See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224968532

Modeling and Dynamics of a Horizontal Axis Wind Turbine

Article in Journal of Vibration and Control · November 2010

DOI: 10.1177/1077546309350189

CITATIONS READS

44 1,026

4 authors:

Sameh Kessentini S. Choura

University of Sfax Ecole Nationale d'Ingénieurs de Sfax
32 PUBLICATIONS 553 CITATIONS 55 PUBLICATIONS 1,031 CITATIONS

SEE PROFILE SEE PROFILE

Fehmi Najar Matthew Franchek

Ecole Polytechnique de Tunisie University of Houston
132 PUBLICATIONS 1,824 CITATIONS 250 PUBLICATIONS 2,729 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Fehmi Najar on 03 October 2015.

The user has requested enhancement of the downloaded file.

Modeling and Dynamics of a Horizontal Axis Wind
Turbine

S. KESSENTINI
Department of Mathematics and Physics, Sfax Preparatory Engineering Institute, BP 805,
Sfax 3000, TUNISIA

S. CHOURA
Micro-Electro-Thermal Systems Research Unit, National Engineering School of Sfax, B.P. W,
3038 Sfax, TUNISIA

F. NAJAR
Applied Mechanics Research Laboratory, Tunisia Polytechnic School, BP 743, La Marsa 2078,
TUNISIA

M. A. FRANCHEK
Department of Mechanical Engineering, University of Houston, Houston, TX 77204-4006, USA
(Received 1 December 20081 accepted 18 June 2009)

Abstract: In this paper, we develop a mathematical model of a horizontal axis wind turbine (HAWT) with
flexible tower and blades. The model describes the flapping flexures of the tower and blades, and takes into
account the nacelle pitch angle and structural damping. The eigenvalue problem is solved both analytically
and numerically using the differential quadrature method (DQM). The closed-form and numerical solutions
are compared, and the precision of the DQM-estimated solution with a low number of grid points is con-
cluded. Next, we examine the effects of pitch angle and blade orientation on the natural frequencies and
mode shapes of the wind turbine. We find that these parameters do not incur apparent alteration of the natural
frequencies. Then, we examine the linear dynamics of the wind turbine subjected to persistent excitations ap-
plied to the tower. We investigate the effects of the pitch angle and blade orientation on the linear vibrations
of the wind turbine. We demonstrate that the time response of the coupled system remain nearly unaffected.
We show that small vibrations of the tower induce important blade deflections, and thus, the dynamic tower–
blade coupling cannot be considered insignificant.

Key words: HAWT, mathematical model, pitch angle, DQM.

1. INTRODUCTION

Growing populations, regional uncertainty, and the transformation of developing countries

into manufacturing nations have accelerated the consumption of world-wide energy usage.
Further complicating this increase in energy usage are the resulting emissions spilled into the

Journal of Vibration and Control, 16(13): 2001–2021, 2010 DOI: 10.1177/1077546309350189

1
12010 SAGE Publications Los Angeles, London, New Delhi, Singapore
Figures 2, 4–6 appear in color online: http://jvc.sagepub.com
2002 S. KESSENTINI ET AL.

environment from fossil-based fuels such as coal and oil. While nuclear energy is a viable
clean energy alternative, renewable energy solutions are also being sought in an effort to
create a broader energy base for those locations where power transmission from tradition or
nuclear sites may be cost prohibitive.
Wind energy is one of the renewable energy options. Recently, wind turbines are finding
a more appealing home in offshore settings whereby the issues pertaining to landscape aes-
thetics, noise generation and blade size can be addressed. Moreover, these offshore locations
have an additional advantage for wind energy in that the wind boundary layer is greatly re-
duced over the rolling surfaces of the ocean. Therefore, more energy can be mined in offshore
facilities than for the onshore counter parts for the same free steam wind speeds.
As wind energy solutions have gained new momentum, additional technical challenges
must be resolved to make wind energy a viable, cost competitive, reliable alternative. Per-
haps the key engineering innovation that could lead to a revolutionary discovery in gigawatts
wind energy resides within the turbine blades. Essentially wind turbine blades that are larger,
lighter and stiffer are ideal albeit each are diametrically opposed specifications. In addition,
only providing maintenance when needed is paramount. Hence, wind turbine blades require
an interdisciplinary approach spanning materials engineering, structural health monitoring,
aerodynamics, and multivariable control for systems integration. Recent research studies on
wind turbines highlighted the need for optimizing the aerodynamic performance and mini-
mizing the vibratory loads (Negm and Maalawi, 20001 Maalawi and Badr, 20031 Younsi et
al., 2001), thereby reducing fatigue stress, improving service life and reducing maintenance
costs. This work seeks to develop dynamic models for a horizontal axis wind turbine. In
particular, the coupling between deflections and flapping of the tower and the blades are
captured. The proposed models will serve as an analytical test bed for structural health mon-
itoring of the blades and for power transmission control systems. These models demonstrate
the complexity of wind turbine blades as evidenced by the parametric dependence of their
natural frequencies on pitch angle.
There has been a developing interest in modeling flexible beams or a set of beams at-
tached to a rotating base or hub. These structures are modeled to describe the dynamics of
flexible robot manipulators, blades of turbines, helicopters and windmills, etc (Al-Bedoor,
19991 Christensen and Santos, 20051 Suryanarayanan and Dixit, 2007). Among the compre-
hensive studies on modeling of windmill blades is the one by Baumgart (2002). This study
developed a mathematical model for an elastic wind turbine blade mounted on a rigid test
stand and compared it with experimental results. Using the principle of virtual work, the
author derived a set of equations and formulated the eigenvalue problem, and the results are
compared with an experimental modal analysis of a 19 meter long blade. The modal analysis
showed that the torsional dynamics has insignificant effect on the overall response of the
blade. Recently, a dynamical model of a rotor blade described by a set of partial differen-
tial equations was developed to include the effects of gravity, pitch action and varying rotor
speed (Kallesoe, 2007). However, this paper models a single blade and does not describe
how these equations can be used to model a wind turbine where the blade–nacelle–tower
coupling becomes important.
For design purposes, it is important to develop accurate models for wind turbines with
flexible components. Owing to its practical importance, many studies have focused on the
problem of vibrations of flexible beams with different boundary conditions and engineering
applications (Nayfeh and Pai, 2004). In the literature, flexible blades are usually modeled
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2003

