See discussions, stats, and author profiles for this publication at: [Link]

net/publication/236867830

Modeling and Control of Wind Turbine Tower Vibrations

Conference Paper · May 2010

DOI: 10.1063/1.3515600

CITATIONS READS
18 3,786

1 author:

Julian Asenov Genov

Technical University of Sofia
41 PUBLICATIONS 76 CITATIONS

SEE PROFILE

All content following this page was uploaded by Julian Asenov Genov on 16 May 2014.

The user has requested enhancement of the downloaded file.

 Modelling and Control of Wind Turbine Tower Vibrations
J. Genov1), B. Gilev2), Y. Slavchev3), G. Venkov2)

1) Department of Mechanics, Faculty of Transport, Technical University – Sofia

2) Department of Mathematical Analysis and Numerical Methods, Faculty of Applied
Mathematics and Informatics, Technical University – Sofia
3) Department of Engineering Logistics, Materials Handling and Building Machines,
Mechanical Engineering Faculty, Technical University – Sofia

Abstract: Developed is a model for studying the vibrations of a wind turbine tower subjected to
random wind loadings. The tower is modeled as an elastic Euler-Bernoulli beam, viscously damped and
fixed at the bottom. The concentrated mass of the turbine is located at the top of the tower where also
the wind loading is applied as a step function. The partial differential equation of the model is managed
by the method of separation of variables and along with the applied modal analysis produces an
efficient numerical-analytic solution algorithm. Demonstrated is the ability for dissipating the vibrations
through a properly designed controller. Numerical results are provided.

Keywords: method of separation of variables, tower vibrations, PID, LQR, linearized model

1. Introduction
The major wind turbine (WT)
components are shown in Fig.1.
The main operating parameters of a
WT are the generated power and the
loadings on the basic structural
assemblies. In order to keep the
parameters within nominal ranges it is
necessary to apply suitable control
systems.
Classical control systems, based on
the use of PID controllers, have been
applied for several huge WTs in the 70-s
and 80-s [6] with the objective to control
the power production [7]. It has been
estimated that a wide frequency band
controller is to be used to control the
generator operation in the case of
Fig. 1 turbulent flow [8].

A linear quadratic regulator (LQR) combined with a state observer has been applied
in [8] to control the power production of a constant speed WT by controlling the blade
pitch. The results obtained are good except for the freq intervals near the first mode of
the drive train vibrations.
There are different non-linear responses in the WT dynamic model that
significantly complicate the synthesis of a control system. A regulator that is designed
for a single WT operation point could turn out to be quite ineffective and even degrade
performance for the other points. As a possible solution to the problem, it has been
proposed to utilize several controllers for different operating points [3] and to switch
between the controllers when the corresponding states (parameters) change. Sudden
switching between any two controllers is associated with undesirable transients. A
method for avoiding this is proposed in [1], where the next controller that is to be
switched is tuned in advance for smooth transition.
Another, up to date control technique that considers the variable gain coefficients is
the adaptive control where the control coefficients are “adapted” to the changing
conditions. In [2] and [5] a method is presented for adaptive control of a three blade
WT. In [4] a predictive control is discussed for a non-linear WT system model.
Investigations have confirmed that the elasticity of the WT building elements is of
vital importance and must be included when looking for an effective WT design and
control. An approach with similar virtues is assumed in the current study.

2. Modeling tower vibrations as a system with distributed parameters

The tower with the nacelle are modeled as a beam, with

М F(t) distributed mass, elastic and damping properties, that is
fixed at the bottom, has a concentrated mass at the tip and
the tip is subjected to a corresponding wind loading - Fig.2.
The governing equation of the model is
l ∂2 y ∂y ∂ 2  ∂2 y 
m 2 + 2n m + 2  EJ 2  = δ ( x − l ) F (t ) (1)
∂t ∂t ∂x  ∂x 
for x ∈ [0, l ] and t ≥ 0
where F (t ) is external force, N/m; δ ( x − l ) -delta function;
l -tower height; t -time, s ; m = ρ S - linear mass, kg / m ,
Fig. 2
( ρ - density and S - cross-section area); J - area moment
of inertia, m ; n -damping coefficient ( sm) −1 ; E elasticity modulus, Pa .
4

