Proceedings of the 3rd International Conference on ThBB.

4
Systems and Control, Algiers, Algeria,
October 29-31, 2013

Genetic Algorithm Optimized PI and Fuzzy Logic Speed Vector

Control of Dual Stator Induction Generator in
Wind Energy Conversion System
Fatima.Ameur , Katia. Kouzi
Semi conductors and functional materials Laboratory,
Electrical Engineering Department, Laghouat
University BP 37G, Ghardaia Street, Algeria
ameurfatima1@gmail.com katia_kouzi@yahoo.fr

Abstract—This paper focuses on the design and analysis
ofoptimized PI conventional and PI fuzzy logic speed vector
control of Dual Stator Induction Generator (DSIG) in wind
energy conversion system. In first step a field-oriented control
of a DSIG is presented. In second step, in order to ensure an
optimum operating point and a Maximum Power Point Track
(MPPT) giving online a maximum production of electric power
for different wind speeds, a conventional PI and then a fuzzy PI
speed regulators are used. In the final step, to improve the
design of these two regulators, the technique of genetic
algorithms is used to facilitate the design parameters
determination and reduced time consumption comparatively to
the "trial-error" method.
Simulation results show clearly the effectiveness and the
performance of the suggested control algorithm.
Index term: Dual stator Induction Generator, Variable Speed
Wind Turbine, Field oriented control, Conventional PI and PI
fuzzy controllers, Optimization Method, Genetic Algorithms
,PI and PI fuzzy controllers optimized.
I.INTRODUCTION
In order to meet increasing power demand, taking into
account economical and environmental factors, wind energy
conversion is gradually gaining interest as a suitable source
of renewable energy. At the same time there has been a rapid
development of related wind energy technology [1],[2].An
induction generator, with its lower maintenance demands
and simplified controls, appears to be an effective solution
for small hydro and wind power plants [3].
To increase the power rating of an ac drive system a
multi-phase induction machine is seemed an ultimate
solution. Multiphase (more than three phases) drives possess
several advantages over conventional three-phase drives,
such as reducing the amplitude and increasing the frequency
of torque pulsations, reducing the rotor harmonic currents,
reducing the current per phase without increasing the voltage
per phase, lowering the dc-link current harmonics, and
higher reliability. By increasing the number of phases it is
also possible to increase the power /torque per rms ampere
for the same volume machine [3]. A common type of multi-
phase machine is the dual stator induction machine (DSIM),
also known as the six phase induction machine.
Generally, variable speed wind energy conversion
system with the DSIG require bothwide operating range of
speed and fast torque response, regardless of any
disturbances anduncertainties (turbine torque variation,
parameters variation and un-modeled dynamics). This leads
to more advanced control methods to meet the real demand
[3].
The fuzzy logic control (FLC) is basically nonlinear and
adaptive in nature, giving robust performance under
parameter variation and load disturbance effect.
As an intelligent control technology, fuzzy control
provides a systematic method to incorporate human
experience and implement nonlinear algorithms,
characterized by a series of linguistic statements, into the
controller [4].
The major drawback of FLC is the lack of design
techniques. Most fuzzy rules are based on human knowledge
and differ among persons despite the same system
performance. The selection of suitable fuzzy rules,
membership functions, and their definitions in the universe
of discourse invariable involves painstaking trial-error [5].
We address using genetic algorithms to overcome these
drawbacks to make design tasks easier. Genetic Algorithms
(GAs) is a search heuristic that mimics the process of
evaluation. GAs can be applied to process controllers for
their optimization using natural operators [6].
The main objective of the use genetic algorithms is to
facilitate the design parameters determination and reduced
time consumption comparatively to the "trial-error" method.
Such an optimal FLC could provide ideal control
performance and achieve desired speed.
This paper is constructed as follows: in section II, the
modelling of the wind generator and the MPPT are
presented. Section III deals with the field oriented control
(FOC) of a DSIG. The design of a FLC for speed regulation
of a DSIG is presented in section IV. In section V genetic
algorithms based PI conventional and FLC is proposed. In
section VI the performances of the proposed control are
illustrated by some simulation results. Finally some
concluding remarks are given in section VII.
II. MODELING OF THE WIND GENERATOR