using the Euler–Bernoulli beam theory (Al-Bedoor, 19991 Larsen and Nielsen, 2006). Few
studies have treated the tower as a flexible structure (Larsen and Nielsen, 20061 Murtagh et
al., 2005). The early design of windmills either ignored completely the effect of structural
dynamics or included it through the use of estimated dynamic magnification factors (Lee
et al., 2002). In order to account for structural flexibility, Lee et al. (2002) represented a
horizontal axis wind turbine (HAWT) structure as a multi-flexible-body system formed of
rigid and flexible body subsystems. A more recent study by Murtagh et al. (2005) investi-
gated the along-wind forced vibration response of an assembly of wind turbine tower and
rotating blades subjected to a rotationally sampled stationary wind loading. They considered
that wind turbine assembly consists of three rotating rotor blades connected to the top of
a flexible annular tower modeled as discretized multi-degree-of-freedom bodies. The paper
also emphasized the importance of the blade/tower interaction on the structural dynamics.
Larsen and Nielsen (2006) developed a model of a HAWT structure in which the flexibility
of both the blades and tower were taken into account. In particular, they considered flapping
and edgewise vibrations of the tower and blades. They also conducted a stability analysis by
assuming harmonic motion of the tower support.
Large-scale wind turbines are subjected to vibration due to internal and/or external load-
ings. Modal data can be efficiently investigated for many purposes. It can be used for check-
ing modal parameters and/or verifying and improving analytical models contrasting exper-
imental to analytical data (Perdersen and Kristensen, 2003). Moreover, it can be used to
predict the change in dynamic properties due to physical modifications, and estimate the
necessary physical modifications required to obtain a set of desired dynamic property of
the wind turbine (Negm and Maalawi, 20001 Lee et al., 20021 Maalawi and Negm, 20021
Maalawi, 2007). In addition, modal data can be used for the purposes of modal control de-
sign (Christensen and Santos, 2005) and structural health monitoring (Lele and Maiti, 20021
Sinha et al., 2002).
The current study aims at developing a comprehensive mathematical model of a HAWT
structure by considering the tower and blade flexibilities in the flapping direction and the
effect of the nacelle pitch angle. The model assumes the Euler–Bernoulli beam approxima-
tion and considers the flapping deflections of the tower and blades. Using energy methods,
we develop a mathematical model for a HAWT structure described by a set of one ordinary
differential equation and four partial differential equations with 16 boundary conditions. The
modal analysis of the nonrotary system is then conducted using analytical and numerical so-
lutions. The convergence of the numerical solution is contrasted to the closed-form solution,
and thus, an accurate discretized model is retained for the dynamics analysis. We conduct a
parametric study to examine the influence of the pitch and blade orientation angles on the
HAWT eigenstructure. An analysis of the HAWT dynamics is provided by examining its
time response due to persistent wind loadings. A second parametric study is established to
investigate the effects of the pitch angle and blade orientation. Since it cannot be neglected,
structural damping is added to the linear model and then we examine its effect on the struc-
tural response.
2004 S. KESSENTINI ET AL.

Figure 1. HAWT coordinate systems.

2. MATHEMATICAL MODELING

We develop a mathematical model of a large-scale HAWT formed of a flexible tower of

uniform hollow cross section, a rigid hub–nacelle system and three flexible blades of uniform
cross section (see Figures 1a–c). The Euler–Bernoulli beam approximation is adopted for
modeling the dynamics of the tower–blade structure, which is assumed to undergo flapping
flexures 1 T and 1 Bi (i 2 12 22 3). Let 34t5 be the rigid body rotation, and 3 i 2 34t5 3
2
3
6 4i 4 15 be the orientation of the ith blade 4i 2 12 22 35 with respect to the y N axis, as
shown in Figure 1c.
In Table 1, we provide the HAWT parameters used in the simulations.
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2005

Table 1. Wind turbine parameters.

