10 1016@j Isatra 2016 07 009

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Mechatronic modeling of a 750 kW fixed-speed wind energy conversion


system using the Bond Graph Approach
Zakaria Khaouch, Mustapha Zekraoui, Jamaa Bengourram, Nourreeddine Kouider,
Mustapha Mabrouki n
Industrial Engineering Laboratory, Faculty of Science and Technology, Sultan Moulay Slimane University, Beni Mellal, Morocco

art ic l e i nf o a b s t r a c t

Article history: In this paper, we would like to focus on modeling main parts of the wind turbines (blades, gearbox,
Received 5 February 2016 tower, generator and pitching system) from a mechatronics viewpoint using the Bond-Graph Approach
Received in revised form (BGA). Then, these parts are combined together in order to simulate the complete system. Moreover, the
1 July 2016
real dynamic behavior of the wind turbine is taken into account and with the new model; final load
Accepted 25 July 2016
simulation is more realistic offering benefits and reliable system performance. This model can be used to
This paper was recommended for publica-
tion by Dr. Jeff Pieper develop control algorithms to reduce fatigue loads and enhance power production. Different simulations
are carried-out in order to validate the proposed wind turbine model, using real data provided in the
open literature (blade profile and gearbox parameters for a 750 kW wind turbine).
Keywords: & 2016 ISA. Published by Elsevier Ltd. All rights reserved.
Mechatronic modeling
Interdisciplinary systems
Bond graph
Wind turbine
Blades
Gearbox
Tower
Generator
Pitching system

1. Introduction should be usable for detailed control system design, which will be
necessary as the design goes deeper. Indeed, linear simplified
Many engineering activities, including mechatronic design, models, e.g., a simple aerodynamic, mechanical or electrical
require that a multi-domain or multi-physics system and its con- models, are not sufficient to take into account the necessary cou-
trol system should be designed as an integrated system [1]. pling effects among the components.
Mechatronics is “a technology which combines mechanics with For the combined analysis of the aerodynamic, structural and
electronics and information technology to form both functional control generating systems, or the mechanical, control and aero-
interaction and spatial integration in components, modules, pro- dynamic subsystems many works either use a very simple
ducts and systems” [2]. mechanical model with a detailed electrical model, or a compli-
Wind turbine is a complex mechatronic system, in which dif- cated mechanical model with a simple control system model [3,4].
ferent technical areas are involved (mechanics, aeronautics, elec- Hence, the interactions and the dynamics coupling cannot be
trical, among others). There is no doubt that a mechatronic accurately predicted. In [5] four different models of wind turbines
are introduced. On one hand, two static models – the simple static
approach is essential in the field of wind turbine design. This
model and the static mechatronic model – represent two different
approach implies that a mechanic, aerodynamic, electric sub-
instances of a simplified behavior. On the other hand, two dynamic
systems and eventually their control subsystem should be
models – the mechanical model and the dynamic mechatronic
designed as an integrated system. This integration is important for
model – describe the dynamic behavior of a wind turbine in more
a more accurate evaluation of the extreme loads and the fatigue
detail.
life, and this might reduce the failure rate in the design stage.
In [6] a control-generator-structure coupled with analysis in
However, the aerodynamic, mechanical and electrical models wind turbines using a monolithic modeling and simulation is
proposed. The integrated system models were developed on
n
Corresponding author. Samcef for Wind Turbines (S4WT) through a nonlinear finite ele-
E-mail address: mus_mabrouki@yahoo.com (M. Mabrouki). ment method (FEM) formalism. Another contribution [7,8] reviews

http://dx.doi.org/10.1016/j.isatra.2016.07.009
0019-0578/& 2016 ISA. Published by Elsevier Ltd. All rights reserved.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
2 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

