You are on page 1of 8

Proceedings of the ASME 2010 Dynamic Systems and Control Conference DSCC2010 September 12-15, 2010, Cambridge, Massachusetts,

USA

DSCC2010WIND TURBINE POWER CAPTURE CONTROL WITH ROBUST ESTIMATION


Tony Hawkins Mechanical and Nuclear Engineering Kansas State University Manhattan, KS 66506 gah8833@ksu.edu Guoqiang Hu Mechanical and Nuclear Engineering Kansas State University Manhattan, KS 66506 gqhu@ksu.edu Warren N. White Mechanical and Nuclear Engineering Kansas State University Manhattan, KS 66506 wnw@ksu.edu Faryad Darabi Sahneh Mechanical and Nuclear Engineering Kansas State University Manhattan, KS 66506 faryad@ksu.edu

ABSTRACT A robust control scheme to optimize the power capture of a wind turbine is proposed. A novel identification technique is used both to estimate the unknown aerodynamic properties and to regulate the turbine speed. A Lyapunov-based approach is adopted to choose the set point change direction so that power capture can be maximized for the given wind speed and environmental parameters. The set point consists of the turbine tip speed ratio and the turbine blade pitch. The efficacy of the controller is demonstrated through simulation. NOMENCLATURE A Rotor Swept Area m2 CD Damping Coefficient - kgm2/s Rotor Power Coefficient dimensionless Cp * Optimal Power Coefficient dimensionless Cp J Rotor Inertia kgm2 N Controller Auxiliary Term Power Captured by Rotor - Nm/s Paero Power available from wind - Nm/s Pavail R Swept Area Radius m e Controller Error Signal rad/sec Estimation of Aerodynamic Torque - Nm Control Gain Wind Velocity m/s Wind Velocity Measurement Noise m/s

d * c d aero c * d

Control Gain Blade Pitch angle degrees Desired Blade Pitch degrees Optimal Blade Pitch degrees Control Gain Blade Pitch Adaptation Constant dimensionless Tip Speed Adaptation Constant dimensionless Tip-Speed Ratio dimensionless Desired Tip-Speed Ratio dimensionless Air density kg/m3 Aerodynamic Torque - Nm Generator (control) torque - Nm Angular Shaft Velocity rad/s Optimal Angular Velocity rad/s Desired Angular Velocity rad/s

ks v w

I. INTRODUCTION Wind turbines are machines that provide the energy industry with a clean alternative to burning fossil fuels in the generation of electric power. The implementation of wind turbines has increased drastically throughout the past decade. This has driven researchers to examine how the power is being captured and how that energy capture efficiency can be improved. Wind turbine control has proven to be challenging because the power provided to the wind capture system has a

Copyright 2010 by ASME

difficult to measure nonlinear aerodynamic property, which is a function of the tip-speed ratio and propeller blade pitch. To extend the challenging nature of the problem, the optimal operating point is not always common between similar machines operating under the same conditions. The optimization of energy capture efficiency in a wind turbine becomes not only a problem of nonlinear system trajectory tracking, but also estimation and extremum seeking. It is probable that long term, time varying conditions could come into play to alter this optimal operating point. Changes such as component aging, oxidation, icing, and frictional modification can attribute to some of these time varying parameters in the physical system. Controllers that will operate the turbine at peak power production under time varying conditions are desirable. Increasing the power capture in wind turbines provides stakeholders with a greater return on investment and the industry with more efficient machines which require less controller maintenance and tuning. With this motivation, control systems in wind turbines have advanced through the years. Using blade pitch control, generator torque control, or a combination of both, control systems have been built into wind power capture systems to stabilize their operation and maximize power capture. Linear control methods have been used to regulate wind turbines [1] modeled as linear systems. Such approaches use PID or similar controllers. State feedback gains are then tuned to each wind turbine system. These control methods assume prior knowledge of the optimal operating point, as well as manual re-tuning in the event of any parameter variation. Different implementations of sliding mode control have also been used as a nonlinear control approach. In [2] and [3], sliding mode controllers were proposed for a system in order to provide tracking in the presence of the unknown nonlinear input torque. Sliding mode results proved better in the tracking control, however, the approach still needed a reference track to be provided. Perhaps one of the greatest challenges that wind turbine control presents is the lack of available information. In fact, the extent of power capture which is intended to be maximized is a quantity that is not readily measured. This has caused researchers to develop adaptive techniques to aide in identifying unknown quantities. Johnson et al. proposed a nonlinear controller, [4] and [5], where an adaptive gain is tuned based on an estimation of the power captured over a period of time. This control scheme was applied to a real turbine and was able to achieve convergence to the region of maximum power capture. This approach had a lengthy adaptation period, on the order of hours, and a convergence time on the order of days. Blade pitch also was not considered in this application making this a single dimensional optimization problem. Other estimation approaches have also been studied such as Mas wind turbine controller given in [6]. The Kalman filter is used to estimate the unknown nonlinearities while a proportional-integral controller regulates the turbine to a

