Bond Graph Model of Wind Turbine Blade

S. Agarwal* L. Chalal**,
G. Dauphin-Tanguy**, X. Guillaud***

* B-401, Garden Mansion, Old Airport Road, Kodihalli, Bangalore, 560008 India
(s27.agarwal@gmail.com)
** LAGIS FRE 3303, Ecole Centrale de Lille, BP 48, 59651 Villeneuve dAscq cedex, France
(laminechalal@gmail.com, genevieve.dauphin-tanguy@ec-lille.fr)
*** L2EP, Ecole Centrale de Lille, BP 48, 59651 Villeneuve dAscq cedex, France
(xavier.guillaud@ec-lille.fr)

Abstract: The design of a wind turbine system (sizing, control) according to specifications in terms of
quality of produced electricity, stability and safety, needs the use of performing simulation tools able to
change from simple static calculations assuming a constant wind to dynamic simulation that, from the
unsteady aerodynamic loads, models the aero elastic response of the entire wind turbine system, including
blades, tower, drive train, rotor and control system. In the paper, a bond graph model of a blade is
proposed. It consists in considering the blade as a Rayleigh beam composed of a number de sections
submitted to aerodynamic forces calculated using the BEM theory. The model is validated using data
from a NACA 4415 blade profile.
Keywords: wind turbine blade, BEM theory, bond graph, dynamic model

1. INTRODUCTION
The design of the wind turbine system (sizing,
control) according to specifications in terms of
quality of produced electricity, maximization of
power capture, regulation of power to the turbines
rated output, stability and safety, needs the use of
performing simulation tools able to change from
simple static calculations assuming a constant wind
to dynamic simulation that, from the unsteady
aerodynamic loads, models the aero elastic response
of the entire wind turbine system, including blades,
tower, drive train, rotor and control system.
However, even the most recent papers in the
literature dealing with control strategy for wind
turbines (as examples (Boukhezzar et al, 2007),
(Fernandez et al, 2008), (Balas et al 2003) take as
model of the wind turbine only a mechanical power
generator extracting power Pwt from the wind as
1 2. PROBLEM STATEMENT
Pwt = R 2 C p ( , )V03 (1)
2 In recent years, progress has been made both in
R Europe and the United States in developing and
where is defined as = r (2)
V0
Such approaches give simplified representation
models well adapted for control designing, but
they cannot be used as knowledge models or
virtual test benches allowing the validation of the
control strategies in terms of stability and reliability
of the power system with high wind power
penetration, vibration analysis, or simulation of
faulty behaviours of the wind turbine system.
The objective of the present paper is to propose a
bond graph model of a wind turbine blade.
The paper is organized in five sections.
In the second section, the main simulation codes for
wind turbines are briefly presented and the
objectives of the proposed modelling approach are
defined.
Then the building of the bond graph model of a
wind turbine blade is described. A generic bond
graph sub model is proposed for a section of a
blade; the complete BG model of the blade is
obtained by connecting the sub models.
In the next section, the aerodynamic load for the
blade is calculated using Blade Element Momentum
(BEM) theory.
Then simulations are performed using 20Sim
software (Controllab Product), with data from
NACA 4415 profile. Several results show the
validity of the proposed model.
validating simulation codes for predicting wind
turbine loads and response (BLADED (Bossanyi,
1997), ADAMS adapted for wind turbine use
(Elliott and Wright, 1994) (Bir, 1999), FAST-AD
(Wilson et al., 1999), subroutine package AeroDyn
used in ADAMS and FAST-AD (Hansen et al,
1998)).
These design tools have been developed as
specialized tools dealing with only structural part or
aerodynamic part. So, it is necessary to move
towards a more unified approach for modelling
wind turbine systems as a whole for understanding,
analyzing and hence designing this The stiffness matrix is given by (2) in terms of
multidisciplinary system. flexural rigidity EI and element length l i ,
Bond graph methodology allows its further
integration to bond graph model of drive train,
tower and generator providing a common platform
to access the whole turbine system and to deal with
control problems (Dauphin-Tanguy et al, 1999).