recent research of technical issues on the development of wind functional blocks, i.e., the drive-train and the support structure. The
farms. These contributions highlight the importance of having a drive-train transfers the aerodynamic torque on the blades to the
mechatronic model of the wind turbine, but the models lack some generator shaft. It encompasses the rotor, the transmission and the
detail. In order to analyze the system in the same reference frame, mechanical parts of the generator. The structure is made up of the
there are several methods and tools for the modeling and the tower and the blades. The electrical subsystem describes the con-
simulation of physical systems and their controllers, with para- version of mechanical power at the generator shaft into electricity.
meters directly related to physical components. Moreover, it is Finally, the actuator subsystem models the pitch servo behavior.
desired that these (sub-) models should be reusable. Common In order to analyze the system in the same reference frame, the
Block-Diagram (CBD) or Equation-Based Simulation (EBS) packa- Bond-Graph Approach (BGA) [10,11] is used.
ges do not easily support these features. The Energy-Based
Approach (EBA) for modeling physical systems allows the con- 2.1. Fundamental concepts of bond graphs
struction of reusable and easily extendable models.
In this work, a fixed speed wind turbine mechatronic model Bond graph is a graphical way of modeling physical systems. All
has been developed. The main components of the system are these physical systems have in common the conservation laws for
modeled in detail using the Bond-Graph Approach (BGA). The mass and energy. Bond graph, originated by Paynter [10] in 1961,
implementation of the complete model has been carried out by deals with the conservation of energy. This gives a unified
means of the 20-Sim simulation program (a demo version is freely approach to modeling physical systems. Further follows a short
available on the Internet). Parameters of a 750 kW-power wind introduction to this modeling tool, more information can be found
turbine were taken from NACA 4415 profile [9]. Verification of in [9,10]. The bond graph based modeling has several advantages
results is reported. Conclusions of the conducted research inves- over conventional simulation methods as follows:
tigation are drawn.
1. providing a visual representation of the design;
2. controlling the consistency of the topological settings of the
2. Mechatronic modeling of the wind energy conversion design;
systems 3. providing the hierarchical modeling of designs;
4. extracting the system equations symbolically in a structured way.
In the next sections, the mechatronic modeling of the main
components of a WECS will be presented. Our first aim is to show Within physical systems, energy is transported from one item
the complete wind turbine in a unified framework. The WECS is to another. This energy is either stored or converted to other
organized into four main areas, namely the aerodynamic, mechan- forms. But the important thing is that it does not dissipate. If the
ical, electrical, and pitch servo subsystems. The main components of energy is changing in one place, it also changes in an opposite way
a WECS are rotor, transmission system and power generator unit. at another location. The definition of power is the change in
The rotor comprises the blades where the aerodynamic conversion energy ðEÞ with respect to time:
takes place, the hub that links the blades to the transmission and the
pitch servos, which are located inside the hub that rotates the blades dE
P¼ ð1Þ
around their longitudinal axes. The transmission system transmits dt
the mechanical power from the rotor to the electric machine. It This power is transferred between the different parts in Bond
comprises the low- and high-speed shafts, the gearbox and the Graph model with the use of power bonds, see Fig. 2. Power can be
brakes. The gearbox increases the rotor speed to values more sui- expressed as the product of an effort and a flow variable, thus the
table for driving the generator, typically from 20–50 rpm to 1000– general expression:
1500 rpm. The electric generator is the device that converts
P ¼ eðtÞf ðtÞ ð2Þ
mechanical power into electricity. Its electric terminals are con-
nected to the utility network. In the case of variable-speed WECS, an The symbols eðtÞ and f ðtÞ are used to denote effect and flow
electronic converter is used as interface between the AC grid and the quantities as functions of time. Table 1 shows that the effort and
stator or rotor windings. flow quantities can be in some familiar domains.
A mechatronic model, that takes into consideration the dynamic
behavior of the entire WECS, can be structured as several inter- 2.1.1. System elements
connected subsystem models as shown in Fig. 1. The aerodynamic In Bond Graph modeling there are a total amount of nine
subsystem describes the transformation of the three-dimensional different elements. We will also here introduce the causality
wind speed field into forces on the blades that generate the rota- assignments, but first we have to explore the cause and effect for
tional motion. The mechanical subsystem can be divided into two each of the basic bond graph elements.

Fig. 1. Subsystem-level block diagram of a WECS.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3

Fig. 2. Power bond with effort and flow.

Table 1
Effort and flow variables in different domains.
Fig. 4. 1-junction.
Domain Effort, eðtÞ Flow, f ðtÞ

Mechanical translation Force, F(t) Velocity, v(t)


Mechanical rotation Toque, τðtÞ Angular velocity, ωðtÞ
Hydraulic Pressure, P(t) Volume flow rate, Q(t)
Electric Voltage, U(t) Current, i(t)

Fig. 5. Effort and flow source with their causality assignment.

Fig. 3. 0-junction.

Fig. 6. Example of a compliance element with integral causality.