The problem is solved for zero initial conditions and the following boundary
conditions (BCs):
- BCs for the fixed end
∂y
y (0, t ) = 0, (0, t ) = 0 (2)
∂x
- BCs for the end with the concentrated mass M
 ∂ ∂2 y ∂y ∂2 y 
 ( EJ ) − 2 n M − M  =0 (3)
 ∂x ∂x 2 ∂t ∂t 2  ( x =l ,t )
Equation (1) is to be solved in the form:
∞
y ( x, t ) =  Ti (t ) X i ( x) (4)
i =1
Substituting (4) in (1) and applying modal analysis [9] yields:
 F (t ) X i (l )
Ti (t ) + 2nTi (t ) + ωi2Ti (t ) = l
for i = 1, 2,... (5)
m  X ( x)dx + M X (l )
i
2
i
2

The eigenfunctions X i (x) can be derived from the system (1),(2),(3) for
n = 0; F (t ) ≡ 0 and they have the following expression:
ch(α i l ) + cos(α i l )
X i ( x) = ch(α i x) − cos(α i x) + (− sh(α i x) + sin(α i x) )
sh(α i l ) + sin(α i l )
where α i are the roots of
M
1 + cos(α l )ch(α l ) + α ( cos(α l ) sh(α l ) − sin(α l )ch(α l ) ) = 0 .
m
ωi2 m
The eigenfrequencies ωi can be calculated from the expression α i = . 4
EJ
Most commonly tower vibrations are approximated to the first term of (4) which
produces the first mode response of the tower. This is confirmed by the system
response for the free end ( x = l ) shown in Fig. 3, where the deflection corresponding
to the first eigefrequency ω1 = 7.73rad / s (Fig. 3а) is compared with the one
corresponding to the second eigenfrequency ω 2 = 71.41rad / s (Fig.3а) for wind
speeds V=13 m/s and V=17m/s.
-4
x 10
0.18 3.5

0.16 a 3 b
0.14
2.5
0.12
y(l,t) [m ]

y(l,t) [m ]

0.1 2

0.08 1.5

0.06
1
0.04

0.5
0.02

0 0
0 2 4 6 8 10 12 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
t [s] t [s]

Fig. 3

3. Modeling the wind turbine system

The general wind turbine system response is governed by the system:
Jw = M r (V , λ , β ) − M g − M l
 l 
T1 + 2nT1 + ω12T1 = A1 F (V , λ , β ) for A1 = X 1 (l ) /  m  X 12 ( x)dx + M X 12 (l ) 
 0  (6)
 
β + 2ς β + ωs ( β − β c ) = 0
2

1
M g + ( M g − M gc ) = 0
τ
The first equation governs the turbine and gear-train response reduced to the
turbine axis. The second equation governs tower fore-aft vibrations (first mode). The
third equation describes the blade pitch dynamics together with the controlling servo-
mechanism. The fourth equation is a first order approximation of the electric generator
dynamics [3].
Designations are as follows:
β c , M gc control inputs; J reduced mass moment of inertia; w rotor angular speed;
M g reduced torque of the generator; M l reduced drive-train resistive torque;
M r rotor torque; F wind force applied at the tower tip;
T1 (t ) X 1 (l ) displacement of the tower tip (first mode); τ time delay;
X 1 ( x) first mode shape; ω1 first natural frequency of the tower; n damping ratio;
β blade pitch angle; ς , ω s parameters of the servo-system;
The aerodynamic torque and force that act on the rotor and tower are defined as:
1 C (λ , β ) 1
M r = ρπR 3 P (V − T1 X 1 (l )) 2 and F = ρπR 2 Ct (λ , β )(V − T1 X 1 (l )) 2
2 λ 2
wR
where λ = is tip-speed ratio (TSR); ρ air density; R rotor radius; V
V − T1 X 1 (l )
wind speed; C P and Ct look-up table coefficients (Fig.3) experimentally determined.
Torque coefficient Performance coefficient
1 0.5