3. BOND GRAPH MODEL OF THE WIND

TURBINE BLADE
In this paper structural model of wind turbine blade
is prepared based on Rayleigh beam model
(Mukherjee et al, 2000). The blade can be assumed
as a twisted beam and is divided into finite number
of elements (figure 1) where on each element
aerodynamic forces due to the wind are acting.
Fig.1. Turbine Blade with space reticulation
(2)
E is Young modulus of the material of the blade; Ii
second moment of area about the axis of deflection.
A bond graph model of the beam element is created
by lumping the element inertias at the centre of
gravity of the element and appending them to the 1-
junctions representing displacements and rotations
at the centre of gravity of the element (fig.4.). In
this model the C-field represents the stiffness
matrix and the R-field the structural damping
matrix between the centres of gravity of adjacent
elements (Hansen 2007). is the coefficient of

The axial extension deformation and the Coriolis

structural damping. [Ri ] = [K i ] (3)
forces effect are not considered in the present AerodynamicForce I Mi

model. Self weight of turbine blade is neglected

while simulations. A beam element with shear
forces and moments acting on it in Newtonian End1Translation 1 C
Ki
1 End2Translation

convention is shown in Figure 2.

F1i , y1i F2 i , y 2
End1Rotation 1 R 1 End2Rotation
Ri

I Ji

M 1i , 1i EI M 2i , 2i Fig.4. BG model of a section of the turbine blade

with aerodynamic forces
l
Fig.2. Beam ith element with generalized forces and Only the aerodynamic force corresponding to the
displacements differential rotor torque is applied at the centre of
gravity of each section (fig.4.). It depends on the
The stiffness of the beam element relates the blade geometry and the flow stream properties. Its
generalized Newtonian forces to the generalized value has to be calculated for each section.
displacements at the ends of the element as given Cascading of several element bond graph section
by (1). sub models leads to the space reticulated beam
F1 y1 model as shown for three sections in Fig 5.
M (1)
1 = [K ] 1 Sf 1
F2 i
y2 V0

M 2 i 2 i
A cantilever beam with reticulation and lumping of pitch
GY GY GY

inertia is shown in Figure 3.

GY1 GY2 GY3

l l l 1
Sf 1 Section1 Section2 Section3
Vbound
1

Sf 1 1 e 1
M1, M2, M3, wr Se
Se
I
Jwhole Torque

Fig.3. Cantilever beam with reticulation and inertia

lumping (3 elements) Fig.5. BG model turbine blade 3 sections
Sf: V0 and Sf: r represent the wind mean The relationships between the lift and drag forces
velocity and the angular velocity of the rotor, Sf: and the normal and tangential forces are purely
Vbound and Se define the boundary conditions. vectorial as determined by the angle of attack of the
Gyrators are used to calculate the aerodynamic incoming flow stream, i , defined as the angle
forces acting on each section. The calculus of the between the incoming flow stream and the chord
different GY moduli will be developed in the next line of the airfoil at the ith section (fig.7.).
section. Gyrators have two input modulating
signals; one is the pitch of the blade, which is a
control variable, and the other is the angular
velocity of the rotor Jwhole is the rigid body inertia
of the whole blade. It is in derivative causality.