Junctions: There are two different types of junctions that con-
nect the different parts in a Bond Graph model, the 0-junction and
the 1-junction. The 0-junction is an effort equalizing connection,
see Fig. 3 and its corresponding equations in (3) and (4). Since the
efforts are the same, only one bond can decide what it is. The
1-junction is a flow equalizing connection, see Fig. 4 and its cor-
Fig. 7. Example of an inertia element with integral causality.
responding equations in (5) and (6). Since the flows are the same,
only one bond can decide what it is. Which bond decides the flow
and which one decides the effort is indicated with the vertical
causality stroke. If the vertical line is closest to the junction, then
this element decides the effort, furthest away from the junction
decides the flow.

e1 ¼ e2 ¼ e3 ð3Þ

f 3 ¼ f 1 þf 2 ð4Þ

Fig. 8. Example of resistive element.


f1 ¼ f2 ¼f3 ð5Þ
preferred case is seen in Fig. 7 and its corresponding equation in (8):
e3 ¼ e1 þ e2 ð6Þ Z
1 p
f¼ edt ¼ ð8Þ
Source element: We can divide the source elements into two I I
different kinds, effort- and flow-source. The effort source gives an The variable p is called the generalized momentum. For
effort into the system, then it is up to the system to decide the example, this can be rotor inertia times rotor velocity in a wind
flow. This is what is meant with cause and effect, and its vice versa turbine.
for the flow source. Fig. 5 shows how the causality is indicated on Resistive element: It is a bit more freedom when it comes to the
the graphical elements. For the source elements these causality causality assignment for the R-element. Its equation does not
assignments are fixed. include any dynamics, it is only an algebraic expression. The two
Compliance element: The causality assignment for the C-element causality choices are shown in Fig. 8 and its corresponding equa-
has two possibilities, but one is preferred in contrast to the other. tion in (9):
This is discussed at the end of this section. The preferred case is e
seen in Fig. 6 and its corresponding equation in (7). We see from f¼ ; or e ¼ Rf ð9Þ
R
both the equation and the figure that flow is given to the element/
Transformer element: The transformer element can work in two
equation and it gives the effort in return:
ways; either it transforms a flow into another flow or it transforms
Z
1 q an effort into another effort. Fig. 9 corresponds to (10), where m is
e¼ fdt ¼ ð7Þ
C C the transformation ratio. For example, this can represent a
mechanical gearing or an electric transformer.
The variable q is called the generalized displacement. For
example, this can be rotational position of the rotor in a wind f2 e1
f1 ¼ ; e2 ¼ or f 2 ¼ mf 1 ; e1 ¼ me2 ð10Þ
turbine. m m
Inertia element: There are two choices for the causality assignment Gyrator element: The gyrator can also work in two ways; either
for the I-element, also here one is preferred in contrast to other. The it transforms a flow into an effort or it transforms an effort into a

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
4 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 9. Example of the two transformers.

Fig. 10. Example of the two gyrators.

flow. Fig. 10 corresponds to (11), where r is the gyrator ratio.


e2 e1
f1 ¼ ; f2 ¼ or e1 ¼ rf 2 ; e2 ¼ rf 1 ð11Þ
r r
By using bond graph as the modeling tool we get a good
overview of the models' physical structure and we can do simu-
lations in one step, instead of first deriving the equations and then
drawing the block diagram.

2.2. Mechanical subsystem

A horizontal-axis wind turbine is a complex mechanical system


that consists of interacting devices with some degree of flexibility.
Like any flexible structures, such system exhibits many vibration
modes. Some oscillatory movements inherent to these modes are
illustrated in Fig. 11 [12]. The existence of these vibration modes
demands a careful design of the wind turbine and controller.
For the WECS modeling, the most involved part is probably the
mechanical subsystem. The complexity arises from the interaction
of three flexible structures: the blade, the drive train and the
tower. Each of these structures is fixed to a reference frame that Fig. 11. Mode shapes for horizontal-axis wind turbines.
rotates with respect to the other. This leads to high-order non-
linear dynamic models. Moreover, most of the forces applied to the
structures come from a three-dimensional wind field.
The mechanical subsystem model (blade, tower, and gearbox),
that takes into account the dynamic behavior of a horizontal-axis
wind turbine, can be undertaken by various analysis methods
[13,14]. In this section, a Bond Graph model of the mechanical
subsystem is proposed. It considers the blade and the tower as a
Rayleigh beam composed of a number of sections submitted to
aerodynamic forces, and the gearbox as a multi-body dynamic
system. Subsequently, the model can be used to develop pitch
control algorithms, which should minimize the vibrations prob-
ability and also increase wind turbine lifetime by mastering the
load effects and the costs reduction without sacrificing the aero-
dynamic output.
Fig. 12. Turbine blade with space reticulation (a). Dynamic model of blade (b).

