Energy Conversion and Management 50 (2009) 885–892

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: [Link]/locate/enconman

Nonlinear control with wind estimation of a DFIG variable speed wind turbine
for power capture optimization
B. Boukhezzar a,*, H. Siguerdidjane b,*
a
GREYC Control Team, 6 bd du Maréchal Juin, 14050 Caen, Cedex, France
b
Automatic Control Department, Supélec, Plateau du Moulon, 3, Rue Joliot-Curie, 91192 Gif-sur-Yvette, Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history: A cascaded nonlinear controller is designed for a variable speed wind turbine equipped with a Doubly Fed
Received 27 May 2008 Induction Generator (DFIG). The main objective of the controller is wind energy capture optimization
Accepted 2 January 2009 while avoiding strong transients in the turbine components and specially in the drive train. The inner
Available online 8 February 2009
loop controller ensures an efficient tracking of both generator torque and stator flux, while the outer loop
controller achieves a close tracking of the optimal blade rotor speed to optimize wind energy capture. It is
Keywords: combined to a wind speed estimator that provides an estimation of the wind speed and the aerodynamic
DFIG
torque involved in the controller. The global controller is firstly tested with a simplified mathematical
Nonlinear control
Variable speed wind turbines
model of the aeroturbine and DFIG for a high-turbulence wind speed profile. Secondly, the aeroturbine
Wind speed estimation controller is validated upon a flexible wind turbine simulator. These new control strategies are compared
to other existing controllers based on tests upon an aeroelastic wind turbine simulator. The obtained
results show better performance in comparison with the existing controllers.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction one is devoted to the DFIG. These controllers are designed using
the dynamical features of the wind speed, the aeroturbine and
Wind energy conversion systems have quickly evolved over the the DFIG together with their nonlinear characteristics.
last decades, therefore, an efficient and reliable exploitation tools Many approaches have been proposed for DFIG torque and flux
are necessary to make these installations more profitable [1]. It control [6,7]. They are generally based on simplifying assumptions
was shown that the control strategies have a major effect on the that allow the use of a vector control techniques similar to those
wind turbine and the electric grid loads [2], and whatever the kind employed with an induction machine control [8]. In order to
of the wind turbine, the control strategy remains a key factor [3]. achieve high-performance control during the transient period, a
Modern high-power wind turbines (WT) are equipped with adjust- new DFIG controller is proposed. It consists on using a field ori-
able speed generators [4]. The doubly fed induction generator ented technic without any simplifications in the DFIG model.
(DFIG) with a power converter is a common and efficient configu- For the aeroturbine mechanical part control, the control design
ration to transfer the mechanical energy from the variable speed is generally based on a local linearized model of the WT around its
rotor to a constant frequency electrical grid [5]. Many contribu- operating points [9]. Some nonlinear controllers were proposed
tions have been devoted to the control of the aeroturbine mechan- assuming that the wind turbine operates in steady state conditions
ical as well as the electrical components. The global control [10,11]. The dynamical aspect of the wind and the turbine is then
objective mainly consists in optimizing the extracted aerodynamic not taken into consideration. The second contribution of this work
power in partial load area. then consists then on proposing an aeroturbine controller that is
The contribution of this paper consists on proposing a new con- based directly on the nonlinearity and dynamics of the mathemat-
trol structures even for the DFIG and the mechanical part (aerotur- ical model without any need of wind speed measurement. The con-
bine) that overcomes some of the drawbacks of existing control troller is also able to reject the effect of an additive disturbance on
methods. The global controller is organized in two cascaded con- the control input.
trollers. The first one concerns the aeroturbine, while the second The linearization by feedback is well known in the theory of
nonlinear control systems, but in renewable energy domain, the
classical controllers are mainly implemented only. In our previous
* Corresponding authors. Tel.: +33 1 69 85 19 70; fax: +33 1 69 85 13 89 (B.
works, we have proposed the use of the linearization by feedback
Boukhezzar), tel.: +33 1 69 85 13 77; fax: +33 1 69 85 13 89 (H. Siguerdidjane).
E-mail addresses: [Link]@[Link], b_boukhezzar@ based approach and also the use of LQG controller described in
[Link] (B. Boukhezzar), [Link]@[Link] (H. Siguerdidjane). [12], with no consideration of DFIG.