4. DETERMINATION OF THE AERODYNAMIC

LOAD
The aerodynamics of a wind turbine is highly Fig.7. Velocities at rotor plane
complex and non linear. The true fluid flow passing
around and through a wind turbine is governed by
the first principles- based Navier-Stokes equations The velocity of the incoming flow stream relative to
but too complex to use in modelling. As an the ith blade element V reli is thus obtained as
alternative, most wind turbine designers and V0 (1 a i )
analysts have opted to develop models that Vreli = (4)
incorporate two-dimensional airfoil analysis sin i
techniques and quasi-steady flow fields which have V (1 a i )
resulted in most common theory known as Blade with i = tan 1 0
(5)
r ri (1 + a i )
Element Momentum (BEM) theory.
A lot of literature exists on the explanation and r ri
The local speed ratio is defined as ri = (6).
presentation BEM theory (Hansen (2006), Moriarty V0
and Hansen (2005)). BEM theory uses several The differential rotor thrust dTi and the differential
annular stream tube control volumes. At the rotor
plane, the boundaries of these control volumes rotor torque dQi are deduced from lift and drag
effectively split the blade into a number N of forces in axial and tangential direction as
distinct elements (or sections), each of length 1 2
dTi = V reli (C li cos i + C di sin i )c i l i (7)
l i = Rr / N with R r the radius of the rotor (which 2
is also the length of the blade). At each element, 1 2
dQi = V reli (C li sin i C di cos i )c i ri l i (8)
blade geometry and flow stream properties can be 2
related to a differential rotor thrust dTi and a C li and C di are lift and drag dimensionless
differential rotor torque dQi, if some assumptions
are made as: there is no interaction between each coefficients functions of the angle of attack i .
blade element and the forces exerted on the blade
elements by the flow stream are determined solely Only dQi has to be considered here because we do
by the two-dimensional lift and drag characteristics not take into account the deflections in axial
of the blade element airfoil shape and orientation direction induced by the thrust dTi since it will not
relative to the incoming flow.
affect the dynamic torque.
The local load forces on the blade are L and D the
lift and drag forces, and R the normal force (fig.6.). The axial and tangential induction factor are then
calculated using
1
4 sin 2 i
a i = 1 + (9)
i (C li cos i + C di sin i

1
4 sin 2 i
a i = 1 + (10)
i (C li sin i C di sin i

The aerodynamic force Fi (deduced from the

differential torque by dividing by ri ) applied on the
ith, i =1,.., N element, is expressed as:
Fig.6.The local loads on a blade
 dQi 1 (1 ai ) 2
Fi = = V02 (Cli sin i Cdi cosi )ci liV0
r 2 sin 2 i
(11)
or
1 (1 a i ) 2
Fi = V0 (C li sin i C di cos i )c i l i V0
2 2
i
1444sin444 442444444443
GYi
(12)
where GYi denotes le modulus of the Gyrator
element associated with the ith section.
Fig.9. Coefficients of drag and of lift vs angle of
5. SIMULATION AND RESULTS attack (NACA 4415)
We consider the case of a wind turbine in which the
rotor is in upwind configuration. The connection Table 1. Case of three sections
between the blade and the hub is assumed to be
rigid ( Vbound = 0 ) and the flow unsteady. The blade Section 1 Section 2 Section 3
has only one rotational degree of freedom. r1 = 3.9 m r2 = 11.7 m r3 = 19 .5 m
For simulation, NACA 4415 blade data are used c1 = 1.56 m c 2 = 1.459 m c3 = 0.8315 m
(Younsy et al (2001)):
R = 23.2m, E =1.71e10, B =3, = 1.225, = 0.01, t1 =7.65 deg t 2 = 4.53 deg t 3 = 0.72 deg
Jwhole = 1000 kg m2. M1 = 928 kg M2 = 560 kg M3 = 207 kg
J1= 26 kg m2 J2= 13.78 kg m2 J3 = 1.66 kg m2
Distribution of twist and chord along the length of I0=0 I1 = 0.00077 I2 = 0.0005
the blade is shown in Fig.8. I1= 0.00077 I2 = 0.0005 I3 = 6.2 10-5

velocities, angles of attack and thrust coefficients

for each section along the span of the blade. It is
composed of the following steps:

- 1rst step: estimate the initial value of the axial

induction factor a i .
One efficient way is to assume that the inflow angle
. is small: sin i i , the tangential induction a i is
zero, the drag coefficient C d is zero, the lift
coefficient is C li = 2 i where the angle of attack
is i = i i with i the sum of the twist angle
ti depending on the position of the centre of
gravity of the ith section (fig.8) and the pitch
angle which is an input control variable.