2.2.1. Bond graph model of the wind turbine blade


In this section, we study the structural dynamic behavior of the they are integrated in our model. The dynamic model of the blade
wind blade. A blade Bond Graph model is developed. It consists of is shown in Fig. 12(b). For each element in the aerodynamic forces,
considering the blade as a three-dimensional beam composed of a we also add the deformation of the axial extension and the
number of sections (Fig. 12(a)). The sections vary according to pitching moment blade. Therefore, the blade is modeled as a
NACA 4415 aerofoil sections. In [15,16] only constant sections were three-dimensional beam.
used. Many studies on the behavior of the blade dynamic structure The stiffness of the beam element relates the generalized
were conducted. In [17], a structural model of the wind turbine Newtonian forces to the generalized displacements at the ends of
blade is proposed based upon the Rayleigh beam model. The the element as given by
2 3 2 3
blade – assumed as a twisted beam – is divided into three com- F yi yi
ponents, each of which is subject to the aerodynamic wind forces. 6 7 6 7
6 M zi 7 6 θzi 7
6 7 6 7
Nevertheless the deformation of the axial extension and the 6 F yi þ 1 7 ¼ ½K f yi 6 yi þ 1 7 ð12Þ
pitching moment applied to the center of gravity are taken into 4 5 4 5
consideration. These “gaps” are addressed in this paper where
M zi þ 1 θzi þ 1

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 5

2 3 2 3  
F zi zi GJ 1 1
6 M 7 6 θ 7 ½K t  ¼ ð17Þ
6 yi 7 6 yi 7 L 1 1
6 7 6 7
6 F zi þ 1 7 ¼ ½K f zi 6 zi þ 1 7 ð13Þ
4 5 4 5
M yi þ 1 θyi þ 1 where E is the Young modulus of the material, G is the Coulomb
modulus, I y;z the second moments of area about the axis of
" # " # deflection, J is the second polar moment and L is the length of the
M xi θxi finite element. The structural damping matrix is given by
¼ ½K t  ð14Þ
M xi þ 1 θxi þ 1 ½Ri  ¼ μ½K i , where the μ factor represents the coefficient of struc-
tural damping.
Fig. 13 shows the Bond Graph model of the blade.
where ½K fyi  is the flexural stiffness around y, ½K fzi  is the flexural
Fig. 13(a) represents the axial extension deformation of the blade,
stiffness around z and ½K t  is the flexural stiffness around x are
Fig. 13(b) the tangential extension deformation and Fig. 13(c) the
given by
torsional deformation of blade. For Fig. 13(a) and (b), two motions

2     3
12 6 li  1 þ li =2  12 6 li  1 þ li =2
6     7
EðI zi  1 þ I zi Þ 6
6 6 li  1 þ li =2 4ðli  1 þ li =2Þ2  6 li  1 þ li =2
 
 2ðli  1 þ li =2Þ2 7
 7
½K f yi  ¼ 6 7 ð15Þ
ðli  1 þ li =2Þ3 6  12  6 li  1 þ li =2 12  6 li  1 þ li =2 7
4     5
2
6 li  1 þ li =2 2ðli  1 þ li =2Þ  6 li  1 þ li =2 4ðli  1 þli =2Þ2

2    3
12  6 li  1 þ li =2  12  6 li  1 þ li =2
 6
6
    7
E I yi  1 þ I yi 6  6 li  1 þ li =2 4ðli  1 þ li =2Þ2 6 li  1 þ li =2 2ðli  1 þ li =2Þ2 7
½K f zi  ¼ 6     77 ð16Þ
ðli  1 þ li =2Þ3 6  12 6 li  1 þ li =2 12 6 li  1 þ li =2 7
4     5
2 2
 6 li  1 þ li =2 2ðli  1 þ li =2Þ 6 li  1 þ li =2 4ðli  1 þ li =2Þ

Fig. 13. Structural bond graph of blade, axial extension (a), tangential extension (b), torsional extension (c).