978-1-4799-0275-0/13/$31.00 ©2013 IEEE

Proceedings of the 3rd International Conference on ThBB.4

A. Modeling of the Wind Turbine and Gearbox The wind speed varies over time, and to ensure
Wind turbine mechanical power is expressed as follows [7], maximum capture of wind energy incident, the speed of the
[8]: wind turbine should be adjustable permanently with that of
the wind [9].
Pt  C p     SV 3 (1)

Where C p the power coefficient of the turbine, ρ is the air

density, R is the blade length and V is the wind velocity.
The turbine torque is the ratio of the out power to the
shaft speed t , given by:

Pt
Tt 
t
The turbine is normally coupled to the generator shaft
through a gearbox whose gear ratio G is chosen in order to
set the generator shaft speed within a desired speed range.
Neglecting the transmission losses, the torque and shaft
speed of the wind turbine, referred to the generator side of
the gearbox, are obtained as follows:
T 
Tg  t , t  r (3)
G G
Where the Tg driving is torque of the generator and  r
is the generator shaft speed.
The captured wind power is not converted totally by the
wind turbine. C p ( ) Give us the percentage converted
(2)
Fig. 1 Control of turbine
B. Dual Stator Induction Machine Model
The model of dual stator induction machine is composed
of a stator with two identical phase windings shifted by an
electric angle   30°, and a squirrel cage rotor.
Under the assumptions of magnetic circuits linearity, and
assuming sinusoidal distributed air-gap flux density, the
equivalent two-phase model of dual stator induction
machine, represented in asynchronous frame (d,q) and
expressed in state-space form, is a fourth-order model [8]-
[10]:
 I    L
 
1
 BU   gl C  I    D[I ] (6)

which is function of the wind speed, the turbine speed and Where:
the pith angle of specific wind turbine blades [9],[10]. t
U   Vqs1Vds1Vqs 2Vds 2VqrVdr  ;
Although this equation seems simple, C p is dependent t
 I    I qs1I ds1I qs 2 I ds 2 I qr I dr  ;
on the ratio  between the turbine angular velocity t and
d I 
the wind speed V .this ratio is called the tip speed ratio I  
  dt
expressed by:
 
B  diag 1 1 1 1 0 0
t R
 (4) 0 0 0 0 0 0 
V 0
 0 0 0 0 0 
The areodynamique torque (wind) is determined the 0 0 0 0 0 0 
following equation [9]: C    0 0 0 0 0 0 


Paero L 0 Lm 0 0 Lr  Lm 
Tt   C p ( ) S  V 3 / 2t (5)
t  m 
 0 Lm 0 Lm   Lr  Lm  0 
From the previous equations, a functional block diagram
model of the turbine is established. It shows that the turbine
rotation speed is controlled by acting on the electromagnetic
torque of the generator. The wind speed is considered an
entry disruptive to this system (see Fig.1).

978-1-4799-0275-0/13/$31.00 ©2013 IEEE

Proceedings of the 3rd International Conference on ThBB.4

 Ls1  Lm 0 Lm 0 Lm 0 
 0 Ls1  Lm 0 Lm 0 Lm 

 L 0 Ls 2  Lm 0 Lm 0 
 L   0m Lm 0 Ls 2  Lm 0 Lm 


 L 0 Lm 0 Lr  Lm 0 
m
 
 0 Lm 0 Lm 0 Lr  Lm 

 Rs1 s  Ls1  Lm  0 s Lm 0 s Lm 

 
s  Ls1  Lm  Rs1 s Lm 0 s Lm 0 
 0 s Lm Rs 2 s  Ls1  Lm  0 s Lm 
 D   
 s Lm 0 s  Ls1  Lm  Rs 2 0 0 
 0 0 0 0 Rr 0 
 
