(b) and (c) shows the stator ux
linkage and the rotor current oscillation. As expected, by using the
proposed method, a sensible reduction on the stator ux linkage
and the rotor current oscillation is observed when DFIG system
is subject to a disturbance. Moreover, it is important to study the

Fig. 8. Transient response of (a) generator slip speed, (b) stator ux linkage, (c) d-axis rotor current, and (d) DC-link voltage.

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

57

Fig. 9. Transient response of (a) rotor speed, (b) electromagnetic torque, (c) terminal voltage, (d)  (angle difference between innite bus voltage and stator ux angle), (e)
rotor active power, and (f) reactive power to the DFIG system.

oscillations in the DC capacitor voltage. It dictates the converter

rating of DFIG. The more the variation in the DC-link voltage, the
higher the rating of the converters should be. This aspect is important, since it also contributes toward maintaining the converter in
operation during a disturbance, given that DC-link over-voltage
may block converter operation. The DC-link voltage is shown in
Fig. 8(d). As can be seen, a reasonable reduction in the voltage
uctuation can easily be observed, when the gains are adjusted
by method three. Fig. 9(a)(f) shows that the proposed gain adjustment procedure is very efcient. In these gures, it can be observed
that the over-voltage on the DFIG terminal voltage is reduced, the
oscillations of the rotor speed and the electromagnetic torque is
smaller and a decrease in  oscillations is also veried. Moreover,
the dynamic response of rotor active power and line reactive power
are well improved whereas the oscillations after the disturbances
are well-damped.

this condition, rotor back-EMF voltages reect the effects of stator

dynamics on rotor current dynamics, and have an important role on
oscillations of rotor current. These oscillations reduce the DFIG system reliability and gear lifetime. This problem has been addressed
in this paper, by formulating and solving a multi-objective optimization technique based on NSGA-II combined with DFT analysis
to obtain optimal gains for the PI controllers used in the control
loop of the DFIG. With this procedure, it has been intended to
improve the global system dynamic behavior during wind speed
variations, also increasing the transient margin of DFIG system. In
order to validate the proposed strategy, modal analysis and time
domain transient simulation have been carried out, and the results
demonstrated the effectiveness of the proposed algorithm.

7. Conclusion

KPidr = KP

A novel algorithm has been proposed for the coordinated tuning

of the controller parameters of a DFIG WT. The optimization was
performed so that both robust damping performance and sufcient
stability margin for changing operating point conditions can be
achieved. Once DFIG is subjected to a disturbance such as wind
speed variations, the stator ux cannot change instantly. Under

Appendix A. Controller parameters obtained by pole

placement method
iqr

pf

= 0.85,

KI = 1.7,
KIr = 3.4,
KPVdc

iqr

KIidr = KI

= 1.7,

pf

KP = 1.7,

KPr = 13.6
idg

iqg

KP

= KP

= 3.4,

KIVdc

= 1.7,

= 1.7

idg

KI

iqg

= KI

= 1.7,

58

M. Zamanifar et al. / Electric Power Systems Research 108 (2014) 5058

Appendix B. Controller parameters obtained by rst

optimization method
iqr

KPidr = KP

= 1.52,

KPVdc

= 1.21,

pf

KP = 0.74,

KPr = 10.3

pf

KI = 2.54,
KIr = 2.52,

iqr

KIidr = KI

idg

iqg

KP

= KP

= 3.36,

KIVdc

= 2.04,

idg

KI

iqg

= KI

= 2.36,

= 2.1

Appendix C. Controller parameters obtained by second

optimization method
iqr

KPidr = KP

= 0.14,

KPVdc

= 2.16,

pf

KP = 0.74,

KPr = 9.44

pf

KI = 2.71,
KIr = 2.45,

iqr

KIidr = KI

idg

iqg

KP

= KP

= 3.26,

KIVdc

= 1.41,

idg

KI

iqg

= KI

= 1.44,

= 2.06

References
[1] R. Cardenas, R. Pena, S. Alepuz, G. Asher, Overview of control systems for the
operation of DFIGs in wind energy applications, IEEE Transactions on Industrial
Electronics 60 (2013) 27762797.
[2] M. Tazil, V. Kumar, R.C. Bansal, S. Kong, Z.Y. Dong, W. Freitas, H.D. Mathur,
Three-phase doubly fed induction generators: an overview, IET Electric Power
Applications 4 (2010) 7589.
[3] G.D. Marques, D.M. Sousa, Air-gap-power-vector-based sensorless method for
DFIG control without ux estimator, IEEE Transactions on Industrial Electronics
58 (2011) 47174726.
[4] M.Z. Kamh, R. Iravani, Three-phase steady-state model of type-3 wind generation unit-part I: mathematical models, IEEE Transactions on Sustainable Energy
2 (2011) 477486.
[5] V. Akhmatov, Analysis of dynamic behavior of electric power systems with
large amount of wind power, Technical University of Denmark, Kgs, Lyngby,
Denmark, 2003 (PhD thesis).
[6] G.D. Marques, D.M. Sousa, Stator ux active damping methods for eldoriented doubly fed induction generator, IEEE Transactions on Energy
Conversion 27 (2012) 799806.
[7] D.C. Gaona, L.M. Goytia, O.A. Lara, Fault ride-through improvement of DFIG-WT
by integrating a two-degree-of-freedom internal model control, IEEE Transactions on Industrial Electronics 60 (2013) 11331145.