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
6 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

are shown: rotation at the top and translation at the bottom of the
figure. In Fig. 13(c) the model is represented by one rotation
motion. For each model, the displacements and rotations at the
center of gravity of each element are represented by 1-junction.
For each element aerodynamic forces are applied, meaning that
the effort is imposed. The stiffness and the structural damping
matrixes between the centers of gravity of adjacent elements are
represented using C-field and R-field elements, respectively.
Boundary conditions of the model for the tangential extension Fig. 15. Torque simulation of the blade.
deformation are represented by the Sf and Se sources. Connection
between the blade and the hub must be rigid, i.e. Sf ¼ 0, and the
blade has only one degree of freedom, so Se ¼ 0. Boundary con-
ditions for the axial extension deformation are represented by the
Se source. The blade has only one degree of freedom, then Se ¼ 0.
Connection between the blade and the hub as well as between the
hub and the tower are assumed to be rigid. Therefore, the move-
ment at the tower top is the same as the movement at the blades
bottom. This relationship between the blade and the tower can be
modeled by 1-junction. Boundary conditions for the torsional
deformation are represented by the Sf source which represents the
pitch actuator systems velocity. Dynamic equations and natural
frequencies of the blade can be directly obtained from the Bond
Graph model.
Three sources of effort, which represent the aerodynamic for-
ces, need to be calculated in the input wind. This process will be
treated in the aerodynamic subsystem section. Sub-models, as
shown in Fig. 14, can represent the model of the blade (Fig. 13).
To check the model validity, a simulation of the torque at the
rotor blade is shown in Fig. 15. Parameters are available in
Appendix A. A constant tangential aerodynamic force is applied at
Fig. 16. Bond Graph model of the hub.
each elements' center of gravity ðSe ¼ 100 NÞ. These forces are 
P
transformed into a torque at the blade rotor T ¼ 3i ¼ 1 ¼ F i r i .
This simulation shows that the torque in steady state is equal to
3510.25 N m. Based on the parameters in Appendix A, we have
T ¼100*3.9 þ100*11.7þ100*19.5 ¼3510 N m. We find the same
value of the torque. Therefore, the model works very well. The
vibrations presented in the model are due to the dynamic behavior
between the mesh stiffness and the momentum of inertias.

2.2.2. Bond Graph model of the hub


The hub model links the blades' tangential model to the
transmission, the blades' axial model to the tower and the blades'
pitching moment to the pitch servos system. The latter is located
inside the hub and rotates the blades around their longitudinal
axes. The Bond Graph model of the hub is shown in Fig. 16.
Fig. 16(a) sketches the connection between the tangential
model of the blades and the transmission. I-elements and
R-elements represent the inertia and main bearing of the hub
respectively. Fig. 16(b) shows the connection between the axial Fig. 17. Bond Graph sub-model of the hub.

Fig. 14. Tree sections bond graph blade.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 7

Fig. 18. Rayleigh beam model of the tower (a), analyzing model of the tower (b),
dynamic model of the tower (c).

model of the blades and the tower. Fig. 16(c) deals with the con-
nection between the pitching moment model and the pitch servos
system. Sub-models, as shown in Fig. 17, can represent the model
of the hub (Fig. 16).
Fig. 19. Bond Graph model of the tower.
2.2.3. Wind turbine tower model
Dynamic behavior of a wind turbine tower is subjected to with
2     3
random wind loads – which produces large vibrations that could, 12 6 li  1 þ li =2 12 6 li  1 þ li =2
6     7
on the long run, lead to fatigue and failure is examined [18]. EðI zi  1 þI zi Þ 6
6 6 li  1 þ li =2 4ðli  1 þ li =2Þ2
 