0196-8904/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/[Link].2009.01.011
886 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892

This paper is organized as follows: the aeroturbine and DFIG where xt is the rotor speed, R is the rotor radius and q is the air
mathematical models are presented in Section 2. Section 3 starts density.
with a description of the global control structure including the If a perfectly rigid low-speed shaft is assumed, a single mass
aeroturbine control and the DFIG control loops. The control objec- model of the turbine may then be considered [13]
tives are then detailed. After a short review of some existing con-
Jt x
_ t ¼ T a  K t xt  T g ð3Þ
trollers, the proposed DFIG controlled is exposed in Section 4.
The designed nonlinear static state feedback linearization with where
asymptotic rotor speed reference tracking and PI action are de-
J t ¼ J r þ n2g J g
scribed in Section 5 in order to reach the required specifications
for aeroturbine control. In Section 6, a brief description of the used K t ¼ K r þ n2g K g
aeroturbine simulator and the experimental wind turbine charac- T g ¼ ng T em
teristics are given. After what, the validation results show quite
good performance of the proposed approach upon the whole wind The one mass wind turbine model is shown in Fig. 1.
energy system.
2.2. DFIG modelling
2. Wind turbine modelling
A scheme of a DFIG-based wind turbine is shown in Fig. 2. This
2.1. Aeroturbine modelling kind of wound-rotor machine can be fed from both stator and rotor
side [14]. The most significant feature of the DFIG is that the stator
The aerodynamic torque expression is given by is directly connected to the grid while the rotor winding is inter-
faced through a back-to-back variable frequency, voltage source
1 converters [4]. By decoupling the power system electrical fre-
Ta ¼ qpR3 C q ðk; bÞv 2 ð1Þ
2 quency and the rotor mechanical frequency, the converter system
The torque coefficient C q depends on the blade pitch angle b and allows a variable speed operation of the wind turbine. As com-
the tip-speed ratio k which is defined as follows: monly done in the literature [15], the DFIG is described in the Park
d–q frame by the well known following set of equations:
xt R
k¼ ð2Þ dUsd
v v sd ¼ Rs  isd þ  xs  Usq ð4Þ
dt
dUsq
v sq ¼ Rs  isq þ þ xs  Usd ð5Þ
dt
dUrd
v rd ¼ Rr  ird þ  xr  Urq ð6Þ
Jt dt
dUrq
ωt v rq ¼ Rr  irq þ þ xr  Urd ð7Þ
Ta Tg dt
As the d and q axis are magnetically decoupled, the flux are gi-
ven by
Usd ¼ Ls  isd þ M  ird ð8Þ
Kt Usq ¼ Ls  isq þ M  irq ð9Þ
Urd ¼ Lr  ird þ M  isd ð10Þ
Fig. 1. One mass wind turbine model dynamics. Urq ¼ Lr  irq þ M  isq ð11Þ

Fig. 2. Configuration scheme of a DFIG-based wind turbine.

 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892 887

for the reader convenience, the list of symbols is given in Appendix The system output to be controlled is the rotor speed xt and the
A. control problem is the tracking of an optimal rotor speed reference
xtopt that ensures maximum wind power capture. Notice that the
3. Control problem formulation blade pitch angle could be used as an additional control input to
achieve electrical power regulation, for high-wind speeds.
3.1. Control structure The main objectives of the DFIG controller are:

The WT electric system time responses are much faster than  Regulate the stator flux to its nominal value Uref .
those of the other parts of the WT. This makes it possible to disso-  Track the generator reference torque T gref that is the aeroturbine
ciate the generator and the aeroturbine control designs and thus control input.
define a cascaded control structure around two control loops:

(1) The inner control loop concerns the electric generator via the 4. DFIG control design
power converters.
(2) The outer control loop concerns the aeroturbine that pro- For the DFIG control, classical techniques as vector control were
vides the reference inputs of the inner loop. extensively used [6,16]. In most cases, the proposed control strat-
egies are derived from those used with an induction machine using
Thereafter, these two control levels will be considered sepa- simplifying assumptions on the DFIG model. For this purpose,
rately. As seen in Fig. 3, the DFIG controller is considered as a many kinds of assumptions can be found in the literature: either
low level one. the rotor and stator transients are neglected [17], either the rotor
We will describe the two level controllers in the following currents are assumed to be previously controlled [18], or the stator
sections. resistance voltage loss [19,20] neglected. A reduced order dynamic
DFIG model is then used for controller design.
3.2. Control objectives In order to improve the transient behavior of the controllers, it
is more convenient to use a complete model of the DFIG. To achieve
The objective of this work is first to design a vector controller of this objective, a new vector control technic based on the DFIG
the DFIG without any reducing assumptions in its mathematical model, without any reduction, is presented as a first contribution
model and second to design a nonlinear controller of the aerotur- in this paper.
bine that takes into consideration the dynamical aspect of the wind To make the stator flux in quadrature with the q axis, the Park
speed and the mechanical part, its nonlinear characteristic, with- frame is oriented such that
out a need of wind speed measurement.
Usd ¼ Uref ð14Þ
The objective of the aeroturbine controller is to optimize wind
power capture. The aerodynamic power captured by the aerotur- Usq ¼ 0 ð15Þ
bine rotor is given by The stator and rotor voltages (4)–(7) are then simplified to [15]
1  
Pa ¼ qpR2 C p ðk; bÞv 3 ð12Þ M 1
2 v sd ¼   ird þ þ s  Usd ð16Þ
Ts Ts
The C p ðk; bÞ curve is specific for each wind turbine. It has a un- M
ique maximum C popt at a single point v sq ¼   irq þ xs  Usd ð17Þ
Ts
C p ðkopt ; bopt Þ ¼ C popt ð13Þ v rd ¼ v 0rd þ ed ð18Þ
In order to maintain k at its optimal value for a given wind v rq ¼ v 0rq þ eq ð19Þ
speed v, the rotor speed must be adjusted using the generator tor-
k
que to track the reference xtopt ¼ opt v. with
R

Fig. 3. Wind turbine control levels.

888 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892

v 0rd ¼ ðRr þ Lr rsÞ  ird ð20Þ Ts

The second term 1þT ss
 v sd can be seen as an output disturbance
v 0rq ¼ ðRr þ Lr rsÞ  irq ð21Þ and compensated using an appropriate regulator that generates
irdref [21].
and Once Usd assigned to Uref , the generator torque will only depend
ed ¼ Lr xr r  irq ð22Þ on irq via v 0rq

M pM Uref 1
eq ¼ Lr xr r  ird þ xr  Usd ð23Þ Tg ¼    v0 ð28Þ
Ls Ls Rr þ Lr rs rq
2
r ¼ 1  LMr Ls is the scattering coefficient, T s the rotor time constant a PI regulator is also sufficient to regulate T g via v 0rq and the decou-
and s the Laplace variable. pling term eq . The global DFIG control scheme is shown in Fig. 4.
Using (16) and (5), we deduce an estimate of the stator flux and
xs frequency 5. Nonlinear static state feedback linearization with PI action
and estimator
^ sd ¼ Ts M
U  v sd þ  ird ð24Þ
1 þ Tss 1 þ Tss A nonlinear static state feedback linearization with asymptotic
v  Rs  isq
^ s ¼ sq rotor speed reference tracking and estimator was proposed in
x ^ sd
ð25Þ
U [22] for wind turbine power capture optimization. In addition to
x
^r ¼ x
^ s  p  xt ð26Þ the aerodynamic torque estimate, the estimator gives also an esti-
mate of the rotor speed x ^t and the effective wind speed v
^ (Fig. 5).
From (20), it may be seen that the d component of the rotor cur- As mentioned in [23], using the following control torque
rent ird can be set to irdref via v 0rd using a PI regulator and the decou-
pling term ed . T g ¼ T a  K t xt  J t a0 e  J t x
_ topt ð29Þ
The irdref is the input of the flux subsystem A first order dynamic response is imposed for the rotor speed
tracking error
^ sd ¼ M Ts
U  irdref þ  v sd ð27Þ
1 þ Tss 1 þ Tss e_ þ a0 e ¼ 0; a0 > 0 ð30Þ