 0 0 0 0 0 Rr 
The mechanical modeling part of the system is given
The reference voltages are expressed by [9]
by [10]:
d r vd* _ ond  vdg
*
 vdg  s Lt iqg
J  Tr  Tem  f r r (7) (17)
dt vq* _ ond  vqg
*
 vqg  s Lt idg
With:
To maintain constant the dc link voltage, we have
 p   Lm   recourse to use a proportional integral corrector. It is
Tem    
 2   Lmd  Lr 
 
  iqs1  iqs 2  dr  ids1  ids 2  qr  parameterized according to the capacitor value and the
dynamics of the regulation loop. Network reference currents,
(8)
expressed in d–q frame, are given by :
III.1 GRID SIDE POWER CONTROL * *
*
Pg vdg  Qg vqg
In grid-connected control mode, all the available power idg  2 2
that can be extracted from the wind generator is transferred vdg  vqg
(18)
to the grid. Standard PI controllers are used to regulate the Pg*vqg  Qg* vdg
dc link voltage and the inverter output currents in the (abc) *
iqg  2 2
synchronous frame. To have the grid current vector in phase vdg  vqg
with the grid voltage vector, the reference reactive power Q
should be zero. The dc link voltage control is acting to III. FIELD ORIENTED CONTROL OF A DSIM
supply the reference active power. The output of the current According to the field orientation theory [11], the
controllers sets the voltage reference for an average
machine currents are decomposed into i ds and i qs
conversion control method that controls the switches of the
grid inverter [9]. components, which are respectively, flux and torque
components. The key feature of this technique is to keep
The dc link voltage is given by:
namely dr  r and qr  0 .
dudc 1
 (idc  iond ) (14)
dt Cdc Hence, the flux and the electromagnetic torque are
Where, decoupled from each other, and can be separately controlled
as desired. Then the drive behavior can be adequately
ic*  idc  iond (15) described by a simplified model expressed by the following
equations [9]:
The reference active power injected to the electrical supply
network is given by: r* Lm
idr   (ids1  ids 2 ) (9)
Pg*  udc idc  udcic* (16) Lm  Lr Lm  Lr

978-1-4799-0275-0/13/$31.00 ©2013 IEEE

Proceedings of the 3rd International Conference on ThBB.4

Lm 2. Fitness: the fitness or objectif function defined by

iqr   (iqs1  iqs 2 ) (10)
Lm  Lr this equation [15]:
 tf 
rr Lm  iqs1  iqs 2  
min f obj 

1

*
Ωr  Ωr 
2
dt


sl*   
2
(11)  *
Ωr 
( Lm  Lr ) r*  0


Finally the electromagnetic expression can be represented (18)

by: 3. Selection: we use the stochastic selection [15].
Lm 4. Crossover: in this step we use the scattered
*
Tem P
Lm  Lr

iqs1  iqs 2 .r*  (12) crossover its probability equal 0.8.
5. Mutation: in this step we choose the uniform
IV. DESIGN OF FLC FOR DUAL STATOR INDUCTION mutation and its rate equal 0.001.
GENERATOR SPEED CONTROL
The structure of a standard FLC can be seen as a
traditional PI controller, where the speed error e and its
variation e are considered as input linguistic variables and
the electromagnetic reference torque change k Tem is
considered as the output linguistic variable [12],[13].
For convenience, the inputs and output of FLC were
scaled with three different coefficients k e , k e , k Tem .
These scaling factors can be constants or variables, and play
an important role for FLC design in order to achieve a good
behavior in both transient and steady state.
Seven membership functions with overlap, of triangular Fig.2 Basic structure of the genetic algorithms based fuzzy
shape and equal width, are used for each input and output logic controller for indirect field oriented control of DSIG.
variable, so that a 49 rules base is created (see tableI). The
sum-product inference algorithm is selected to complete the The algorithm restore until the stop condition is reached
fuzzy procedure. And the FLC output is obtained by the
gravity center defuzzification method [14].
  10  , we take the best individual parameter controller.
6

(The number of generations equal 200 and the population

Table I Inference matrix size equal 20).