Symbol Description Numerical value/expression
LT tower height (m) 44.075
DT tower mean external diameter (m) 2.7
dT tower shell thickness (m) 0.015
ET tower modulus of elasticity (N/m2 ) 210 109
7T tower mass density (kg/m3 ) 1 7850 2
IT tower second moment of area (m4 ) 6 DT4 4 4DT 4 2dT 54 864
dN nacelle radius (m) 0.925
lN nacelle half length (m) 1.5
MN nacelle mass (kg) 8000
JN Z nacelle mass moment of inertia along z N (kg m2 ) M N d N 2 82
r0 hub radius (m) 0.75
EB blade’s modulus of elasticity (N/m2 ) 70 109
7B blade mass density (kg/m3 ) 2800
LB blade length (m) 26.25
tB blade shell thickness (m) 0.01
cB NACA1 4415 airfoil/blade chord (m) 2.1

2.1. HAWT Kinematics

In order to describe the kinematics of all windmill components, we introduce an inertial

frame (O, X, Y , Z) attached to the tower base, a moving frame (O N , x N , y N , z N ) fixed to the
nacelle center of mass, and three moving frames (O B , x B , y Bi , z Bi ) (i 2 12 22 3) attached to
the hub right end (see Figure 1). In particular, we let:
1. ix , i y and iz denote the unit vectors along the X , Y and Z axes, respectively1
2. ix N , i yN and iz N denote the unit vectors along the x N , y N and z N axes, respectively1 and
3. ix B , i yBi and iz Bi denote the unit vectors along the x B , y Bi and z Bi axes of the ith blade,
respectively.
The position vector RT of an arbitrary point of the tower is given by

RT 2 1 T 4yT 2 t5 ix 3 yT i y 9 (1)

The position vector R N locating the nacelle center of mass is

R N 2 1 T 5 L T ix 3 L T i y 3 d N i y N 2 (2)

where 1 T 5 L T 2 1 T 4L T 2 t5. Modern HAWT nacelles are equipped with controllers that reg-
ulate the pitch angle about the Z-axis. This angle is mainly varied to maintain a constant
level of harnessed wind power. As the tower deforms, 3the nacelle further rotates about the
1 T 33
Z-axis with an angle that can be approximated by . The transformation matrix from
yT 3 L T
the (O, X , Y , Z) frame to the (O N , x N , y N , z N ) frame is approximated by
2006 S. KESSENTINI ET AL.

4 3 3 7
1 T6 3 L T
cos 3 sin sin 4 cos 1 T6 3 L T 0
5 6 3
3 3 8
A25
6 4 sin 3 cos 1 T L T cos 3 sin 16 3
T LT 0 8
92 (3)
0 0 1

where the prime sign denotes the derivative with respect to the space variable. In the inertial
frame, R N is expressed by
4 7 4 7
1 T 5L T 0
5 8 5 8
RN 2 5
6 LT 8 3 AT 5 d N 8 2
9 6 9 (4)
0 0

where AT is the transpose of A. The position vector R Bi of an arbitrary point of blade i

(i 2 12 22 3) with respect to the inertial frame is given by

R Bi 2 1 T 5 L T ix 3 L T i y 3 l N ix N 3 d N i yN 3 1 Bi 4y B 2 t5ix B 3 4y B 3 r0 5i yBi 9 (5)

The unit vectors ix B , i yBi and iz Bi are related to those of the nacelle frame by the following
transformation matrix:
4 7
1 0 0
5 8
Bi 2 5 6 0 cos 3 i sin 3 i 8
9 4i 2 12 22 359 (6)
0 4 sin 3 i cos 3 i

The position vector R Bi can be expressed in the inertial frame as

4 7 4 7 4 7
1 T 5L T lN 1 Bi 4y B 2 t5
5 8 5 8 5 8
R Bi 2 5
6 LT 8 3 AT 5 d N 8 3 AT BiT 5
9 6 9 6 y B 3 r0 89
9 (7)
0 0 0

2.2. Equations of motion

We use the extended Hamilton’s principle to derive the equations of motion. This principle
states that
t2
4 T 4 V 3 W 5 dt 2 02 (8)
t1

where T is the variation of the total kinetic energy, V is the variation of the total poten-
tial energy and W the variation of the nonconservative energy due to external loads and
damping. The total kinetic energy T and the total potential V are given by
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2007

T 2 TT 3 TN 3 TB1 3 TB2 3 TB3 2 (9)

V 2 VT 3 VB1 3 VB2 3 VB3 3 VN 2 (10)

where TT is the tower kinetic energy given by

1 LT 1 2
TT 2 7 T A T 4R 7 T dyT 9
7 T 5T R (11)
2 0

TN is the nacelle kinetic energy (we assume that the nacelle has a symmetry so that the frame
(O N , x N , y N , z N ) corresponds to the principal axes) given by
1 3
7 N 3 1 JN Z 17 T6 3
2
TN 2 7 N 5T R
M N 4R 2 (12)
2 2 LT

TBi is the blade kinetic energy given by

1 LB 1 2
TBi 2 7 Bi 5T R
7 B A B 4R 7 Bi dy B 4i 2 12 22 35 (13)
2 0

and VT is the tower potential energy given by

1 LT 1 22 LT
VT 2 E T IT 1 T66 dyT 3 7 T A T g yT dyT 2 (14)
2 0 0

where g is the gravity acceleration. VN is the nacelle potential energy given by

VN 2 M N g 8R N 2 i y 92 (15)

where 8 9 denotes the scalar product and VBi is the blade potential energy given by