Fig. 4. DFIG control scheme.

Fig. 5. Nonlinear state space feedback linearization with PI controller and wind speed estimator.
 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892 889

where Table 1
Wind turbine characteristics.
^e ¼ x
^ topt  x
^t ð31Þ
Rotor diameter 43.3 m
This technique is known to lack robustness with respect to per- Gearbox ratio 43.165
Hub-height 36.6 m
turbations [24].
Generator system electrical power 600 kW
In order to improve controller performance in presence of addi- Maximum rotor torque 162 kN m
tive disturbance, a nonlinear state feedback with PI controller is
developed next. Even if a nonlinear state feedback with a higher or-
der asymptotic tracking dynamic can reject the effect of input dis- by Germanischer Lloyd WindEnergie to calculate onshore wind
turbance, it will introduce a more complex control law that will turbine loads for design and certification [28].
contain also higher order derivatives [22,23]. The implementation It uses an assumed mode method to model flexible blades and
of the control algorithm can then be difficult due to numerical der- tower components. Other components are modelled as rigid
ivation sensitivity to noise in the measured signal. Another ap- bodies. In this study, three degrees-of-freedom (DOFs) are simu-
proach is proposed herein based on the use of classical linear lated: the variable generator and rotor speed (two DOFs) and the
controllers upon the linearized system using the nonlinear state blade teeter DOF. The variable generator and rotor speed DOFs ac-
feedback. count for the variations in generator speed and the drive train flex-
If we consider the behavior Eq. (3) with the control torque ibility associated with torsional motion between the generator and
" # hub/rotor. The blade teetering DOF accounts for the teeter motion
T^ a Kt induced by asymmetric wind loads across the rotor plane. FAST
T g ¼ Jt  xt  w ð32Þ
Jt Jt subroutines are coupled in an S-Function to be incorporated in a
Simulink model. Hence, FAST is interfaced with Matlab Simulink
this system can be seen as a simple integrator with input w
allowing users to develop and test high-performance controllers.
x_ t ¼ w
6.2. Simulating environment
linear control technique can then be used for optimal rotor speed
tracking.
The full-field turbulent wind set v used in this study is gener-
A PI controller is then used to track the optimal speed x
^ topt de-
ated using SNwind [29] developed by NREL and coupled with FAST.
duced from estimated wind speed v ^.
The hub-height wind speed profile is illustrated in Fig. 6. It has a
 Z 
1 t mean value of 7 m s1 at the hub-height and a turbulence intensity
wðtÞ ¼ K ^eðtÞ þ ^eðsÞds ð33Þ
Ti 0 of 25%.
The main control objective is to maximize the captured energy
where ^e is the estimated tracking error given in (31). from the wind while limiting the transient loads experienced by
The global controller scheme using the estimator and the PI the turbine (see Section 3). Transient loads reduction amounts