desired set point. In [6], it is assumed that the optimal point is known. Mas paper also excludes blade pitch control. Creaby et al. [7] investigated the power maximization based on the extremum seeking control strategy proposed by Ariyur and Kristic [8] for systems with general input to output mapping. While achieving good convergence to the optimal operating point, this method relies on oscillatory perturbation of the optimizing variables. For the specific application to large wind power capture systems, this may introduce unwanted mechanical vibration and the associated fatigue, thus exacerbating the already high wind turbine maintenance needs. Also, it may not be practical to perturb large inertial systems with high frequency disturbances due to lengthy system time constants. Other nonlinear robust strategies have been developed to estimate the nonlinearities. Iyasere et al. [9, 12] use a robust controller to provide an estimation of the unknown aerodynamic torque while employing Powells optimization method to provide a reference trajectory to follow. The application of this approach was for a small scale wind turbine and showed good convergence toward the optimal values of blade pitch and tip-speed ratio but did not consider operation under a realistic wind condition or the presence of measurement error. The control approach presented in the current work uses a Lyapunov - based strategy for maximizing the power capture given time varying wind conditions. This approach will control the blade pitch and shaft speed of a large-scale wind turbine (greater than 1.0 MW) so that the turbine is operating very near the peak power capture. The controller has two responsibilities. The first is the inner loop which provides regulation of the turbine shaft angular velocity to a given set point together with the estimation of the wind produced aerodynamic torque. The second is the outer loop which provides the regulator with a setpoint trajectory for the tip speed ratio and blade pitch for improving the rotor power capture. System parameter changes and time-varying wind velocity are present in practical wind power capture systems. The Lyapunov-based method will ensure the convergence of the captured power to a region near its optimal value in the presence of a time varying wind condition and measurement error. The presentation will consist of a description of the system dynamics of a typical horizontal axis wind turbine, the Lyapunov based methodology for trajectory generation for the desired blade pitch and tip-speed ratio, and the use of a robust controller to regulate the system to the desired set-point as well as to provide the Lyapunov based controller with an estimation of the unknown system quantities. These unknown quantities consist of the rotor power coefficient and the gradient of the power coefficient with respect to the tip speed ratio and blade pitch. Finally the system and controllers are modeled and a simulation is performed. The results of the simulation will be analyzed and discussed.

Copyright 2010 by ASME

II. DYNAMIC MODEL DEVELOPMENT The wind turbine model considered in this paper consists of a lumped inertia, a torque provided by the wind acting on the rotor, a lumped damping coefficient, and a control torque which will be provided by the loading of the electric power generator. The total rotational inertia of this system can be lumped into one single known term which is represented by J. The total viscous damping coefficient is combined into CD, a known constant value. The torque introduced to the system by the wind - rotor interaction is denoted as . The wind turbine system dynamic model is described by the first order differential equation = ( ).
1

(1)

where is the shaft angular velocity and is the state variable of the inner loop control system. The remaining state variables are the tip speed ratio, , the blade pitch, , and the filtered tracking error which will be introduced in the next section. Variable speed wind turbines have 3 main regions of operation. A stopped or just starting turbine is considered to be operating in Region 1. A turbine operating in Region 3 is operating at the generator rated power. It is typically being limited due to high wind speeds which can damage the mechanical system, thus forcing it to capture less power. In Region 2, somewhere between region 1 and 3, the energy captured by a variable speed turbine is limited by its efficiency of energy capture. The efficiency of power capture is a function of the tip speed ratio and the blade pitch. The power captured from the wind follows the relationship = (, ) 3
1 2