1 LB 1 22 LB
VBi 2 E B I B X 1 66Bi dy B 3 7 B A B g 8R Bi 2 i y 9 dy B 4i 2 12 22 359 (16)
2 0 0

The use of Hamilton’s principle yields a nonlinear set of four partial differential equations,
one ordinary differential equation and 16 boundary conditions. Here, we assume small vibra-
tions of the tower and blades, and thus nonlinear terms involving 1 T and 1 Bi are neglected.
By incorporating the effect of structural damping, we obtain the following set of dynamical
equations, expressed in nondimensional form:

L 3T FxT
1 T 3 C T 17 T 3 Y4 1 T 2
6666
2 (17)
E T IT

2 2
LT dN 6
1 Bi 3 C B 17 Bi 3 4sin 4 cos 1 T6 51 Y 1 g
6666
3 Y1 1 Bi 3 cos 1 T 51 3 1 51
LB LB LB T
  
r0 L 3B Fx Bi
3 3 yB cos 3 i 1 T6 51 4 237 sin 3 i 17 T6 51 2 4i 2 12 22 352 (18)
LB E B IB X
2008 S. KESSENTINI ET AL.

  
3r0 3r02 2 1 1 2
13 3 2 32 3 3 L 2B Fz B1 3 Fz B2 3 Fz B3 y B d y B 9 (19)
LB LB 7 B A B L 3B 0

The associated boundary conditions are

at yT 2 02
1 T 2 02 (20)
1 T6 2 02 (21)
at yT 2 12
1 3
L 2B 666 L T
1 4 Y2 43 3 Y 3 5 1 T 4 Y2 cos 1 Bi d y B
L 2T T LB 0 i21
 
dN lN dN
4 Y2 3 cos 3 3 sin 3 Y3 cos 1 T6 2 02 (22)
LB LB LB
 
L B 66 dN L T 1 3r0 3r02
37 1 T6
2
4 1 T 4 Y2 Y3 2 cos 1 T 3 Y2 3 3 2
LT LB 2 2L B 2L B
 
dN L T lN L T d2
4 3Y2 2
cos 3 2 sin 1 T 4 Y2 Y3 N2 1 T6 4 Y2 Y5 1 T6
LB LB LB
 
1 3d N2 3r0 3r 2 3l 2
4 Y2 3 2 3 3 02 3 2N 1 T6
2 LB 2L B 2L B LB
2
3 gY2 [3 sin d N 4 3 cos l N 3 sin d N Y3 ]
L 2B
1 
3  
r0
4 cos 3 i 37 1 Bi 3 1 Bi 3 3 sin 3 i 1 Bi
2
3 Y2 yB 3
0 i21
LB
2

dN
4 1 Bi 4g cos 1 Bi d y B 2 02 (23)
LB LB

at y B 2 02

1 Bi 2 0 4i 2 12 22 352 (24)

1 6Bi 2 0 4i 2 12 22 352 (25)

at y B 2 12

1 66Bi 2 0 4i 2 12 22 352 (26)

1 666
Bi 2 0 4i 2 12 22 352 (27)
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2009

where C T and C B are the nondimensional damping coefficients associated with the tower and
blades, respectively, FxT is an external distributed force applied to the tower, Fx Bi and Fz Bi are
the applied wind loadings on blade i in the flapping and edgewise directions, respectively,
 3 is the control torque and

t yB yT 1 Bi 1T 7 B AB
t 2 2 yB 2 2 yT 2 2 1 Bi 2 2 1T 2 2 2 L 2B 2
LB LT LB LT E B IB Z

E B IB Z E B IB Z L 3B M N L 4T 7 T A T JN Z L B
Y1 2 2 Y2 2 2 Y3 2 2E I
2 Y4 2 2E I
2 Y5 2 2E I
9
E B IB X E T IT B BZ T T B BZ

3. Eigenvalue Problem

To formulate the eigenvalue problem, we set the wind loadings and damping forces to zero
and freeze the rotation of the rotor. To this end, we determine the natural frequencies and
mode shapes of the HAWT structure analytically and numerically using the differential
quadrature method (DQM). Assuming harmonic motion with nondimensional frequency 
and setting 37 to zero, the solution of the eigenvalue problem can be written in the following
form:
1 2 1 2 1 2 1 2
1 T yT 2 t 2  T yT es t 2 1 Bi y B 2 t 2  Bi y B es t 4i 2 12 22 352 (28)

where s 2 2 41, and  T and  Bi are the mode shapes associated with the flapping vibra-
tions of the tower and blades, respectively. The nondimensional eigenvalue problem and the
associated boundary conditions are, therefore, described by the following set of ordinary
differential equations:
6666
 T 4 2 Y6  T 2 02 (29)

2
 6666 6
Bi 4  Y1  Bi 4 cos  T 51 Y 1 g
2
LB
    
dN r0 LT
42 Y1 3 3 y B cos 3 i  6T 51 3 cos  T 51 2 0
LB LB LB
4i 2 12 22 352 (30)

 T 50 2 02  6T 50 2 0 (31)
 
L 2B 666 1 
3
1 2
 51 3 2 Y2 cos  Bi d yB
L 2T T 0 i21

2 Y2  
3 L T 43 3 Y3 5 T 51 3 4d N 43 3 Y3 5 cos 3 3l N sin 5  6T 51 2 0 (32)
LB
2010 S. KESSENTINI ET AL.

L B 66 2 Y2 
4  T 51 3  T 51 46l N L T sin 3 2d N L T cos 43 3 Y3 55
LT 2L 2B
1 2
3  6T 51 6l N2 3 L 2B 3 3L B r0 3 3r02 3 2d N2 43 3 Y3 5 3 2L 2B Y5