controller upon the nonlinear feedback linearization wind turbine mainly to low-speed shaft torsion minimizing, which is equivalent
is shown in Fig. 5. Finally, the control torque T g is given by to reducing the variance of the high-speed shaft torsional moment.
"  Z # The generator torque and the rotor speed must also be both within
T^ a K t 1 t
T g ¼ Jt  xt  K ^eðtÞ þ ^eðsÞds ð34Þ the required constraints of 162 kN m and 58 rpm [30], respectively.
Jt Jt Ti 0
To get more realist operating conditions, an additive measure-
Since the integral action is used to cancel the steady state error ment noise on xt has been introduced. The Signal to Noise Ratio
only, so the integration constant is chosen in such a way that T1i is (SNR) is approximately 7 dB. This is a band-limited, white noise.
far away from the system bandwidth. The adequate value of K is A constant additive disturbance d of 10 kN m is also introduced
49  103 . on the control torque T g . A common phenomenon that can corre-
spond to a constant additive disturbance on the generator torque
is the static friction that can be considered as unknown and con-
6. Validation results
stant additive torque or a bias on the generator torque controller.
Therefore, such kind of disturbance is chosen.
The proposed aeroturbine controller was validated using the
FAST (Fatigue, Aerodynamics, Structures and Turbulence) aeroelas-
6.3. DFIG controller validation
tic simulator (see next section) with the parameters of the Controls
Advanced Research Turbine (CART) which is located at NREL.1 Its
The low level DFIG controller performance are first investigated.
detailed characteristics can be found in [25]. The CART is a variable
As seen in Fig. 7a, the stator flux Usd reachs the constant desired
speed, variable pitch WT with a nominal power rating of 600 kW
reference Uref in less than 10 ms. The DFIG torque T g is then inde-
and a hub height of 36 m. It is a 43-m diameter, two-bladed, tee-
pendently controlled to track the reference torque T gref . The torque
tered hub machine. It is assumed to be coupled to a three-phase
controller achieves good performance while T g and T gref exhibit al-
DFIG. The characteristics are given in [26]. The main parameters
most the same behavior (Fig. 7b).
of CART are summarized in Table 1.
These results proof the ability of the proposed DFIG controller
to achieve high-level performance during the transient period. This
6.1. FAST aeroelastic simulator description
is due to the fact that no assumption is made on the DFIG model
during the controller synthesis as the neglected terms can have a
FAST code is developed by NREL [27]. It is an aeroelastic WT
significant influence during this phase.
simulator capable of modelling two and three bladed propeller-
type machines. This code is used by WT designers to predict both
6.4. Validation using FAST
extreme and fatigue loads. FAST was evaluated and recommended