The function (, ) is an unknown nonlinear relationship described by a three dimensional surface. As a first pass in this development of a robust scheme of power capture control and optimization, a simplified representation of Cp was adopted and the Cp surface is shown in Figure 1. In general, the Cp function shape is convex which approximately resembles an inverted paraboloid-like function having a single maximum denoted as ( , ) > 0 for some * and *. (, ) is maximized as and . While the function of Figure 1 is not convex for its domain of definition, it is in a neighborhood of the extremum point. In Eqs. (2) (6), the air density , and the wind speed v is presumed measurable with some measurement error w. It is also assumed that the shaft angular velocity is a measurable quantity. The angular shaft acceleration, , however, is not available from measurements. Neglecting yaw error and dividing Eq. (6) by , a representation of is obtained as A Lyapunov-based approach is proposed for developing the wind turbine control law. The constraints to be satisfied by the controller include robust estimation of the turbine aerodynamic properties, stabilization of the tip speed ratio about a set point, and optimization of the power capture coefficient, Cp. The controller design will be divided into two steps. The first is the control of the inner loop which is accomplished by the use of a robust identifier-based controller. This controller is responsible for the regulation of the angular shaft velocity, , to the desired shaft velocity, d, This controller also provides an estimate of the unknown aerodynamic torque, which will be denoted as . Secondly is the development of the Lyapunov-based controller which uses this estimation for generating the desired state trajectories for the robust controller to regulate. This is known as the outer loop controller. A Lyapunov candidate function will be used to ensure that the values of the outer loop states, and , converge towards the optimal values of * and * which maximize Cp, thus Cp converges to Cp*. =
1 2 (,)

= .

(6)

2 .

(7)

(2)

where , the tip speed ratio, is given by =


(3)

Cp is the coefficient of performance or rotor power coefficient, is the air density, A is the rotor swept area, v is the wind speed, and Paero is the captured power. The coefficient of performance is defined as the ratio of the power captured to the available power in the wind. According to the Betz limit, this ratio is theoretically bounded at 0.59. The Cp relationship is defined as where is given by and is given by (, ) =
1 2

(4)

= 3

III. IDENTIFIER-BASED CONTROL DEVELOPMENT A nonlinear robust controller is developed to regulate the angular velocity of the system to a desired set-point. As a result, an estimation of the unknown parameter, aero, is also obtained. A reaction torque from the generator, denoted as , is used to control the shaft speed of the system. The open-loop dynamics, as stated in (1), are given by = ( )
1

(5)

(8)

Copyright 2010 by ASME

where is an unknown nonlinear term in the dynamics, which contains the term (, ) as shown in (7). The equation (8) can be rewritten as The objective of this identifier-based control design is twofold: (1) Achieve asymptotic tracking in the sense of () where represents a desired angular velocity. (2) Estimate the unknown value of the nonlinear function . A tracking error of this desired trajectory is defined as To facilitate the subsequent control design and analysis, a filtered tracking error, denoted as (), is defined as where denotes a positive constant. The filtered tracking error r(t) is not measurable because it depends on (t) which is also not measurable. The quantity r(t) is the last state variable. Multiplying (11) by J gives = + + + = + () = + (). (12) = + (11) = . (10) + = . (9)

where ks, , and c are positive control gains, and sgn( ) denotes the standard signum function. [Note that was defined previously in (11).] The time derivative of (16) is given by
d () d

After substituting (17) into (15), the following closed-loop error system can be obtained as = ( + 1) sgn() + (, , ) = + . () = . (18)

= ( + 1) + sgn().

(17)

where the auxiliary function N denotes the unmeasurable auxiliary term of (19)

Before analyzing the stability of the closed-loop system, we perform the following manipulation on (18). Let us introduce a new, unmeasurable auxiliary parameter Nd that is defined as (20)

The reason to introduce N and Nd is to facilitate the stabilizing analysis. The main result can be stated in the following theorem. Theorem 1: The controller given in (13) and (16) achieves semi-global asymptotic position tracking in the sense that provided that is selected according to the sufficient condition > | | +|

where (9) and (10) were utilized. Based on the expression in (12) the control torque is designed as where () denotes a subsequently designed control term. By substituting (13) into (12), the result is (13)

e(t)0 as t

From (14) it is evident that as () 0, () will identify the unknown input torque . Therefore, it is desirable to design a controller such that () 0. To facilitate the design of (), we differentiate (14) as

(14)

In addition, all system signals are bounded, and can be identified in the sense that. Proof: The proof is similar to that shown in [10]. (() )0 as t.

|.