2 Y2 1 
3
1 2 
3 r0 3 L B y B  Bi cos 3 i 3 d N  Bi 4g 2
cos  Bi d y B 2 02 (33)
LB 0 i21

 Bi 50 2 02  6Bi 50 2 02  66Bi 51 2 02  666Bi 51 2 0 4i 2 12 22 359 (34)

Solving equations 29 and 31 yields

 T 4 yT 5 2 K 1 cos4 yT 5 4 K 1 cosh4 yT 5 3 K 2 sin4 yT 5 4 K 2 sinh4 yT 52 (35)

where  2  Y6184 . We substitute equation 35 into equations 30 to derive expressions for

 Bi . Solving equations 30 and the associated boundary conditions at y B 2 0, the expressions
for  Bi are given by

 Bi 4 y B 5 2 K 23i cos4 B y B 5 4 K 23i cosh4 B y B 5 3 K 53i sin4 B y B 5

LT g 2 1 6
2
4 K 53i sinh4 B y B 5 4 cos  T 51 3 sin 4 cos  51
LB 2 L B Y1 T

 
r0 dN
4 cos 3 i 3  6T 51 4 cos 3 i  6T 51 y B 4i 2 12 22 352 (36)
LB LB

where  B 2  Y1184 . Hence a set of eight unknowns K k (k 2 12 9 9 9 2 8) is formed. The use

of the remaining eight boundary conditions

 66Bi 51 2 02  666Bi 51 2 0 4i 2 12 22 35

and equations 32–33 yields a set of eight linear equations with eight unknowns K T 2
K 1 9 9 9 K 8 . This set can be written in matrix form as

F K 2 [0]9 (37)

The characteristic equation is obtained by equating the determinant of matrix F to zero, i.e.

P45 2 det[F] 2 09 (38)

Because of the complexity of its governing equations, it is necessary to use a powerful nu-
merical method to simulate the time response of the wind turbine. In this study, we propose
the use of the DQM for one-dimensional systems. The DQM was first introduced by Bellman
and Casti in 1971 and 1972 (Bellman et al., 1972). Since then, the method has been applied
successfully in various research areas. A significant merit of the method is its high efficiency
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2011

in computing nonlinear problems. Compared with standard numerical techniques, such as

the finite element and finite difference methods, the DQM produces a solution of reasonable
accuracy with relatively small computational effort. This method also does not require one to
seek trial functions satisfying boundary conditions as in the Galerkin method. The basic con-
cept of the DQM is to approximate the derivative of a function f 4 5 with respect to the space
variable  at a given sampling point as a weighted linear combination of the function values
at all sampling points in the domain of  . Differential equations are then transformed to a
set of algebraic equations for time-independent problems and a set of ordinary differential
equations in time for initial-value problems (Tomasiello, 1998). Hence, for a dimensionless
variable   [02 1] and a set of N discretization points over the spatial domain, the r th-order
derivative of f 4 ) at  2  i is approximated by
3 
d r f 33
N
2 Ari j f i 4i 2 12 22 9992 N 52 (39)
d r 3 2 i j21

where Ari j are the corresponding DQ weighting coefficients of the rth-order derivative. More-
over, the DQM approximates the integral of the function f 4 5 over [02 1] by

1 
N
f 4 5d 2 Ck f k 9 (40)
0 k21

The weighting coefficients Ari j are obtained directly and most accurately by the formulae
taken from Shu and Richards (Shu and Richards, 1992a1 Shu and Richards, 1992b). The
off-diagonal terms of the weighting coefficient matrix of the first-order derivative turn out to
be
N 1 2
121212i  i 4  1
Ai j 2 1
1
2 N 1 2 4i2 j 2 12 22 9992 N2 i 2 j 59 (41)
i 4  j 121212 j  j 4  1

The off-diagonal terms of the weighting coefficient matrix of the higher-order derivative are
obtained through the recurrence relationship:
 
Ar41
Ari j 2 r Ar41
ii Ai j 4
1
1 ij 2 4i2 j 2 12 22 9 9 9 2 N 2 i 2 j5 42  r  4N 4 1559 (42)
i 4  j

The diagonal terms are given by


N
Arii 2 4 Ari1 4i 2 12 22 9992 N 5 41  r  4N 4 1559 (43)
121212i

The weighting coefficients Ck of the integral are derived using the following Newton–Cotes
integration formula (Tomasiello, 1998):
2012 S. KESSENTINI ET AL.