After some rearranging, it results in

Fig.8. Twist and chord along the blade 1 2
ai = 2 + ri i 4 4 ri i + ri i (8 i + i
The drag and lift coefficients, obtained from look- 4
up tables are shown fig.9.
r ri
with ri = , and i = Bc i 2r calculated
Table 1 regroups the numerical values used in the V0
simulation program in the case of three sections of using chord values ci depending on ri (fig.8)
length l i = 23.2 / 3 = 7.8 m.
- 2nd step: estimate the inflow angle i using an
Calculation of the GYi modulus (programmed
initial zero value for the tangent induction a i .
inside the code of each GY- element):
An iteration procedure, proposed in (Moriarty and V 0 (1 a i )
i = tan 1

Hansen, 2005)), is used to calculate the induced r ri
- 3rd step: determine the thrust coefficient for the
element using
(1 a i ) 2 (C li cos i + C di sin i )
C Ti = i ,
sin 2 i
where C di ( i ) and C li ( i ) are determined using
look-up tables (fig.9)
Fig.11. Cp for different numbers of sections
- 4th step: calculate a i
1
It is clear from Fig.11that to estimate the accurate
4 sin 2 i torque 10 to 14 sections are enough. Following
a i = 1 +
results are produced using ten sections.
i (C li cos i + C di sin i

1
4 sin 2 i
a i = 1 +
i (C li sin i C di sin i

This process is iterated until the values of induction
factors and inflow angle have converged to their
final values. It is repeated for each GYi.