VI. SIMULATION RESULTS AND DISCUSSION

V.GENETIC ALGORITHMS BASED CONVENTIONAL
AND FUZZY LOGIC CONTROLLER
The genetic algorithm is applied to automate and
optimize the PI conventional and FLC design process (see
Fig.2). This optimization requires a predefined of different
parameter of GAs sash us coding, fitness, selection,
mutation, and specifying of (Crossover probability, mutation
rate, Population size and Number of generations), that detail
explain in flowing:
1. Coding: we use real coding of each individual
composed by genes.
In order to verify the validity of the proposed controllers,
the computer simulation results for a 1.5 MW DSIG using a
PI and Fuzzy controllers optimized by the GA technique is
compared to a conventional PI and fuzzy controllers whose
parameters are determined by trial-error" method. The dual-
stator induction generator parameters used in the simulation
are given in the Appendix B.
The results of simulations are obtained for reactive
power Q⁄ = 0 and DC link voltage Udc = 1130 V: Figure.3
shows the angular speed of the DSIG. The waveform of the
generator torque is shown in Fig.4. The stator currents and
voltages waveforms of the DSIG and the related expended
plots are shown, respectively, in Fig.5. The feature of vector
control is illustrated in Fig.6 and the grid voltage and
current are given in Fig.7, The DC link voltage is maintained
at a constant level (1130 V) (see Fig.8). It is clear from
figures10, 12 that the introduction of the GA can improve
greatly the performance of suggested system. In fact, it can
be noted that the specific speed and coefficient speed are
considerably ameliorated, which leads us to say that

978-1-4799-0275-0/13/$31.00 ©2013 IEEE

Proceedings of the 3rd International Conference on ThBB.4

optimization by GA gives us the possibility of designing a Table III Normalization factors of FLC found by genetic algorithms.
powerful PI and fuzzy controllers by optimizing their
parameters. Hence that the real power extracted from the ke k e k c
wind energy conversion systems can pass through the grid. FLC 0.14 0.003 8.584
Figure 13 illustrates the grid active and reactive powers. FLC-GA 6 4 70000

Table II. Normalization factors of PI found by genetic algorithms. Table IV. Summary of result

ki kp Cp error  error The necissaire time to

PI 200000 40 % % restore its initail value%
PI-GA 300000 500
PI 0.07 2.4 15
GA-PI 0.04 2.4 15
FLC 0.0001 0.36 0.2
GA-FLC 0.0001 0.36 0.001

180 2

160 0
r* , r (rad/s)

Tem* , Tem (KN.m)

-2
140
-4
120
-6
100
 * T
r -8 em
80  T
em
*
r
-10
0 200 400 600 0 200 400 600
T(s) T(s)

Fig.3 DSIG speed and its reference.

Fig.4 DSIG Torque and its reference.

4000 3
3000
2
dr , qr (Wb)

2000
ias1 (A) , 4Vas1(V)

1000 1
0
0
-1000

-2000 
Vas1
-1 dr
-3000 
ias1 qr
-4000 -2
0 100 200 300 400 500 600 0 200 400 600
T(s) T(s)
Fig.6 Direct and quadratic rotor flux.
Fig.5 Stator voltage and current for stator 1.
3000 1200

1000
2000
ig1 (A) , 4Vg1 (V)

800
1000
Udc(V)

600
0
400
-1000 200
Vg1
ig1
-2000 0
0 200 400 600
T(s)
-3000
0 200 400 600
T(s) Fig. 8 DC link voltage

Fig. 7 Grid voltage and current

978-1-4799-0275-0/13/$31.00 ©2013 IEEE

Proceedings of the 3rd International Conference on ThBB.4

9.001 0.5241

9.0008
0.524

Cp
9.0006


0.524
9.0004

9.0002 0.524
0 200 400 600 0 200 400 600
T(s) T(s)
Fig.9 The tip speed and coefficient speed with classical PI.
0.5236
9.0006
0.5236
9.0004
0.5236
9.0002
0.5236

Cp


9
0.5236
8.9998
0.5236
8.9996

8.9994 0.5236
0 200 400 600 0 200 400 600
T(s) T(s)

Fig.10 The tip speed and coefficient speed with PI optimized.