1 
N
 4 i
Ck 2 d 9 (44)
0 i212i2k k 4 i

A decisive factor in the accuracy of the DQM solution is the choice of the sampling points.
Here, the Chebyshev–Gauss–Lobatto distribution of the grid points is adopted:
  
1 i 41
i 2 1 4 cos 6 4i 2 12 9992 N 59 (45)
2 N 41

It can be clearly seen that the boundary points are very close to their adjacent points. This
distribution, applied to problems with boundary conditions, was shown to yield more accu-
rate results with less grid points than other sampling schemes such as the equally spaced grid
point (Bert and Malik, 1996).
We first apply the DQM to equations 29–34 to test its convergence with respect to the
analytical solution obtained from equations 37 and 38. This provides the minimum number
of grid points needed for an acceptable discretization of the dynamic equations 17–27. We
let  T2i ,  B12i ,  B22i and  B32i be the mode shapes evaluated at node i. Using N sampling
points and applying the DQM to the governing equations and boundary conditions, we get the
following system of algebraic equations associated with the interior points i 2 32 9 9 9 2 N 42:

N
Ai24 j  T2 j 4 2 Y6  T2i 2 0 4i 2 32 9992 N 4 252 (46)
j21
    

N
dN r0
N
Ai24 j  Bl2 j 4  Y1  Bl2i 3
2
3 3 i cos 3 l A1N 2 j  T2 j
j21
LB LB j21

  
LT g 2 
N
3 cos  T2N 3 sin 4 cos Ai21 j  T2 j 20
LB LB j21

4l 2 12 22 35 4i 2 32 9992 N 4 259 (47)

The boundary conditions, expressed in DQM-discretized forms, become


N
 T21 2 0 and A112 j  T2 j 2 02 (48)
j21

LB  2 2 Y2 
N
4 A N2 j  T2 j 3  T2N 46 sin l N L T 3 2 cos d N L T 43 3 Y3 55
L T j21 2L 2B


N
1 2
3 A1N 2 j  T2 j 6l N2 3 L 2B 3 3L B r0 3 3r02 3 2d N2 43 3 Y3 5 3 2L 2B Y5
j21

2 Y2   1 2 
N 3
3 Ci r0 3 L B  i  Bl2i cos 3 l 3 d N  Bl2i 4g 2
cos  Bl2i 2 02 (49)
L B i21 l21
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2013

 
L 2B  3
N N
dN 3l N
A  3  Y2 2
43 3 Y3 5 cos 3 sin A1N2 j  T2 j
L 2T j21 N 2 j T2 j LB LB j21

  
3L T L T Y3 
N
1 2
3  T2N 3 3 Ci cos 4 B12i 3  B22i 3  B32i 5 2 02 (50)
LB LB i21


N 
N
 Bl21 2 0, A112 j  Bl2 j 2 0, A2N 2 j  Bl2 j 2 0
j21 j21


N
and A3N 2 j  Bl2 j 2 0 4l 2 12 22 359 (51)
j21

We take advantage of the linear form of the boundary conditions to express a set of dis-
placements as linear combinations of the remaining ones. In particular, we use equations 51
to write the  Bl displacements at nodes 1, 2, N 4 1 and N as linear combinations of those
at the other nodes. Similarly, we use equations 48 to express the  T displacements at nodes
1 and 2 in terms of the remaining displacements. The DQM-discretized eigenvalue problem
of dimension 4N 4 14 can be written in the following matrix form:

Q R 2 02 (52)

where
 
RT 2  T23 2 9992  T2N 2  B123 2 9992  B12N 42 2  B223 2 9992  B22N 42 2 999 B323 2 9992  B32N 42 9

The characteristic equation is obtained by setting the determinant of matrix Q to zero,

i.e.

det[Q] 2 09 (53)

We use Mathematica to determine the natural frequencies  by solving numerically the roots
of the characteristic equations 38 and 53 for the analytical and DQM-estimated solutions. In
Table 2, we display the first four DQM-estimated natural frequencies using different numbers
of sampling points along with the exact solution for 2 0 and 3 2 0. The results indicate
that the natural frequencies converge to the exact values using nine and 12 grid points with
three and six digits after the decimal, respectively.
Figure 2 compares the DQM-estimated mode shapes (dashed curves), associated with the
first natural frequency, with the exact ones (solid curves). The results show great accuracy
of the DQM with only nine grid points. Here, it should be pointed out that the fundamental
frequencies associated with the corresponding uncoupled cantilevered tower and blade are
2.65835 and 0.655249, respectively. This indicates that the fundamental frequency of the
tower coupled with the nacelle and blades went down from 2.65835 to 1.57643. This is
expected since the tower inertia was increased by adding the nacelle and blades at the top
2014 S. KESSENTINI ET AL.

Figure 2. Comparison of the analytical and DQM-estimated mode shapes (nine grid points).

Table 2. Convergence of the natural frequencies using DQM.

1 2 3 4
Exact solution 0.628094 0.654108 0.655249 1.57643
DQM 6 grid points 0.638309 0.667017 0.668280 1.62677
DQM 7 grid points 0.623689 0.648609 0.649704 1.56115
DQM 8 grid points 0.628640 0.654770 0.655917 1.57551
DQM 9 grid points 0.628168 0.654206 0.655349 1.57728
DQM 10 grid points 0.628097 0.654111 0.655253 1.57649
DQM 11 grid points 0.628094 0.654107 0.655249 1.57642
DQM 12 grid points 0.628094 0.654108 0.655249 1.57643

Table 3. Effect of on the natural frequencies.