0.9 0.45
β=0 β=0
0.8 0.4
β=2
0.7 β=2 0.35
4
0.6 0.3
4 6
8
Cp

0.5 0.25
Ct

0.4
6 0.2 10
0.3 8 0.15
12
0.2 10 0.1
14
0.1 0.05

0 0
4 6 8 10 12 14 16 4 6 8 10 12 14 16
λ λ

Fig. 4 Ct and C p vs. TSR for β = 0,2,4,6,8,10,12,14,16,18,20 deg

4. Linearization of the turbine model

In order to synthesize LQ regulator, the equations
Jw = M r (V , λ , β ) − M g − M l and T + 2ξωT + ω 2T = A F (V , λ , β )
of the model in the previous section must be linearized with respect to the rotor speed
ϖ , the wind speed V (input) and the control parameter β (blade pitch angle) [10].
Regarding this, M r is expanded as a Taylor series in the vicinity of a selected point
(nominal point).
∂M r ∂M r ∂M r
M r (V , w, β ) = M r (Vnom , wnom , β nom ) + (V − Vnom ) + ( w − wnom ) + ( β − β nom )
∂V nom ∂w nom ∂β nom

∂M r ∂M r ∂M r
M r (V , w, β ) − M r (Vnom , wnom , β nom ) = (V − Vnom ) + (w − wnom ) + (β − β nom )
∂V nom ∂w nom ∂β nom
 ∂M r ∂M r ∂M r
ΔM r = ΔV + Δw + Δβ
∂V nom ∂w nom ∂β nom

Here M g is assumed to be constant, losses are zero and for these conditions, the
linearized equation becomes:
∂M r ∂M r ∂M r
J Δw = ΔV + Δw + Δβ
∂V nom ∂w nom ∂β nom
Similarly, F is expanded as a Taylor series in the vicinity of the nominal point
∂F ∂F ∂F
F (V , w, β ) = F (Vnom , wnom , β nom ) + (V − Vnom ) + ( w − wnom ) + ( β − β nom )
∂V nom ∂w nom ∂β nom
∂F ∂F ∂F
ΔF = ΔV + Δw + Δβ
∂V nom ∂w nom ∂β nom
Then, the governing equation of the tower tip motion is:
 ∂F ∂F ∂F 
ΔT + 2ξω ΔT + ω 2 ΔT = A  ΔV + Δw + Δβ 
 ∂V nom ∂w nom ∂β nom 
Since the process in mind is stationary, the last equation in the model (6) could be
neglected and the linearized model turns out to be:
 1 ∂M r   1 ∂M r 1 ∂M r 
 Δw   0 0 0 0  Δw   
   J ∂w nom    J ∂V nom J ∂β nom 
 ΔT   0 0 1 0 0  ΔT   0 0 
 ΔT  =  ∂F  ΔT  +  ∂F ∂F  ΔV  (7)
   A1 −ω 2
− 2n 0 0    A1 A1  Δβ 
 Δβ   ∂w nom  Δβ   ∂V nom ∂β nom  c 
 Δβ   0 0 0 0 1     0 0 
    Δβ   
 0 0 0 − ω s2 − 2ς   0 ωs 2 
where the partial derivatives at the nominal point are
∂M r ρR 2πV 3 ∂C p ∂F ρR 2πV 2 ∂C t
= =
∂β nom 2w ∂β nom ∂β nom 2 ∂β nom

∂M r ρR πV  C p
2 3
R ∂C p  ∂F ρR πV ∂C t
3
=  − 2 +  =
∂w nom 2  w wV ∂λ  nom ∂w nom 2 ∂λ nom