9 0.5236

9 0.5236

9 0.5236
Cp

0.5236
9


0.5236
9
0.5236
9
0.5236
9 0 200 400 600
0 200 400 600
T(s) T(s)
Fig.11 The tip speed and coefficient speed with fuzzy PI.
9 0.5236

9 0.5236

9
0.5236
Cp

9


0.5236
9
0.5236
9

0.5236
0 200 400 600 0 200 400 600
T(s) T(s)
Fig.12 The tip speed and coefficient speed with PI fuzzy optimized.
VII. CONCLUSION
0.5
A novel genetic algorithm optimized PI and Fuzzy logic
Pg (MW) , Q g(MVAR)

0 speed vector control of dual stator induction generator in

-0.5 wind energy conversion system connected to the grid is
presented in this paper. The genetic algorithms was
-1 proposed to automate and optimize the PI and fuzzy logic
-1.5
controllers, and to impose the ideal feedback control
solution despite model uncertainties. The validity of the
-2 Pg proposed controllers is confirmed through the simulation
Qg results. It testifies that optimization method is not only
-2.5
0 200 400 600 robust, but also can improve dynamic performance of the
T(s)
system. The GA-PI and GA-FLC proposed approach
Fig. 13 Grid active and reactive power achieves:
 Good pursuit of reference speed;

978-1-4799-0275-0/13/$31.00 ©2013 IEEE

Proceedings of the 3rd International Conference on ThBB.4

 Starting without over shoot;

 Good support for changes of the turbine and the G = 90.
generator as well as to electric grid disturbances. REFERENCES