1 2 3 4
0 0.628094 0.654108 0.655249 1.57643
5 0.628207 0.654110 0.655249 1.57120
10 0.628591 0.654116 0.655249 1.56229
15 0.629239 0.654126 0.655249 1.55002

end. On the other hand, the set of frequencies of the blades coupled with the nacelle and
tower were slightly scattered in the neighborhood of the uncoupled frequency 0.655249.
We also considered a case in which the tower is flexible and all three blades are rigid. The
corresponding eigenvalue problem led to a slight drop in the first frequency from 1.57643 to
1.29.
Next, we investigated the effects of pitch angle and blade orientation on the first four
natural frequencies and corresponding mode shapes. The operating mode of wind turbines
requires that its blades rotate at a constant angular velocity, and thus their orientation with
respect to the reference frame varies with time. For this, the wind turbine at different frozen
orientations must be carefully examined. We first set 3 to zero and examined the sensitivity
of natural frequencies to alteration of the nacelle pitch angle . In practice, the pitch angle
is allowed to take values in the range [0 2 15 ]. According to Table 3, we notice that there
are no significant changes in the natural frequencies for the range of pitch angles under
consideration. The same observation was deduced with regard to the mode shapes.
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2015

Figure 3. Effect of 3 on the first four mode shapes.

Now, we let be zero and investigated the influence of changing 3 on the eigenstructure.
We checked that 3 has no remarkable effect on the natural frequencies. We note that the blade
configuration, which leads to the same eigenstructure, is reproduced for every angular change
of 120 . Figure 3 illustrates the effect of three values of 3 (0 , 30 , and 45 ) on the first four
mode shapes of the wind turbine. For the first configuration, characterized by symmetry
about the tower axis, the third and second blades have equal mode shapes. This equality is
no longer valid when symmetry of any two blades about the tower axis is not maintained. On
the other side, the mode shapes of the tower are not affected by the blade orientation.
2016 S. KESSENTINI ET AL.

4. DYNAMICAL ANALYSIS

In this paper, we also give particular emphasis to the time response of the nonrotating struc-
ture taking into consideration the structural damping (C T 2 0907497 and C B 2 0930417).
The coupling between the different parts of the wind turbine subjected to a harmonic loading
applied to the tower (equation 17) is carefully examined.
The application of DQM to equations 17–27 yields a set of ordinary differential equa-
tions, which are numerically integrated using Mathematica. Let 1 T2i , 1 B12i , 1 B22i , and 1 B32i
denote the mode shapes evaluated at node i. Using N grid points, this set is given by


N
Ai24 j 1 T2 j 3 C T 17 T2i 3 Y6 1 T2i 2 a sin t 4i 2 32 9992 N 4 252 (54)
j21

where a is the amplitude of the harmonic motion set to 09005 and  is its frequency:


N    
N
dN r0
Ai24 j 1 Bl2 j 3 C B 17 Bl2i 3 Y1 3 3 i cos 3 l A1N2 j 1 T2 j
j21
LB LB j21

  
LT g 2 N
3 cos 1 T2N 3 1 Bl2i 3 sin 4 cos A N 2 j 1 T2 j 2 0
1
LB LB j21

4l 2 12 22 352 4i 2 32 9992 N 4 259 (55)

The DQM-discretized forms of the boundary conditions are


N
1 T21 2 0 and A112 j 1 T2 j 2 02 (56)
j21


LB  2
N
Y2
4 A 1 T2 j 4 46l N L T sin 3 2d N L T 43 3 Y3 5 cos 5 1 T2N
L T j21 N2 j 2L 2B


1 2
N
3 6l N2 3 L 2B 3 3L B r0 3 3r02 3 2d N2 43 3 Y3 5 3 2L 2B Y5 A1N 2 j 1 T2 j
j21

g 2
3 Y2 43 sin d N 4 3 cos l N 3 sin d N Y3 5
L 2B

Y2   1 2 
N 3
4 Ci r0 3 L B  i 1 Bl2i cos 3 l 3 d N 1 Bl2i 3g 2
cos 1 Bl2i 2 02 (57)
L B i21 l21
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2017

Figure 4. Convergence of the DQM for linear behavior.

 
L 2B  3
N N
dN 3l N
A 1 4 Y 43 3 Y 5 cos 3 sin A1N 2 j 1 T2 j
L 2T j21 N 2 j
T2 j 2 3
LB LB j21

  
3L T L T Y3 
N
1 2
3 1 T2N 3 3 Ci cos 41 B12i 3 1 B22i 3 1 B32i 5 2 02 (58)
LB LB i21


N 
N
1 Bl21 2 0, A112 j 1 Bl2 j 2 0, A2N2 j 1 Bl2 j 2 0
j21 j21


N
and A3N2 j 1 Bl2 j 2 0 4l 2 12 22 35 (59)
j21

where the dot denotes the derivative with respect to t. We take advantage of the boundary
conditions 56 and 59 to reduce the number of unknown functions. We let the tower and all
blades be originally undeflected, i.e. all initial conditions are set to zero.
The convergence of the discretized model is tested for different numbers of DQM grid
points for 2 0, 3 2 0, and  2 195. The steady-state response of the end point of
the second blade (1 B2 51 ) is displayed in Figure 4 for different numbers of grid points. We
conclude that the DQM convergence is attained with N 2 8, which is adopted for further
dynamical analysis.
Figure 5 shows the time responses of the tower and the first blade for two different
loading frequencies,  2 0962 and  2 1957, and two pitch angles, 2 0 and 2 10 .
The values of these frequencies are selected close to the first and fourth frequencies of the
HAWT structure, which are closely related to the blades and the tower, respectively. We
observe that as the frequency is brought closer to the fourth frequency, the tower and blade
amplitudes are amplified by factors of 10 and 4, respectively. Despite small shifting of the
2018 S. KESSENTINI ET AL.