Based on (14) and the subsequent analysis, the control law () is designed as [8,9] + [( + 1)() + sgn(())]
0

= +

d () d

(15)

() = ( + 1)() ( + 1)(0)

As a result, () will be used to estimate the quantity aero, which is used in the determination of .

(16)

IV. ESTIMATION STRATEGY OF CP Now that there is a method which can regulate the state, , of the inner loop system about a desired set point, a control law must be written to generate a trajectory, d, for the robust controller to follow. It is desirable that the trajectory converges to the optimal point by means of manipulating the outer loop state variables, blade pitch and tip-speed ratio. We begin with

Copyright 2010 by ASME

the development of the Lyapunov controller. The Lyapunov candidate function is

where

Taking the time derivative of V provides (23)

= .

1 2 2

where is a positive constant. For the prop pitch, we choose


=

(21)

(29)

(22)

both set points so that CpCp*. Because the gradients of and


and is also a positive constant. The time derivatives, and , provide a way of updating

We assume that the optimal operating point, Cp*, is constant or slowly time varying. Therefore, the time derivative of Cp* is small and can be neglected. Replacing with the Lyapunov time derivative expands to
= + . = +

= .

must be estimated, as shown in (26) and (27), a discrete

update law must be written to represent (28) and (29)., The subscript k is used to denote the current time step and k+1 to denote the next future time step. For the tip speed ratio, the discretized update law is given by where is the current value of the desired tip-speed ratio set point and +1 is the new desired tip-speed ratio set point. A similar expression for the discretized prop pitch update is found to be +1 = + sgn(

(24)

+1 = + sgn(

(30)

. The term is guaranteed to always be greater than zero because of its definition in (22). Cp* has been denoted as the maximum value which Cp can attain. The outer loop Lyapunov controller developed in this section works in conjunction with the identifier-based controller of Section III. The Lyapunov controller will choose a new set-point once the shaft angular velocity has been stabilized within a given tolerance by the inner loop robust controller. Because of the waiting time for the stabilization of d, the system becomes discrete in nature. Using numerical
namely,

To ensure that always remains negative semi-definite, and are selected to parallel their respective gradients of Cp, and

(25)

differentiation,

and

and

are determined approximately as

where is the estimate of Cp. The time derivatives of the state variables, and , are chosen so that the Lyapunov time derivative, , remains negative semi-definite. For the tip speed ratio, we choose
=

1 1

(26)

1 1

V. ESTIMATION OF SYSTEM NONLINEARITIES From Section III, the quantity is used to approximate the unknown system nonlinearities contained in the aerodynamic torque . Under the stability analysis of the identifier-based controller, , as t becomes large. For a given set point, + when = d , where is a specify = stabilization error tolerance and represents the resulting estimation error in . At this point very closely approximates . Now define a bounded estimation of as By substituting Eq. (33) into Eq. (4), we now have an estimation of the unknown function of energy capture
=

The signum function is used to extract the sign information from the estimated gradient partial derivatives. The dynamics of actuating , the rotor blade pitch, have been neglected under the assumption that they are small compared to the stabilization of d. The set point d is determined by rewriting (3) in terms of +1 as =
+1

).

(31)

(32)

(27)

= .

(33)

(34)

or rewritten, by using Eqs. (33), (5), and (6), as

(28)

Copyright 2010 by ASME

VI. SIMULATION A wind power capture system is considered with a time varying wind velocity input and measurement error. The control approach which has been proposed is used to maximize the coefficient of performance, Cp. As an initial test of the method presented in this paper, the coefficient of performance is chosen as (, ) = 0.45 0.3(7.5) 0.3(+1) (36)

The new desired operating point , +1 , and +1 are chosen to maximize the function .

= 1 3.
2

(35)

The controller block implements the identifier-based control procedure presented in Section III. The error signal between shaft angular velocity and the desired shaft angular velocity becomes the input. The outputs are the control torque, c, and the estimation of the unknown input torque, aero. The control law computing c is taken from Eq. (13). This controller regulates the turbine to the desired set-point, driving e(t) to zero.

which provides simplicity in calculation as well as a reasonable representation of the shape of the parameter in a neighborhood of the extremum. Figure 1 shows a plot of Cp as a function of and . The function in Eq. (36) is chosen because it exhibits the form of an actual Cp curve with a single maximum at some * and *. From Eq. (36), the optimal values of and are, = 7.5 and = 1 degree, where the maximum value of Cp = 0.45.

