Professional Documents
Culture Documents
Nonlinear Control of Variable Speed Wind Turbines
Nonlinear Control of Variable Speed Wind Turbines
Nonlinear Control of Variable Speed Wind Turbines
net/publication/224627312
CITATIONS READS
103 977
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Boubekeur Boukhezzar on 09 September 2016.
Abstract— A nonlinear approach, for variable speed wind wind speed measurement.
turbines control, is presented by considering that there is no This paper is organized as follows : Section II describes the
wind speed measurement. The control objective below rated wind turbine modelling, the control objectives are exposed.
wind speed is to maximize the extracted energy from the
wind while reducing mechanical loads. The existing control Section III briefly recalls some existing control techniques,
techniques do not take into account the dynamical aspect of the aerodynamic torque and wind speed estimators are
the wind and the turbine, which leads to significant power described. A nonlinear static and dynamic state feedback
losses, besides, they are not robust. In order to bring some linearization with asymptotic rotor speed reference tracking
improvement and to evaluate the applicability efficiency of the are developed in order to reach the required specifications : In
nonlinear controllers, so then, nonlinear static and dynamic
state feedback controllers with wind speed estimator are de- section IV, simulation results show quite good performance
rived. The controllers are tested upon a mathematical model of the proposed approach on the basis of the mathematical
and are validated with a wind turbine simulator, in presence model and have been validated through a wind turbine
of disturbances and measurement noise. The results indeed simulator.
show significant improvements in comparison with the existing
controllers.
II. W IND TURBINE MODELLING
I. INTRODUCTION A. system modelling
Advances in wind turbine technology [1] made necessary
The power extraction of wind turbine is being known to
the design of more powerful control systems. This is in order
be a function of three main parameters : the wind power
to improve wind turbines behavior, namely to make them
available, the power curve of the machine and the ability
more profitable and more reliable.
of the machine to respond to wind fluctuations [10]. The
Compared to fixed speed turbines, variable speed wind tur-
global scheme of a variable speed wind turbine is displayed
bines feature higher energy yields, lower component stress,
in Fig. 1. Roughly speaking, a variable speed wind turbine
and fewer grid connection power peaks. To be fully exploi-
(FPVS), mainly consists of an aeroturbine, a gearbox and a
ted, variable speed should be controlled in meaningful way.
generator.
Many of the works in the wind energy conversion systems
The expression for aerodynamic power captured by the wind
control deal with the optimization of the extracted aerody-
turbine is given by the nonlinear expression
namic power in partial load area. For this purpose, classical
controllers have been extensively used, particularly the PI 1
regulator [2], [3]. Optimal control has been applied in the Pa = ρπ R2C p (λ , β )v3 (1)
2
LQ [4], and LQG form [5], [6].
where
Robust control of wind energy conversion systems (WECS) ωr R
has been introduced in [7] and also used in [8] - [9]. λ=
v
By assuming the wind turbine operating in steady state condi-
tions, most of these papers do not take into consideration the is the so-called tip speed ratio, namely the ratio between the
dynamical aspect of the wind and the turbine, because wind linear blade tip speed and the wind speed v. R is the rotor
energy conversion systems have strong nonlinear characteris- radius, β is the blades pitch angle.
tics. Furthermore, in the major part of these investigations, For the reader convenience, table of the symbols description
the wind speed is assumed to be measurable. In spite of using is given at the end of this paper.
an anemometer to measure the wind speed, the obtained So, any change in the rotor speed or the wind speed induces
value is, as matter of fact, different from the one that appears change in the tip-speed ratio, thus leading to the power
in the wind turbine model equations, since its a mean value. coefficient C p (λ , β ) variation and therefore to the generated
The objective of this work is to design robust nonlinear power one. The C p (λ , β ) curve is displayed in Fig. 3.