Figure 5. Effect of the pitch angle on the linear time response (—, 2 0 1 - - -, 2 10 ).

time response, observed as the pitch angle is increased from 0 to 10 , we notice that the
pitch angle does not yield significant alteration (Figures 5c and 5d). Careful investigation of
the influence of the blade orientation on the HAWT time response, displayed in Figures 6,
depicts noticeable changes in the blade vibration amplitudes. From these simulations, it can
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2019

Figure 6. Effect of 3 on the linear time response (—, 3 2 0 1 - - -, 3 2 30 1 – –, 3 2 45 ).

also be concluded that small amplitudes of the tower induce relatively large amplitudes of
the blades.

5. CONCLUSIONS

Developed in this paper is a HAWT linear model that takes into consideration the flapping
deflections of the tower and blades. We analyzed the HAWT eigenstructure via analytical
2020 S. KESSENTINI ET AL.

and numerical methods. We tested the convergence of the numerical solution (DQM) and
contrasted it to the exact solution. We concluded that retaining few grid points is sufficient to
estimate the eigenvalues and mode shapes of the structure. We conducted a parametric study
to examine the effects of pitch angle and blade orientation on the HAWT eigenstructure
and time response. We showed that both parameters do not have apparent influence on the
structural performance. The simulations of the time response showed that small vibrations
of the tower can induce important blade deflections.

NOTE

1. National Advisory Committee for Aeronautics

REFERENCES
Al-Bedoor, B. O., 1999, “Dynamic model of coupled shaft torsional, blade bending deformations in rotors,” Com-
puter Methods in Applied Mechanics, Engineering 169, 177–190.
Baumgart, A., 2002, “A mathematical model for wind turbine blades,” Journal of Sound and Vibration 251, 1–12.
Bellman, R., Kashef, B. G., and Casti, J., 1972, “Differential quadrature: a technique for the rapid solution of
nonlinear partial differential equations,” Journal of Computational Physics 10, 40–52.
Bert, C. W. and Malik, M., 1996, “Differential quadrature method in computational mechanics: a review,” American
Society of Mechanical Engineers 49, 1–27.
Christensen, R. H. and Santos I. F., 2005, “Design of active controlled rotor-blade systems based on time-variant
modal analysis,” Journal of Sound and Vibration 280, 863–882.
Kallesoe, B. S., 2007, “Equations of motion for a rotor blade, including gravity, pitch action and rotor speed varia-
tions,” Wind Energy 10, 209–230.
Larsen, J. W. and Nielsen S. R. K., 2006, “Non-linear dynamics of wind turbine wings,” International Journal of
Non-Linear Mechanics 41, 629–643.
Lee, D., Hodges, D. H., and Patil, M. J., 2002, “Multi-flexible-body dynamic analysis of horizontal axis wind
turbines,” Wind Energy 5, 281–300.
Lele, S. P. and Maiti, S. K., 2002, “Modeling of transverse vibration of short beams for crack detection, measure-
ment od crack extension,” Journal of Sound and Vibration 257, 559–583.
Maalawi K. Y., 2007, “A model for yawing dynamic optimization of a wind turbine structure,” International Journal
of Mechanical Sciences 49, 1130–1138.
Maalawi, K. Y. and Badr, M. A., 2003, “A practical approach for selecting optimum wind rotors,” Renewable Energy
28, 803–822.
Maalawi, K. Y. and Negm H. M., 2002, “Optimal frequency design of wind turbine blades,” Journal of Wind Engi-
neering and Industrial Aerodynamics 90, 961–986.
Murtagh, P. J., Basu, B., and Broderick, B. M., 2005, “Along-wind response of a wind turbine tower with blade
coupling subjected to rotationally sampled wind loading,” Engineering Structures 27, 1209–1219.
Nayfeh, A. H. and Pai, P. F., 2004, Linear and Nonlinear Structural Mechanics, Wiley, New York.
Negm, H. M. and Maalawi, K. Y., 2000, “Structural design optimization of wind turbine towers,” Computers &
Structures 74, 649–666.
Shu, C. and Richards, B. E., 1992a, “Application of generalized differential quadrature to solve two-dimensional
incompressible Navier–Stokes equation,” International Journal for Numerical Methods in Fluids 15, 791–
798.
Shu, C. and Richards, B. E., 1992b, “Parallel simulation of incompressible viscous flows by generalized differential
quadrature,” Computer Systems in Engineering 3, 271–281.
Sinha, J. K., Friswell, M. I., and Edwards, S., 2002, “Simplified models for the location of cracks in beam structures
using measured vibration data,” Journal of Sound and Vibration 251, 13–38.
 MODELING AND DYNAMICS OF A HORIZONTAL AXIS WIND TURBINE 2021

Suryanarayanan, S. and Dixit, A., 2007, “A procedure for the development of control-oriented linear models for
horizontal-axis large wind turbines,” ASME Journal of Dynamic Systems, Measurement and Control 129,
469–479.
Tomasiello, S., 1998, “Differential quadrature method: application to initial-boundary-value problems,” Journal of
Sound and Vibration 218, 573–585.
Younsi, R., El-Batanony, I., Tritsch, J. B., and Naji, H., 2001, “Dynamic study of a wind turbine blade with hori-
zontal axis,” European Journal of Mechanics, A/Solids 20, 241–252.

View publication stats