0.5 0.4 0.3


C
p

0.2 0.1 0 5 0 -5 5 -10 0 15 10

FIGURE 1. 3-Dimensional Plot of Cp(,). Peak Cp = 0.45 at = 7.5, = -1.

Figure 2 illustrates the implementation of the control strategy as a block diagram. The wind velocity, v(t), and blade pitch, (t), are inputs to the nonlinear block which applies the unknown Cp function shown in Figure 1. This generates the aerodynamic torque applied to the wind turbine. The dynamics of the wind turbine are described by Eq. (1). Note that the output of the nonlinear block, aero, is applied directly to the wind turbine plant model and is not applied to any other block. No other block in Figure 2 has any knowledge of the function in Eq. (36).

From the assumptions in Section V and the convergence proof of the identifier-based controller, () aero as t, thus, providing the Lyapunov controller with an estimate of the unknown nonlinear input torque. This torque estimation information is then used to estimate the coefficient of performance and compute the necessary gradients in Eqs. (26) and (27). The controller uses this estimated torque, the wind velocity, v(t), and the shaft angular velocity, , to choose the next set-point for the controller, d, and the new value of by the method presented in section IV. Because Cp is maximized as a function of and , the Lyapunov method is able to converge toward * and *. Recall from Eq. (3) that is a function of and v(t). Also, recall when the set-point is chosen by using Eq. (32), this allows the controller to track a time varying wind velocity. A Kaimal wind model generator is used from the Simulink Wind Blockset, [13], to create a realistic wind condition with a mean value of 10 m/s and a turbulence of 12%. A zero mean, white noise is also generated and added to the wind velocity signal. The parameters used to generate this wind signal can be seen in Table 1. It should be noted that when an error in measurement of the wind velocity was introduced, the results were not noticeably altered. When using an estimation of Cp to determine d, the wind velocity is cubed and in the

FIGURE 2. Wind capture system with proposed controller. The robust controller and wind turbine makes up the inner control loop. The Lyapunov-based extremum seeking controller is the outer loop controller.

Copyright 2010 by ASME

TABLE 1. SIMULATION PARAMETERS AND VALUES

m/sec

denominator. Thus an error in wind velocity measurement must be very large to skew the estimation. A numerical simulation is conducted to observe the performance of the proposed control strategy using MATLAB and Simulink. The maximization process begins with the initial conditions shown in Table 1. The focus of the simulation results are the bounded convergence of Cp to Cp*, to *, and to *, given that the control algorithm has no knowledge of the optimal value of Cp*, *, or *, or the nonlinear behavior of the Cp power coefficient curve. The parameter values used in the simulation are given in Table 1. These values of J, CD, and R are chosen to approximate the values for a large (> 1.0 MW) wind turbine. Figures 3 6 show the performance of system. For the large-scale wind turbine, we are able to show Cp converge from an initial value near zero to a region very near to the optimal value within approximately 135 seconds. The initial conditions are intentionally chosen far away from * and * to demonstrate the performance of the controller. In a practical application, the turbine would most likely be initialized very close to its optimal operating region. This would greatly shorten convergence times.

was accurate. It is also observed that this control strategy is able to hold the operation of the turbine very close to the optimum under given disturbances. For future research, an actual data-comprised Cp function will be used whereas in this initial application, Eq. (36) was used to emulate the Cp behavior. Another assumption made was the time varying nature of the optimal value of Cp*. In the development of this controller it was considered to be constant. Examining how the power capture optimization performs with physical system changes or variations of the optimal operating point are relevant to the development of a robust physical controller. Finally, recording the behavior of this controller implemented with a production turbine will be a useful proof of concept.
13 12 11 10 9 8

Parameter Variable J CD R v(t) w(t) (meas noise) 0 (init. value)

Value 100,000 1 35 10m/s with 12% turbulence N(0, 0.05) 7 0.5

Units kg m2 kg m /s m m/s m/s deg 2

50

FIGURE 3. A graph of the wind conditions input to the turbine, 10 m/s average with 12% turbulence.