controllers that take into consideration the dynamical aspect The aerodynamic power is
of the wind speed and the wind turbine, without a need of
Pa = ωr Ta (2)
This work has been carried out within the project A
Cnergie launched by
Supélec. where the aerodynamic torque Ta expression is
The authors are with Automatic Control Department, Supélec,
91192-F, Gif-sur-Yvette, cedex boubekeur.boukhezzar, 1
Houria.siguerdidjane@supelec.fr Ta = ρπ R3Cq (λ , β )v2 (3)
2
0-7803-9568-9/05/$20.00 ©2005 IEEE 3456
Ta Tg
Wind v Rotor Drive
Pg
speed aerodynamics trains Generator 0.45
ωr ωg 0.4
0.35
0.25
C (λ,β)
p
0.2
ωr 0.15
0.1
0.05
ωg 0
−10
0 12 14
10 8 10
Bls
20 4 6
30 0 2
pitch angle β (deg)
tip speed ratio λ
Tem
Fig. 3. Power coefficient curve
K ls
where
Kr Jt = Jr + n2g Jg
Kt = Kr + n2g Kg
Fig. 2. Drive train dynamics Tg = ng Tem
3457
Estimator
generator torque. The choice of Tem as a control input is
motivated by the fact that when controlling the firing angle Newton-Raphson
estimated
state
Kalman
of the converter, it is possible to control the electrical torque algorithm Filter
in the generator. Well known control strategies have been wind speed
estimated
used and achieve a fast and decoupled responses of torque s t a t e
and flux like vector control [11], [12] and Direct Torque wopt
d y n a m i c
Control (DTC) [13]. The torque control using the frequency state feedback integrator wi nd
linearization turbine wr
converter allows the wind turbine to run at variable speed wr disturbance
3458
TABLE I
B. wind speed estimation
W IND TURBINE CHARACTERISTICS
The effective wind speed v is related to Ta by the aerody-
namic torque expression Rotor diameter 43.3 m
1 gearbox ratio 43.165
Ta = ρπ R3Cq (λ )v2 (18) Hub height 36.6 m
2 Generator system electrical power 650 KW
where Maximum rotor torque 162 KN.m
Cq (λ ) = Cq (λ , βopt )
Cq (λ ) is defined using a look-up table. Nevertheless, to solve 12
test wind profile of 7 m/s
a polynomial in λ . 9
Cq (λ ) = ∑ ai λ i 7
i=0 6
1 ω̂r R 2 3
0 100 200 300 400 500 600
Ta − ρπ R Cq
ˆ 3
v̂ = 0 (19) time(s)
2 v̂
Fig. 5. Wind speed profile of 7 m.s−1 mean value
The Newton-Raphson algorithm is then used to solve equa-
tion (19). As this equation has a unique solution, that is
contained in the below rated wind speed range, the conver-
of this dynamics must lead to track the average wind speed
gence of the algorithm is obtained after a few iterations only.
above small time intervals. The scheme of the dynamic
v̂ is used to obtain ω̂opt
state feedback controller, without wind speed measurement
λopt v̂ is given in Fig. 4. All the derivatives that appear in the
ω̂opt = (20)
R generator torque expression (23) and (25) are obtained using
C. nonlinear static state feedback control with estimator an approximated filtered derivatives.
From the estimate of the optimal rotor speed and the aero-
IV. VALIDATION RESULTS
dynamic torque, a static state feedback linearization control
technique with asymptotic rotor speed reference tracking is As already mentioned, the wind turbine considered in this
used. A chosen dynamics is then imposed to the rotor speed study is variable speed, variable pitch one. It consists of two
tracking error ε̂ . blades rotor coupled with a gearbox. The high speed shaft
ε̂˙ + a0 ε̂ = 0 (21) drives an induction generator connected to the grid via a
power electronics device. The numerical simulations have
where been performed using a wind turbine whose characteristics
ε̂ = ω̂opt − ω̂r (22) are given in TABLE I. The validation tests have been
from equations (9) and (21), the obtained control torque is performed using a wind turbine simulator.1 . The applied
input wind speed profile of 7 m.s−1 mean value is shown
Tg = T̂a − Kt ω̂r − Jt a0 ε̂ − Jt ω̂˙ opt (23) in Fig. 5
D. nonlinear dynamic state feedback control with estimator
A. Using the model
The static state feedback linearization control technique is
not so robust with respect to perturbations [18]. First, the mathematical model is used with the proposed
Even if the system can be linearized by a static state controllers. A measurement noise on ωr as well as a dis-
feedback, one can impose a higher order dynamics to the turbance of 10 KN.m on the control torque Tg have been
tracking error for getting a dynamic state feedback. Let introduced. The simulation results show that the required
us chose a second order dynamics, the tracking error ε̂ is performance are reached using the nonlinear dynamic state
governed by the following differential equation feedback controller with estimator (NDSFE). In Fig. 6(a),
one can check that the rotor speed ωr tracks the mean ten-
ε̂¨ + b1 ε̂˙ + b0 ε̂ = 0 (24) dency of the optimal rotational speed ωopt without tracking
the turbulent component. This led to acceptable control loads,
From (9) and (24), it comes out
as depicted in Fig. 6(b), that remains below the maximum
Ṫg = T̂˙a − Kt ω̂˙ r − Jt ω̂¨ opt − Jt b1 ε̂˙ − Jt b0 ε̂ (25) value of Tg (162 KN.m).