[8] Z.S. Zhang, Y.Z. Sun, J. Lin, G.J. Li, Coordinated frequency regulation by doubly fed induction generator-based wind power plants, IET Renewable Power
Generation 6 (2012) 3847.
[9] A.J.S. Filho, E.R. Filho, Model-based predictive control applied to the doublyfed induction generator direct power control, IEEE Transactions on Sustainable
Energy 3 (2012) 398406.
[10] A.J.S. Filho, M.E.D.O. Filho, E.R. Filho, A predictive power control for wind energy,
IEEE Transactions on Sustainable Energy 2 (2011) 97105.
[11] S.Z. Chen, N.C. Cheung, K.C. Wong, J. Wu, Integral variable structure direct
torque control of doubly fed induction generator, IET Renewable Power Generation 5 (2010) 1825.
[12] J.P.D. Costa, H. Pinheiro, T. Denger, G. Arnold, Robust controller for DFIGs of
grid-connected wind turbines, IEEE Transactions on Industrial Electronics 58
(2011) 40234038.
[13] S. Li, T.A. Haskew, K.A. Williams, R.P. Swatloski, Control of DFIG wind turbine
with direct-current vector control conguration, IEEE Transactions on Sustainable Energy 3 (2012) 111.
[14] L. Yang, Z. Xu, J. Ostergaard, Z.Y. Dong, K.P. Wong, X. Ma, Oscillatory stability
and eigenvalue sensitivity analysis of a DFIG WT system, IEEE Transactions on
Energy Conversion 26 (2011) 328339.
[15] L. Yang, Z. Xu, J. Ostergaard, Z.Y. Dong, K.P. Wong, Advanced control strategy of
DFIG wind turbines for power system fault ride through, IEEE Transactions on
Power Systems 27 (2012) 713722.
[16] L. Yang, G.Y. Yang, Z. Xu, Z.Y. Dong, K.P. Wong, X. Ma, Optimal controller
design of a doubly-fed induction generator wind turbine system for small signal
stability enhancement, IET Generation, Transmission & Distribution 4 (2010)
579597.
[17] J.P.A. Vieira, M.V.A. Nunes, U.H. Bezerra, A.C. Nascimento, Designing optimal
controllers for DFIGs using a genetic algorithm, IET Generation, Transmission
& Distribution 3 (2009) 472484.
[18] Y. Tang, P. Ju, H. He, C. Qin, F. Wu, Optimized control of DFIG-based wind
generation using sensitivity analysis and particle swarm optimization, IEEE
Transactions on Smart Grid 4 (2012) 112.
[19] F. Wu, X.P. Zhang, K. Godfrey, P. Ju, Small signal stability analysis and optimal
control of a wind turbine with doubly fed induction generator, IET Generation,
Transmission & Distribution 1 (2007) 751760.
[20] Y. Mishra, S. Mishra, M. Tripath, N. Senroy, Z. Dong, Improving stability of a
DFIG-based wind power system with tuned damping controller, IEEE Transactions on Energy Conversion 24 (2009) 650660.
[21] P.C. Krause, Analysis of Electric Machinery, McGraw-Hill, New York, 1986.
[22] M. Rahimi, M. Parniani, Dynamic behavior analysis of doubly-fed induction generator wind turbines the inuence of rotor and speed controller parameters,
Electrical Power and Energy Systems 32 (2010) 464477.
[23] T. Ackermann, Wind Power in Power Systems, John Wiley & Sons, New York,
NY, USA, 2005.
[24] B. Pokharel, Modeling, control and analysis of a doubly fed induction generator
based wind turbine system with voltage regulation, Tennessee Technological
University, USA, 2011 (MS thesis).
[25] P.S. Flannery, G. Venkataramanan, A fault tolerant DFIG wind turbine using a
parallel grid side rectier and series grid side converter, IEEE Transactions on
Power Electronics 23 (2008) 11261134.
[26] E. Brigham, The Fast Fourier Transform and its Application, Prentice-Hall, Englewood Cliffs, 1988.
[27] S. Panda, Multi-objective PID controller tuning for a FACTS-based damping stabilizer using non-dominated sorting genetic algorithm-II, Electrical Power and
Energy Systems 33 (2011) 12961308.