APPENDIX.A. NOMENCLATURE [1] M. Benadja ,A Chandra, “A New MPPT Algorithm for PMSG Based
Grid Connected Wind Energy System with Power Quality” in
Proc.5th Annu. IEEE Conf, Power India,Morthal,India , Dec 19-22
G Gear ratio 2012, pp.1-6.
V Wind velocity [2] R.Pena, R.Cardenas, R. Blasco, G. and J. Clare “A cage induction
Pn Nominal power generator using back to back PWM converters for variable speed grid
connected wind energy system”, in Proc. 27th Annu. IEEE Conf,
 Tip speed ratio Industrial Electronics Society (IECONO1), Denver, CO, États-Unis,
S Area of the rotor Nov29-Dec 02 2001, vol.2, pp. 1376 - 1381.
Power coefficient
[3] G. K. Singh, K. Nam, S. K. Lim, “A Simple Indirect Field-Oriented
Cp Control Scheme for Multiphase Induction Machine” in Proc. 27th
Annu. IEEE Conf, Industrial Electronics Denver, CO, États-Unis, Aug
r Mechanical speed of the DSIG
2005, vol.52, No. 4, pp. 1177 - 1184.
t Turbine speed [4] K. Kouzi, M – S. Naϊt-Saϊd, M.Hilairet, E. Berthlot., " A Fuzzy
Sliding-Mode Adaptive Speed Observer for Vector Controlof an
Aerodynamic torque
Tt Induction Motor,” IEEE-IECON, Annu. Meeting, Florida , USA, pp.
362-366, Nov 2008.
Tg Generator torque [5] I.K.Bousehana,A.Hazzab,ABoucheta,B.Mazari,R.Mostefa, “Optimal
Fuzzy Self-Tuning of PI Controller Using Genetic Algorithm for
Rs1 , Rs 2 Per phase stators resistances Induction Motor Speed Control”, Int. J. of Automation
Technology,Vol.2,No.2, 2008, pp. 85-95.
Per phase stators leakages inductances
L1 , L2 [6] R .Malhotra, N.Singh ,Y.Singh “Genetic Algorithms: Concepts,
Magnetizing inductance Design for Optimization of Process Controllers”,Canadian Center of
Lm Science and Education Trans, Vol. 4, No. 2,March 2011, pp. 39–54.
Rr Per phase rotor resistance [7] A. Abo-Khalil, D.C.Lee, J.K.Seok “Variable Speed Wind Power
Generation System Based on Fuzzy Logic Control for Maximum
Lr Per phase rotor leakage inductances Output Power Tracking” in Proc. 35th Annu. IEEE Conf. Power
Electronics Specialists ,Aachen, Germany,Vol.3,pp 2039-2943, June
J Inertia (turbine + DSIG),
20-25 2004.
fr Viscous coefficient [8] S.Chekkal, D.Aouzellag, K.Ghedamsi, H.Amimeur “New control
strategy of wind generator based on the dual-stator induction
P Number of pole pairs generator” in Proc. 10th Annu. IEEE Conf. Environmental and
electrical engineering (EEEIC), Rome, Italy, May 8– 11 2011. pp. 68–
tf final time 71.
[9] H. Amimeur D. Aouzellag ,R. Abdessemed K. Ghedamsi “Sliding
p Derivative operator mode control of a dual-stator induction generator for wind energy
s Speed of the synchronous reference conversion systems” ELSEVIER Trans, electrical Power and Energy
frame Systems, Vol.42, Issue 1, Nov 2012,pp. 60–70.
r Rotor electrical angular speed [10] V. Pant, G.K. Siugh, S.N. Singh “Modeling Of A Multi-Phase
Induction Machtne Under Fault Condjtion” in Proc. IEEE Conf,
* Electromagnetic torque reference Power Electronics and Drive Systems, July 1999, vol.1,pp. 92 - 97.
Tem
proceeding of IEEE International Conference on Power Electronics
Vqs1Vds1 Vqs 2 , “d–q” stators voltages and Drive Systems, PEDS'99, , Hong Kong.
[11] P.Vas “Sensorless Vector_and_Direct_Torque” Oxford
Vds 2 University Press, U.K , 1998, Ch.4, pp. 267.
“q” stators currents [12] K. Kouzi, L. Mokrani, M-S. Naït-Saïd, " A New Design of Fuzzy
I qs1 I ds1 I qs 2 I ds 2 Logic Controller With Fuzzy Adapted Gains Based on Indirect
Vector Control for Induction Motor Drive,” IEEE-SSST, Annu.
Vqr Vdr “d–q” rotor voltages
Meeting,WestVirginia , USA, pp. 362-366, March. 2003.
“d–q” rotor currents [13] K. Kouzi, L. Mokrani, M-S. Naït-Saïd, " High Performances of
I qr I dr Fuzzy Self-Tuning Scaling Factor of PI Fuzzy Logic Controller Based
NS,NS,NS Negative Small, negative medium, negative on Direct Vector Control for Induction Motor Drive without Flux
large Measurements", Proceeding of IEEE International Conference on
Industrial Technology (ICIT),pp. 1106-1111, Dec 2004.
PS,PM,PL Positive Small, positive medium ,positive
large [14] L.Mokrani, K.Kouzi, “ Influence of The Fuzzy Adapted Scaling
Factors on The Performance of A Fuzzy Logic Controller Based on
Z Zero
An Indirect Vector Control for Induction Motor Drive ”In JEEC,
Journal of Electrical Engineering Control , Vol. 55, No.7-8,
APPENDIX.B PARAMETERS 2004,Slovaquie, pp. 188-194.
[15] B.Toual, “ Modeling and Optimized Fuzzy Control of a Doubly Fed
DSIG: 1.5 MW, 400 V, 50 Hz, 2 pole pairs, Rs1 = Rs 2 = 0.008 X, L1 = Induction Generator, Driver by a Variable Speed Wind Turbine ”
Memory Magister, Dept. electrotechnical. Eng., Univ, Batna, Algeria,
L2 = 0.134 mH, Lm = 0.0045 H, Rr = 0.007 X, Lr = 0.067 mH, J = 2010.
104 kg m2 (turbine + DSIG), f r = 2.5 N m s/rd: (turbine + DSIG ).
Turbine: Radius = 35 m, Number of blades = 3, Hub height =85m ,

978-1-4799-0275-0/13/$31.00 ©2013 IEEE