A very fast dynamics will lead to the tracking of the wind 1 developed by NREL (National Renewable Energy Laboratory),
speed turbulence and also to large control loads. The choice Golden, CO.
3459
rotor speed generator torque
45 wind turbine power (KW)
90 400
ω
model Pe
ωopt model
40 80 350
300
35 70
rotor speed (rpm)
250
30
Tg(KN.m)
60
P (Kw)
200
e
25 50
150
20 40
100
15 30
50
10 20 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600 0 100 200 300 400 500 600
time(s) time(s) time(s)
Fig. 6. Tests of the nonlinear dynamic state feedback controller, with estimator, upon the mathematical model
40 70
250
35 60
200
30
Tg(KN.m)
50
Pe (KW)
ω (rpm)
150
25 40
100
20 30
50
15 20
10 10 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600 0 100 200 300 400 500 600
time (s) time (s) time (s)
Fig. 7. Validation of the nonlinear dynamic state feedback controller, with estimator, by using the wind turbine simulator
B. Validation by using the wind turbine simulator bit more significant. The low speed shaft torque oscillations
The validation of the controllers has been carried out upon are also shown to be reduced using the NDSFE strategy, as
the wind turbine simulator. The obtained performance are depicted in Fig. 8(b). One notice that the performance of
shown to be as expected using the dynamic state feedback the three first controllers remain close. On the other hand,
controller with estimator. As Fig. 7(a) shows, the rotor speed one can note that the dynamic state feedback controller with
variations remain smooth, without big fluctuations, while estimator allows an improvement of 10% of the efficiency
tracking the mean tendency of the optimal rotational speed. compared to the indirect speed controller with a decrease of
That ensures an optimal power extraction from the wind, so 1.18 KN.m on the standard deviation of the low speed shaft
the oscillations will be reduced in the drive train components. torsional torque Tls .
The generator torque Tg with the simulator (Fig. 7(b)) ,
remains below the maximum acceptable value, with smooth
variations. That induces low frequency variations in the V. CONCLUSION
generator currents, in other words, a better preservation of the
electrical wind turbine devices. The Nonlinear Dynamic State
Feedback Linearization, with estimator (NDSFE) controller Nonlinear feedback controllers, with wind speed estimator
ensures the rejection of the disturbance on the control torque. have been developed for wind turbine control. The objectives
In the same way, the Kalman filter used with the Newton- are to synthesize robust controllers that maximize the energy
Raphson algorithm provides a good estimate of the wind extracted from the wind while reducing mechanical loads.
speed, the aerodynamic torque as well as the rotor speed It has been proposed, herein, controllers based on dynamic
from a noisy measurement of ωr . state feedback linearization technique, with asymptotic rotor
In order to evaluate the contribution level of the propo- speed reference tracking, combined with an aerodynamic
sed controller, a comparison is made between the different torque and wind speed estimator. The dynamic aspects of
controllers. The simulations have been carried out using the the wind and the turbine were taken into account leading to
wind turbine simulator, under the same operating conditions good performance. The control strategy has been validated
on disturbance and measurement noise. The obtained per- with an aeroelastic wind turbine simulator and has led to
formance with the different controllers are shown in Fig. 8 satisfactory results, which indeed points out the efficiency
and summarized in TABLE II. One can observe in Fig 8(a) level of the developed controllers in comparison with those
that the produced electric power using the NDSFE is a little presented in the existing literature.
3460
wind turbine power (KW) low speed shaft torsional torque
300 90
NDSFE NDSFE
NSSFE NSSFE
ATF 80 ATF
ISC ISC
250
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. 8. Electrical power and low speed shaft torque using the different control strategies
TABLE II
R EFERENCES
C OMPARISON OF THE DIFFERENT CONTROL STRATEGIES
[1] L. V. Divone, Evolution of modern wind turbines, pp. 73–138. David
A. Sepra, amse press ed., 1995.
Ths standard [2] X. Ma, Adaptive extremum control and wind turbine control. PhD
Efficiency [%]
deviation [KN.m] thesis, Technical University of Denmark, 1997.
ISC 59.01 12.77
[3] M. M. Hand and M. J. Balas, “Non-linear and linear model based
NSSFE 61.28 11.87 controller design for variable-speed wind turbines,” pp. 1–8. NREL
ATF 63.49 13.19 Report No. CP-500-26244.