150 100 Time

200

250

0 (init. value)

VII. CONCLUSIONS A Lyapunov control strategy was proposed to optimize the power coefficient, Cp, of a wind turbine operating in region 2. The identifier-based controller stabilizes the wind turbine shaft angular velocity about the set-point and estimates the unknown nonlinear input torque provided by the wind. Once within a given tolerance, the Lyapunov based algorithm chooses the next set points, and . Based on the convergence analysis provided, will approach *, will approach *, and Cp will approach Cp* as time goes to infinity within a bounded region. The simulation has implemented this control approach under realistic wind conditions and wind velocity measurement error. As a first pass at this wind power capture problem, the results are very promising for this type of control strategy. The convergence times were fast for this scale of turbine and the bounded convergence to the optimal coefficient of performance

6 4

Deg

2 0 -2

50

FIGURE 4. A graph of as it converges toward the optimal * o value of = -1 .

100 150 Time (Seconds)

200

250

Copyright 2010 by ASME

50

100 150 Time (Seconds)

200

250

FIGURE 5. A graph of as it converges toward the optimal * value of = 7.5.

0.5 0.4 0.3


C
p

0.2 0.1 0

50

150 100 Time (Seconds)

200

250

FIGURE 6. A graph of Cp converging toward the optimal value of * Cp = 0.45.

VIII. ACKNOWLEDGEMENTS This work is supported by the Kansas State University Electric Power Affiliates Program. The authors gratefully acknowledge this support.

REFERENCES [1] B. Malinga, J. Sneckenberger, J, Feliachi, Modeling and Control of a Wind Turbine as a Distributed Resource, Proc. 35th Southeastern Symp. Syst. Theory, Morgantown WV, 2003, pp. 108-112 [2] Brice Beltran, Tarek Ahmed-Ali, Mohamed El Hachemi Benbouzid, Sliding Mode Power Control of VariableSpeed Wind Energy Conversion Systems, IEEE Transactions on Energy Conversion Vol. 23, No. 2, June 2006 [3] Boubekeur Boukezzar, Mohamed MSaad, Robust Sliding Mode Control of a DFIG Variable Speed Wind Turbine for Power Production Optimization, 16th Mediterranean Conference on Control and Automation, June 2008 [4] Kathryn E. Johnson, Lucy Y. Pao, Mark J Balas, Lee J. Fingersh, Control of Variable-Speed Wind Turbines, IEEE Control Systems Magazine, June 2006 [5] Kathryn E. Johnson, Lee J. Fingersh, Mark J. Balas, Lucy Y. Pao, Methods for Increasing Region 2 Power Capture on a Variable-Speed Wind Turbine, Transactions of the ASME Vol. 126, November 2004 [6] Xin Ma, Adaptive Extremum Control and Wind Turbine Control, Ph. D. Thesis Denmark Technical University, May 1997 [7] Justin Creaby, Yaoyu Li, John E. Seem, Maximizing Wind Turbine Energy Capture using Multivariable Extremum Seeking Control, Wnd Engineering Vol. 33, No. 4, pp 361-388, 2009 [8] E. Iyasere, M. Salah, D. Dawson, J. Wagner, Nonlinear Robust Control to Maximize Energy Capture in a Variable Speed Wind Turbine, American Control Conference, June 2008 [9] C. Makkar, G. Hu, W. G. Sawyer, W. E. Dixon, Lyapunov-Based Tracking Control in the Presence of Uncertain Nonlinear Parameterizable Friction, American Control Conference, June 2005 [10] B. Xian, D. M. Dawson, M.S. de Queiroz, J. Chen, A Continuous Asymptotic Tracking Control Strategy for Uncertain Nonlinear Systems, IEEE Transactions on Automatic Control Vol. 49, No. 7, July 2004 [11] E. Iyasere, M. Salah, D. Dawson, J. Wagner, Nonlinear Robust Control to Maximize Energy Capture in a Variable Speed Wind Turbine Using a Separately Excited DC Generator, College of Engineering and Science Control and Robotics Technical Report, June 2008 [12] Kartik B. Ariyur, Miroslav Kristic, Real-Time Optimization by Extremum-Seeking Control, John Wiley and Sons Inc, 2003 [13] F. Iov, A.D. Hansen, P. Sorensen, F. Blaabjerg, Wind Turbine Blockset in Matlab/Simulink, tech. rep. Aalborg University, Denmark, 2004.

Copyright 2010 by ASME

You might also like