The proposed nonlinear state feedback-PI with estimator (NSFE-

PI) controller is implemented as well as existing controllers for
1
NREL (National Renewable Energy Laboratory), Golden, CO. comparison using the FAST wind turbine dynamic simulation.
890 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892

test wind profile of 7 m/s

12

11

10

wind speed (m.s−1)

9

3
0 100 200 300 400 500 600
time (s)

Fig. 6. Wind speed profile of 7 m s1 mean value.

Stator d axis flux Electromagnetic Torque

250 100
Φ sd Tg
Φ ref Tg ref
90
200 80
T (kN.m)

70
flux

150
60
g

50
100
40

50 30

20
0 0.005 0.01 0.015 0.02 0.025 20 40 60 80 100 120 140
time (s) time (s)

Fig. 7. DFIG controller performance.

Fig. 8a–d compares the performance obtained using the nonlin- of the wind speed through the aerodynamic torque and rotor speed
ear static state feedback with asymptotic error tracking (NSSFET) estimates even with the noisy measurements of xt and T g . Con-
and the proposed controller. The NSSFET controller is obviously cerning the control torque disturbance, it is only rejected with
unable to reject the unknown disturbance applied to the control the nonlinear state feedback-PI controller with estimator. The NSS-
input (Fig. 8a). This causes the deviation of the rotor speed from FET and NSFE-PI performance are compared with the two baseline
its optimal reference. The deviation is more visible in between controllers: the Indirect Speed Controller (ISC) [13] and the Aero-
250 and 400 s. In spite of the presence of the disturbance, the non- dynamic Torque Feed Forward (ATF) control strategy [13].
linear state feedback-PI controller with estimator succeeded in The validation tests have been performed by using the wind
keeping the rotor speed around of the optimal reference speed turbine simulator FAST with the same wind speed profile under
(Fig. 8a). One may observe in Fig. 8b that the control torque used identical operating conditions concerning disturbance, and mea-
in the NSFE-PI controller to reject the additive disturbance is more surement noise.
important than that developed in the NSSFET. However, it still re- The obtained performance with the different controllers are
mains below the upper bound, though. The disturbance rejection shown in Fig. 9 and summarized in Table 2. The selected compar-
results in an increase of the captured energy for the NSFE-PI con- ison criteria are the power capture efficiency and the low-speed
troller compared to the nonlinear static state feedback one shaft torque torsional standard deviation. One can observe in
(Fig. 8c). As seen in Fig. 8a, the rotor speed tracks mean tendency Fig. 9a that the produced electric power using the NSFE-PI is a little
of the optimal rotational speed. That causes its variations to remain bit more significant. The low-speed shaft torque oscillations are
smooth. The differences between the optimal rotor speed reference also shown to be reduced using this control strategy, as depicted
and rotor speed occur during the start-up transient. The generator in Fig. 9b. On the one hand, one may notice that the performance
torque from the FAST model (Fig. 8b) is greater than that obtained of the three first controllers remain close. On the other hand, one
with the mathematical model while remaining below the maxi- can note that the NSFE-PI controller with estimator allows an
mum acceptable value. Inversely, the electrical power produced improvement of 18% of the efficiency compared to the indirect
is less than the one given by the simplified mathematical model speed controller with a quietly similar standard deviation of the
(Fig. 8c). This decrease can be interpreted by the highest complex- low-speed shaft torsional torque T ls .
ity of the wind turbine simulator that includes more flexible ele- Both of the first three controllers are unable to reject the pres-
ments and consequently induces more power losses than the ence of the control input disturbance. As the ISC controller synthe-
simplified mathematical model. tized in steady state regime, it does not take into account the high-
The efficiency of the Kalman filter used with the Newton algo- turbulence of the wind speed which leads to a decrease in power
rithm estimator can be seen in Fig. 8d. It provides a good estimate capture efficiency.
 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892 891

rotor speed generator torque

45 ωopt
100
Tg NSSFE−PI
ω NSSFE Tg NSSFE
40 ω NSSFE−PI 90
80
35
rotor speed (rpm)

70

Tg (kN.m)
30
60
25 50

20 40
30
15
20
10
10
5 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
time (s) time (s)

wind turbine power real and estimated wind speed

350 12
Pe NSSFE−PI estimated wind
Pe NSSFE 11 real wind
300

wind speed (m.s−1)

10
250
9
Pe (kW)

200 8

150 7

6
100
5
50
4
0 3
0 100 200 300 400 500 600 0 100 200 300 400 500 600
time (s) time (s)

Fig. 8. Nonlinear state feedback with asymptotic error tracking and PI action controllers, with estimator.

wind turbine power (KW) low speed shaft torsional torque

300 90
NSSFE−PI NSSFE−PI
NSSFE NSSFE
ATF 80
ATF
250 ISC ISC
70

200 60
Tls(KN.m)
Pe (KW)

50
150
40

100 30

20
50
10

0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
time (s) time (s)

Fig. 9. Comparison of the different control strategies.