6 li  1 þ li =2 2ðli  1 þ li =2Þ2 7
 7
½K f yi  ¼ 36 7
Therefore, the control system should be designed in such a way to ðli  1 þli =2Þ 6 12 6 li  1 þ li =2 12  6 li  1 þ li =2 7
4     5
minimize the effects of external loads and eliminate vibrations of 6 li  1 þ li =2 2ðli  1 þ li =2Þ2 6 li  1 þ li =2 4ðli  1 þ li =2Þ2

the flexible tower. ð19Þ


The tower is modeled as a tubular Rayleigh beam with three
Where E is the Young modulus of the material, and Izi the
different sections (Fig. 18(b)). However, it can be divided into
second moment of area about the axis of deflection. The structural
more segments. The wind loads on the rotors are transferred to damping matrix is given by ½Ri  ¼ μ½K i , where the μ factor repre-
the tower as bending moments at the tower top. In the right- sents the structural damping coefficient. The Bond Graph model
handed Cartesian coordinate system shown in Fig. 18(a), the z- and submodel of the tower is shown in Fig. 19. We assume tower
axis is assumed to coincide with the elastic axis, as well as the movement only in the horizontal direction.
cross sections center of the thin-walled beam, the y-axis is in the Boundary conditions of the model are represented by the Sf
direction of wind. The model is constructed on the assumption source. The tower-base is fixed on the ground, i.e. Sf ¼ 0. Con-
that every segment has uniform cross-sectional properties nection between the tower-top and the hub for the axial defor-
(Fig. 18(b)), with distributed parameters by the Bond Graph mation is ensured by the 1-junction. It is assumed that the tower
Approach. This model can be used to develop pitch control movement will influence the wind speed of the aerodynamic
algorithms, which would minimize the probability of extremely subsystem input.
high levels of tower vibrations.
2.2.4. Bond Graph model of the wind turbine gearbox
In order to evaluate the structures' vibration properties, the
Gearbox is used in the WECSs to increase the angular velocity.
wind turbine tower is modeled as an elastic tubular beam with a
It is the most important part of a wind turbine, where most of the
rigid rotor on the top and a fixed-base on the ground. The dynamic
faults occur. In [17] it is reported that a wind turbine has around
model of the tower is shown in Fig. 18(c). 20 years of life span, but normally the gearbox needs to be
The stiffness of the beam element relates the generalized replaced every 5 years. For this reason, big manufacturing com-
Newtonian forces to the generalized displacements at the ends of panies of wind turbines intend to use a direct drive wind turbine
the element as given by without a gearbox.
2 3 2 3 Nowadays, most of the installed wind turbines are equipped
F yi yi with a gearbox. Actually, most of the 270 GW of wind turbine power
6 7 6 7
6 M zi 7 6 θ 7 is installed around the world with this configuration [19]. Based on
6 7 ¼ ½K f yi 6 zi 7 ð18Þ
6 F yi þ 1 7 6 yi þ 1 7 this context, a model of the gearbox is presented in this paper. In
4 5 4 5
M zi þ 1 θzi þ 1 terms of a Bond Graph Approach, the Bond Graph model of a
gearbox has been addressed in several works [20,21]. The gearbox

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
8 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