Power coefficient
The non dimensional power coefficient
P
Cp = represents the fraction of
1
R 2V03
2
available power in the wind P that is extracted by
the turbine. It can be rewritten as C p = 4a (1 a) 2 Fig.12. Cp(,)
because the power P extracted from the wind by the Fig.12 shows the variations of Cp in function of
rotor is the product of the force different pitch angles and tip-speed ratios.
1
T = R 2V02 4a(1 a ) and the wind velocity at
2 Because of its modularity, the bond graph model
the rotor plane Vrel = V 0 (1 a ) . can very easily be modified to take into account
The theoretical maximum power coefficient from some previously neglected phenomena.
an idealized rotor C p max , known as Betz limit, can
A different reticulation of the blade could be made,
be found by setting the derivative of C p with by using different lengths for the elements,
respect to a equal to zero, which leads to : a = 1 / 3 according to their position in the blade.
and C p max = 0.596 , which corresponds to the Wind field simulation should be improved taking
into the account wind shear and stochastic
maximum possible efficiency for an idealized wind component of wind. It will induce changes in the
turbine of 59.6%. bond graph model only at the upper part.
In practice, three effects prohibit a real wind turbine
from achieving this efficiency: rotation of the wake Further each element should be modified to include
caused by the spinning rotor, finite number of more degrees of freedom to take into account
blades, viscid flow causing nonzero aerodynamic different forces such as centrifugal force, inertial
drag. The simulation program calculates C p as loading and aerodynamic damping to improve the
Pmech OutputTorque * r results. Centrifugal stiffening effect should be
Cp = = incorporated only by modifying the stiffness matrix
Pair 1
R 2V03 Ki which will have a form different for each section
2 of the blade.
where OutputTorq ue is three times the steady state
torque obtained from the BG model of single blade. Blade model should be integrated in a global BG
model including models of the rotor, shaft, gearbox,
Fig.11 shows the values of C p predicted using and mechanical part of the electric generator, only
by replacing Sf: r as shown fig.13 (well-known
different numbers of sections.
two-masses model).
 shaft Boukhezzar B., Lupu, L., Siguerdidjane, H., Hand, M.,
Kls C R bls
(2007). Multivariable control strategy for variable
Sf V0
speed, variable pitch wind turbines, Renewable
I Jg 1 I Jr
Energy, Vol 32, pp 1273-1287
Controllab Product, 20 Sim software, NL
Twt Dauphin-Tanguy, G., Sueur, C., Rahmani, A., (1999).
Se 0 1 TF 0 1 WindTurb
Tem wg
n_gb
wr Bond graph aided design of controlled systems, J.
pitch
Simulation Practice and Theory, Elsevier Pub,
R gearbox R Dekker Ed, volume 7, Issue 5-6
bg br
Elliott, A., Wright, S., (1994). ADAMS/WT: an Industry-
gene rotor
specific modeling interactive interface for wind
turbine analysis, Wind Energy, ASME, pp 111-122
Fig.13. Example of extension of the model
Fernandez, L.M., Garcia, C.A, Jurado, F., (2008).
Comparative study on the performance of control
6. CONCLUSIONS systems for doubly fed induction generator (DFIG)
Wind turbine blade has been analyzed using bond wind turbines operating with power regulation,
Energy, vol 33, pp 1438-1452
graphs taking into consideration the flexibility, to
Hansen, A.C, Laino, D.J., (1998). Users guide to the
predict the torque output for different wind flow Wind Turbine Dynamics Computer Programs
conditions and rotor angular speed. The value for YawDyn and AeroDyn for ADAMS, University of
torque obtained and the pattern of power coefficient Utah
curve indicate the validity of the proposed BG Hansen, M.O.L., Srensen, J.N., Voutsinas, S., Srensen,
model of wind turbine. The model has been N., Madsen, H.A. (2006). State of the art in wind
validated with the available data (NACA 4415), but turbine aerodynamics and aeroelasticity. 4,
the proposed blade model is a generic model and Progress in Aerospace Sciences, Vol. 42, pp. 285-
can be used with any profile of blade. Power 330.
Hansen, M.O.L. (2007). Aerodynamics of wind turbine.
Coefficient curve can be predicted and hence the s.l. : Earthscan
optimum conditions to operate. Moriarty, P.J., Hansen, A.C. (2005). AeroDyn Theory
REFERENCES Manual. National Renewable Energy Lab., Golden,
CO (US)
Balas, M.J., Wright, A., Hand, M.,Stol, K. (2003). Mukherjee, A., Karmakar, R., Samantaray, A.K.( 2000).
Dynamics and control of horizontal axis wind Modelling and Simulation of Engineering Systems
turbines. Proceedings of the American Control through Bond Graphs. s.l. : Narosa Publishing
Conference. Vol. 5, pp. 3781- 3793. House,
Bir, G.S, (1999). Code development for control design Wilson, R.E, Walke,r S.N, Heh, P., (1999). Technical and
applications, Proceedings of the 37th AIAA Users Manual for the FAST-AD Advanced
Aerospace Sciences Meeting and Exhibit, Reno, Dynamics Code, Oregon State University
USA Younsi, R., El-Batanony, I., Tritsch, J.B., Naji, H.,
Bossanyi, E.A., (1997). BLADED for Windows Theory Landjerit, B. (2001). Dynamic study of a wind
Manuel, Garrad Hassan and partners Limited, turbine blade with horizontal axis. 2, European
Bristol England journal of mechanics. A. Solids, Vol. 20, pp. 241-
252.
n gb gearbox ratio ri position of the centre of
Appendix NOMENCLATURE Rr rotor radius (blade gravity, m
length), m
i wind angle of attack,
pitch angle, deg For ith section
air density, kg m -3 deg
B number of blades ai axial induction factor
a i tangential induction
i wind inflow angle, deg
br rotor ext damping, Nm Tem generator
factor (electromagn) torque, Nm ti blade twist angle, deg
rad-1 s
bls low speed shaft ci cord, m Tr rotor torque, Nm
damping, Nm rad-1 s C di drag coefficient Twt wind turbine torque,
C p power coefficient C Li lift coefficient Nm
E Young modulus of the CTi coefficient of thrust V0 wind speed, ms-1
blade material ,Pa Ii moment of inertia, kg m2 g generator speed, rad s-1
G modulus of elasticity for
shear
i local solidity r rotor speed, rad s-1
FT thrust force, N dTi differential thrust y& velocity in the y-axis
J g generator inertia, kg m 2 force, N
dQi differential rotor
J r rotor inertia, kg m2 torque , Nm
K ls low speed shaft ri local speed ratio
stiffness, Nm rad-1 Fi aerodynamic force, N
tip speed ratio