7. Conclusion
Table 2
Comparison of the different control strategies.
The presented research work deals with variable speed wind
Controller ISC NSSFET ATF NSFE-PI
control design, in order to achieve the objectives of maximizing
Efficiency (%) 51.79 61.23 63.43 70.00 the extracted energy from the wind, below the rated power area.
stdðT hs Þ (kN m) 12.73 11.87 13.19 12.64
Cascaded two controllers have been suggested for both DFIG and
maxðT g Þ (kN m) 112.35 70.70 76.99 81.09
aeroturbine control.
892 B. Boukhezzar, H. Siguerdidjane / Energy Conversion and Management 50 (2009) 885–892

The proposed inner control loop concern the DFIG control. It [3] Burton T, Sharpe D, Jenkins N, Bossanyi E. Wind energy handbook. John Wiley
& Sons; 2001.
achieves a good tracking of rotor reference flux and torque even
[4] Muller S, Deicke M, De Doncker RW. Doubly fed induction generator systems
during the transient phase. The uses of a nonlinear state feed- for wind turbines. IEEE Ind Appl Mag 2002:26–33.
back-PI controller jointly with an estimator for the aeroturbine [5] Lindholm M. Doubly fed drives for variable speed wind turbines. Ph.D. thesis,
outer control loop allow to take into consideration the dynamic as- Technical University of Denmark; 2004.
[6] Cardenas RJ, Pena RS, Asher J, Asher GM, Clare JC. Sensorless control of a
pect of the wind turbine, of its aerodynamic behavior nonlinearity doubly-fed induction generator for stand alone operation. In: IEEE 35th annual
and of the turbulent nature of the wind. This controller allows the power electronics specialists conference, vol. 5; 2004. p. 3378–83.
rejection of the effect of the disturbance and achieved the goal. [7] Pena RS, Cardenas RJ, Clare JC, Asher GM. Control strategy of doubly fed
induction generators for a wind diesel energy system. In: IEEE 2002 28th
The designed nonlinear state feedback-PI controller with esti- annual conference of the industrial electronics society, vol. 4; 2002. p. 3297–
mator ensures the best performance in terms of efficiency with 302.
an acceptable drive train transient loads and better implementa- [8] Poitiers F, Machmoum M, Doeuff RL. Simulation of wind energy conversion
system based on a doubly-fed induction generator. In: 10th European
tion simplicity compared to some existing controller strategies. conference on power electronics and applications, CD-ROM proceedings,
France, Toulouse; 2003.
[9] Ma X. Adaptive extremum control and wind turbine control. Ph.D. thesis,
Appendix A. List of symbols
Denmark, May; 1997.
[10] Song YD, Dhinakaran B, Bao XY. Variable speed control of wind turbines using
nonlinear and adaptive algorithms. J Wind Eng Ind Aerodynam
v wind speed, m s1 2000;85(3):293–308.
q air density, kg m3 [11] Valenciaga F, Puleston PF. Variable structure control of a wind energy
conversion system based on a brushless doubly fed reluctance generator.
R rotor radius, m
IEEE Trans Energy Convers 2007;22(2):499–506.
Pa aerodynamic power, W [12] Boukhezzar B, Lupu L, Siguerdidjane H, Hand M. Multivariable control strategy
Ta aerodynamic torque, N m for variable speed variable pitch wind turbines. Renew Energy
k tip-speed ratio 2007;32(8):1273–87.
[13] Boukhezzar B. Sur les stratégies de commande pour l’optimisation et la
b pitch angle, deg régulation de puissance des eoliennes à vitesse variable, Ph.D. thesis,
Cp power coefficient Université de Paris Sud-Supélec; February 2006.
Cq torque coefficient [14] Tapia A, Tapia G, Ostolaza X, Sáenz JR. Modelling and control of a wind turbine
driven doubly fed induction generator. IEEE Trans Energy Convers
xt aeroturbine rotor speed, rad s1 2003;18(2):194–204.
T em DFIG torque, N m [15] Aimani SE. Modélisation de différentes technologies d’éoliennes intégrées