NDSFE 70.01 11.59 [4] T. Ekelund, Modeling and Linear Quadratic Optimal Control of Wind
Turbines. PhD thesis, Chalmers University of Technology, Sweden,
April 1997.
[5] W. E. Leithead, S. DeLaSalle, and D. Reardon, “Role and objectives of
N OTATION AND S YMBOLS control of wind turbines,” IEE Proceedings Part C, vol. 138, pp. 135–
148, 1991.
[6] P. Novak, T. Ekelund, I. Jovilk, and B. Schmidtbauer, “Modeling and
control of variable speed wind turbine drive systems dynamics,” IEEE
v wind speed (m · s−1 ). Control Systems Magazine, vol. 15, no. 4, pp. 28–38, 1995.
ρ air density (kg · m−3 ). [7] M. Steinbuch, Dynamic modelling and robust control of a wind
energy conversion system. PhD thesis, Delft University of Technology,
R rotor radius (m). Netherlands, 1989.
Pa aerodynamic power (W). [8] P. Bongers, Modeling and identification of flexible wind turbines
Ta aerodynamic torque (N · m). and a factorizational approach to robust control. PhD thesis, Delft
University of Technology, Netherlands, 1994.
λ tip speed ratio. [9] D. Connor, S. N. Iyer, W. E. Leithead, and M. J. Grimble, “Control
β pitch angle (deg). of horizontal axis wind turbine using H∞ control,” in First IEEE
Cp (λ , β ) power coefficient. Conference on Control Applications, 1992.
Cq (λ , β ) torque coefficient. [10] Y. D. Song, B. Dhinakaran, and X. Y. Bao, “Variable speed control
of wind turbines using nonlinear and adaptive algorithms,” Journal of
ωr rotor speed (rad · s−1 ). Wind Engineering and Industrial Aerodynamics, vol. 85, pp. 293–308,
ωg generator speed (rad · s−1 ). 2000.
Tem generator (electromagnetic) torque (N · m). [11] A. Tapia, G. Tapia, J. X. Ostolaza, and J. R. Saenz, “Modeling and
control of a wind turbine driven doubly fed induction generator,” IEEE
Tg generator torque in the rotor side (N · m). Transactions on energy conversion, vol. 18, no. 2, pp. 194–204, 2003.
Tls low speed shaft (N · m). [12] R. Pena, J. C. Clare, and G. M. Asher, “A doubly fed induction
Ths high speed shaft (N · m). generato using back-to-back pwm conveters supplying an isolated load
from a variable speed wind turbine,” IEE Proc.-Electr. Power Appl,
Jr rotor inertia (kg · m2 ). vol. 143, no. 5, pp. 380–387, 1996.
Jg generator inertia (kg · m2 ). [13] S. Arnalte, J. C. Burgos, and J. L. Rodriguez-Amendeo, “Direct torque
Jt turbine total inertia (kg · m2 ). control of a doubly-fed induction generator for variable speed wind
Kr rotor external damping (Nm · rad−1 · s−1 ). turbines,” Electric Power Components and Systems, vol. 30, no. 2,
pp. 199–216, 2002.
Kg generator external damping (Nm · rad−1 · s−1 ). [14] H. Vihriälä, R. Perälä, P. Mäkilä, and L. Söderlund, “A gearless wind
Kt turbine total external damping (Nm · rad−1 · s−1 ). power drive : Part 2 : Performance of control system,” in European
Kls low speed shaft damping (Nm · rad−1 · s−1 ). Wind Energy Conference, vol. 1, pp. 1090–1093, 2001.
[15] H. Vihriälä, Control of Variable Speed Wind Turbines. PhD thesis,
Bls low speed shaft stiffness (Nm · rad−1 ). Tampere University of technology, 2002.
[16] W. E. Leithead and D. Connor, “Control of variable speed wind
ACKNOWLEDGMENT turbines : Design task,” International Journal of Control, vol. 73,
no. 13, pp. 1173–1188, 2000.
The authors would like to thank M. Hand and E. Muljadi [17] E. L. van der Hooft, P. Schaak, and T. G. van Engelen, “Wind
from NREL (Golden, CO.) for their help in providing turbine control algorithms : Dowec-f1w1-eh-0303-094/0,” report, ECN
the simulator FAST, thus giving us the opportunity to validate Windenergie, December 2003.
[18] A. Isidori, Nonlinear Control Systems. Springer-Verlag, 3 ed., 1995.
our theoretical results.
3461