∂M r ρR π 
2 ∂C p  ∂F ρR π 
2
∂C 
=  3V 2 C p − VwR  =  2VC t − wR t 
∂V nom 2w  ∂λ  nom ∂V nom 2  ∂λ  nom

The nominal values of the parameters are:

Vnom = 8 m / s , λ nom = 14.94 , β nom = 4 deg , wnom = 2.95 s −1 , Ct nom = 0.4496 ,
∂C t ∂C t ∂C p ∂C p
Cp nom = 0.3572 , = - 0.0028 , = - 0.0507 , = 0.0039 = - 0.0225 .
∂λ nom ∂β nom ∂λ nom
∂β nom

Parameters that stay fixed are: R = 40 m , l = 72 m , J = 5 *10 4 kg m 2 ,

l = 72 m , M g = 2 * 10 4 N m , EI = 3 *1010 N m 2 , ρ = 1.25 .
 0.05 0.05

0.048 0. 048

0.046 0.046

0.044 0.044

0.042 0.042

0.04 0.04

0.038 0.038

0.036 0.036

0.034 0.034

0.032 0.032

0.03 0.03
5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50

a b
Fig. 5 Tower tip vibration for the nominal point: (a) non-linear model ,(b) linearized model

The Fig.4 results are obtained for the following wind velocities: 8m/s for the time
span 1s to 20s, 8.2m/s for 20s to 40s and 7.8 m/s after 40sec. It is assumed, in this
case, that there is no blade pitch control, but the angle is set to 4deg.

5. Control synthesis
The generated WT power depends highly on the wind speed. If there is a gentle
wind, the power output is lower than the nominal output of the generator. Therefore,
within this region, the major control system objective is to increase the WT output
power. If the wind is strong, the generated output power is greater than the nominal
output and power generation must be limited so as to ensure safe operation without
any overloading. An efficient way of managing the power output is to control the
blade pitch angle β , which in turn changes the aerodynamic efficiency of the turbine.
The boundary between the two mentioned regions is defined by the lowest wind speed
for which the turbine reaches its nominal power.
This work deals with regulators that will operate in the region of wind speeds above
the mentioned minimal speed. In addition, the major objective of the synthesized
regulators is to stabilize the WT shaft speed and thus achieve indirect dissipation of
tower vibrations.

5.2 Linear quadratic regulator (LQR)

This section explains the application of the classical LQR synthesis strategy for
controlling the blade pitch angle of a given wind turbine [10].
The linearized model from the previous section is rewritten shortly as:
Δx = A Δx + B Δu
 1 ∂M r   1 ∂M r 1 ∂M r 
 Δw   J ∂w 0 0  Δw   
     J ∂V J ∂β  ΔV 
 ΔT  =  0 0 1  ΔT  +  0 0  
 ΔT   ∂F    ∂F ∂F  Δβ 
   A1 −ω2 − 2n  ΔT   A1 A1 
 ∂w   ∂V ∂β 
   
where Δx = (Δw ΔT ΔT ) , Δu = (ΔV Δβ ) , the derivatives are calculated at the
T T

nominal point and the first three equations from the system (7) are used. Here, ΔV is
input signal, Δβ c is control signal and the control is synthesized with respect namely
to that control signal.
The linearized model performance usually degrades significantly when moving
away from the nominal point. Hence, this model is to be computed in series of
different points for various wind speeds V . Then, matrices A and B will depend on
the averaged values Vk of the wind, and the model will be reduced to the piecewise
linear form Δx = A(Vk )Δx + B(Vk ) , for k = 1,2,..., n .
For the present case, the objective function that is to be minimized with respect to
Δβ ( ΔV = 0 ) at each of the selected points is
∞
J 2 =  (Δx T QΔx + Δu T RΔu )dt
1
20

where Q and R are real symmetric, respectively non-negative definite and positive
definite matrices.
In order to control the system (6) the feedback loop
β c = β nom − K (Vk )( x − x nom )

is introduced, where K is also dependent on the average wind values Vk . The matrix
K is computed through the solution of stationary Riccati equation for various Vk
values. The matrices K (Vk ) are computed in advance and stored in memory so as to
realize an effective on-line control. K (V ) for the current wind V is obtained by linear
interpolation, i.e.
K (V ) = (1 − a) K (Vk ) + aK (Vk +1 ) , for a = (V − Vk ) /(Vk +1 − Vk )