Tg DFIG torque in the rotor side, N m dans un réseau moyenne tension. Ph.D. thesis, Ecole Centrale de Lille-
Université des Sciences et Technologies de Lille 1; December 2004.
T ls low-speed shaft torque, N m [16] Hopfensberger B, Atkinson DJ, Lakin RA. Stator flux oriented control of a
T hs high-speed shaft torque, N m cascaded doubly-fed induction machine. IEE Proc – Electric Power Appl
Jr rotor inertia, kg m2 1999:597–605.
[17] Slootweg JG, Polinder H, Kling WL. Dynamic modelling of a wind turbine with
Jg DFIG inertia, kg m2 doubly fed induction generator. In: IEEE power engineering society summer
Jt turbine total inertia, kg m2 meeting; 2001. p. 644–9.
1
Kr rotor external damping, N m rad s [18] Forchetti D, Garcia G, Valla MI. Vector control strategy for a doubly-fed stand-
1 alone induction generator. In: IEEE 2002 28th annual conference of the
Kg DFIG external damping, N m rad s
1 industrial electronics society; 2002. p. 991–5.
Kt total external damping, N m rad s [19] Pena R, Clare JC, Asher GM. Doubly fed induction generator using back-to-back
1
K ls low-speed shaft damping, N m rad s PWM converter and its application to variable-speed wind-energy generation.
ng gearbox ratio IEE Proc B 1996;143(3):231–41.
[20] Poitiers F, Machmoum M, Doeuff RL, Zaim M. Control of a doubly-fed induction
Pe electrical power, W generator for wind energy conversion systems. In: Proceedings of the
v sd;q stator d—q voltage, V Australasian Universities power engineering conference; 2001.
v rd;q rotor d—q frame voltage, V [21] Goodwin GC, Graebe SF, Salgado ME. Control system design. Prentice Hall;
2000.
isd;q stator d—q frame current, A [22] Boukhezzar B, Siguerdidjane H, Hand M. Nonlinear control of variable-speed
ird;q rotor d—q frame current, A wind turbines for generator torque limiting and power optimization. ASME J
Urd;q rotor d—q frame flux, Wb Solar Energy Eng 2006;128(4):516–30.
[23] Boukhezzar B, Siguerdidjane H. Nonlinear control of variable speed wind
Urd;q rotor d—q frame flux, Wb turbines without wind speed measurement. In: ECC-CDC 2005, Seville, Spain;
Rs stator resistance, X 2005.
Rr rotor resistance, X [24] Nijmeijer H, Schaft AVD. Nonlinear dynamical control systems. Springer-
Verlag; 1996.
Ls stator inductance, H [25] Stol KA. Geometry and structural properties for the controls advanced
Lr rotor inductance, H researchturbine CART from model tuning. Subcontractor report SR-500-
M mutual inductance, H 32087, National Renewable Energy Laboratory, Golden, Colorado; September
2004.
r scattering coefficient
[26] Poitiers F, Machmoum M, Doeuff RL, Zaim M. Direct torque control of a doubly-
p number of poles fed induction generator for variable speed wind turbine power regulation. In:
xs stator d—q reference axes speed, rad s1 2007 European wind energy conference proceedings, EWEA, Milan, Italy; 2007.
xr rotor d—q reference axes speed, rad s1 [27] Jonkman JM, Buhl ML. FAST user’s guide. 6th ed. National Wind Technology
Center, National Renewable Energy Laboratory, Golden, Colorado.
[28] Manjock A. Design codes fast and ADAMSâ for load calculations of onshore
wind turbines. NREL report 72042, Germanischer Lloyd WindEnergie GmbH,
Hamburg, Germany.
[29] Buhl ML. SNwind user’s guide. 1st ed. National Wind Technology Center,
References National Renewable Energy Laboratory, Golden, Colorado; June 2003.
[30] Fingersh LJ, Johnson K. Controls advanced research turbine (CART)
[1] Manwell JF, McGowan J, Rogers A. Wind energy explained: theory design and commissioning and baseline data collection. NREL report TP-500-32879,
applications. John Wiley & Sons; 2002. National Renewable Energy Laboratory, Golden, Colorado; October 2002.
[2] Bianchi FD, Battista HD, Mantz RJ. Wind turbine control systems: principles
modelling and gain scheduling design. 2nd ed. Springer; 2006.