0.05 0.05

0.048 0.048

0.046 0.046

0.044 0.044

0.042 0.042

0.04 0.04

0.038 0.038

0.036 0.036

0.034
0.034

0.032
0.032

0.03
0.03 5 10 15 20 25 30 35 40 45 50
5 10 15 20 25 30 35 40 45 50

a b
Fig. 6 Tower tip vibration in the presence of controller (a) and in the absence of controller (b)
 3.2 5

4.8
3.15
4.6
3.1
4.4
3.05
4.2

3 4

3.8
2.95
3.6
2.9
3.4
2.85
3.2

2.8 3
5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50

a b
Fig. 7 Variation in the rotor angular speed (a) and in the blade pitch (b) in the presence of LQR

6. Conclusion
Developed is a model for studying the wind turbine tower vibrations. The tower is
modeled as a system with distributed parameters and its natural frequencies and mode
shapes are obtained. As a result, the wind turbine is modeled as a system that accounts
for the dive-train, the tower and the blade pitch dynamic processes. The wind turbine
model is linearized so as to facilitate the designing of a linear quadratic regulator that
will be applied to stabilize rotor speed and consequently reduce tower vibrations.
Numerical results are provided that confirm the effectiveness of the controller.

References
[1] Bongers, P.M., 1994, Modeling and Identification of Flexible Wind Turbines and a
Factorizational Approach to Robust Control, Ph.D. Thesis, Department of Mechanical Engineering and
Marine Technology, Delft University of Technology.
[2] Bossanyi, E.A, 1987, “Adaptive Pitch Control for a 250 kW Wind Turbine,” Proceedings of
the 9th BWEA Wind Energy Conference, Edinburgh, Scotland, pp. 85-92.
[3] Bottasso C.L.,A. Croce , Scientific Report DIA-SR 09-01, Dipartimento di Ingegneria
Aerospaziale, Politecnico di Milano,Milano, Italy, January 2009.
[4] Findeisen F., L. Imland, F. Allgoewer, B. Foss, ‘State and Output Feedback Nonlinear Model
Predictive Control: An Overview’, European Journal of Control, 9:190–206, 2003.
[5] Freeman, J.B., and Balas, M.J., 1999, “An Investigation of Variable-Speed Horizontal-Axis
Wind Turbines Using DirectModel-Reference Adaptive Control,” Proceedings of the 37th AIAA
Aerospace Sciences Meeting and Exhibit, Reno,Nevada, pp. 66-76
[6] Kos, J.M., 1978, “On Line Control of a Large Horizontal Axis Energy Conversion System and
its Performance in a Turbulent Wind Environment,” Proceedings of the 13th Intersociety Energy
Conversion Engineering Conference, U.S.A., pp. 2064-2073.
[7] Kraan, I., 1992, Control Design for a Flexible Wind Turbine (in Dutch), Deflt University of
Technology, TUD-WBMRA-613, 1992.
[8] Mattson, S.E., 1984, Modeling and Control of Large Horizontal Axis Wind Power Plants,
Ph.D. Thesis, Department of Automatic Control, Lund Institute of Technology, Lund, Sweden.
[9] Timoshenko S., D. Young, W. Weaver, Vibration Problems in Engineering, Fourth Edition,
John Wiley and Sons, New York, 1974
[10] Wright, D.A.,2004, Modern Control Design for Flexible Wind Turbines, July 2004 ,
NREL/TP-500-35816

View publication stats