can be considered as a simple model in which the conversion ratio or converge. A basic kinematic model of a single planetary gear is
(Np =N g , where Np is the teeth number of pinion and Ng the teeth shown in Fig. 21.
number of gears) can be introduced directly in a transformer (TF According to the kinematic relation of the components shown
element of bond graph). However this model does not describe the in Fig. 21, the translation velocities in contact point A and B can be
dynamic behavior of the gearbox. Therefore, a more accurate model described as follows:
needs to be developed. This model will be used in a future work as a V pA ¼  ωp r p þ ωc r r ð20Þ
base, in order to formulate the law control to compensate any
V rA ¼ ωr r r ð21Þ
perturbation due to a gearbox default.
Wind turbines gearbox consists of three stages. The first one is V pB ¼ ωp r p þ ωc r s ð22Þ
a planetary gear and the two others are parallel gears. The gearbox
scheme is shown in Fig. 20. V sB ¼ ωs r s ð23Þ
In [22] a detailed gearbox Bond Graph model is proposed. The
where ωj and rj stand for the rotating velocity and the base circle
planetary gear model is presented in detail but the parallel stage is radius ð j ¼ r; s; pÞ respectively. A is the meshing point of the ring
modeled by a simple TF element. This resource is taken as a gear and the planet, and B is the meshing point of the sun gear and
reference to develop the wind turbine gearbox model, to which we the planet. Vpm is the velocity of planets in point m along the
add a detailed model of parallel gear stage. Moreover, the dampers tangent direction, where m ¼ A; B. Similarly, VrA and VsB are the
of all components of the system are also considered. linear velocities of ring gear in A and sun gear in B.
For a planetary gear, several identical planet ðpÞ gears are in In a gear systems modeling, the gear mesh interfaces are
mesh with the sun gear and the ring gear ðrÞ, and the carrier ðcÞ usually considered as spring–damper systems [23,24]. The spring
holds all of the planet gears. Planet gears rotate in two kinematic stiffness (also called the mesh stiffness) is one of the major sources
modes called revolution around the sun gear and autorotation of gear vibration. The dynamic model of planetary gearbox is
around its own axis. Power flow into the planetary gear will split shown in Fig. 22(a). Fig. 22(b) depicts the Bond Graph model of the
gearbox.
The planet gears represented by momentum of inertia Jpi are
connected to each other through a common carrier. Therefore,
they get the same rotational speed imposed by the carrier.
1-Junction represents the rotation speed of the carrier ðcÞ, and
I-elements J c denote the rotary inertia of the carrier. These planets
are bounded to sun and ring gears by the mesh stiffness dumping
K sp ; Bsp and K rp ; Brp respectively. Jr is the ring gears momentum of
inertia. The connection between the ring and the hub is carried
out by an elastic joint modeled through a spring–damper. In this
joint, the vibration signal of the gear can be obtained. In the
model, Z i ði ¼ p; s; r Þ represents the teeth number of each gear.
The flow 1-junction represents the angular velocity of planets,
carrier, ring and sun gear, respectively. They are related to each
other through TF-elements. The zero junctions, between the
transformers, represent the effort variable that the planet moves
Fig. 20. Gearbox scheme. in tangential direction.
The parallel gear stage of the wind turbine gearbox is com-
posed of two stages (Fig. 23(a)). The dynamic model and the Bond
Graph model are shown in Fig. 23(b).
The complete Bond Graph model of a wind turbine gearbox and
its sub-models are depicted in Fig. 24.
Fig. 25 represents a simulation of the gearbox behaviors.
A constant torque is applied at the input of the gearbox
(Se¼ 60 N m). This output torque decreases with a ratio of 60. The
same outcome is observed in the angular velocity curves. For a
constant angular velocity at the gearbox output, the gearbox input
is reduced with a ratio of 60. From the teeth number of each gear
(Appendix A), we can calculate the gearbox ratio, which is equal to
60. This comes in line with the predictions of the gears Bond
Graph model.
The system undergoes the same vibration patterns; which are
due to the dynamics involved between the mesh stiffness and the
momentum of inertias.
The model of the mechanical subsystem of WECS is shown in
Fig. 26, where all parts of the mechanical subsystems are related
by the means of the hub.

2.3. Aerodynamic subsystem

The aerodynamic subsystem transforms the three-dimensional


wind field into lumped forces acting on the rotor blades. As observed
Fig. 21. Kinematic model of planetary gearbox. in the block diagram of Fig. 1, the inputs to the aerodynamic

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 9

Fig. 22. Physical model for dynamic meshing problem (a), Bond Graph model of planetary gearbox (b).

Fig. 23. Dynamic model of parallel gear (a), Bond Graph model of parallel gear (b).

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
10 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 24. Bond Graph model and sub-model of a wind turbine gearbox.

employed in the wind turbine design and analysis environment. It


corresponds to the Blade Element Momentum (BEM) theory.
Extensive literature deals with the BEM theory explanation and
presentation [25,26].
The BEM theory combines two methods of examining how a
wind turbine operates. The first method uses a momentum bal-
ance on a rotating annular stream tube passing through a turbine.
Axial thrust dFT and tangential force dFr, in terms of flow para-
meters, are:

dF T ¼ 4að1  aÞρV 2 π rdr ð24Þ

dF r ¼ 4a0 ð1  aÞρV Ωr π r 2 dr ð25Þ


Where V represents the wind velocity, ρ the air density, a the
axial induction factor, a0 the angular induction factor, Ωr the
angular velocity of the blades, and r the radius of an annular ele-
ment, having a thickness dr. The second method is to examine the
forces generated by the airfoil lift and drag coefficients at various
sections along the blade. Axial force, tangential force and aero-
dynamic moment, in terms of the lift, the drag and the pitching
moment coefficients of the airfoil, are as follows:
Fig. 25. Gearbox simulation model.
1 V 2 ð1  aÞ2
dF T ¼ ρ ðC L cos ϕ þ C D sin ϕÞcdr ð26Þ
subsystem are the wind speed V, the pitch angle β, the rotational and
2 sin 2 ϕ
axial speeds of the rotor ωr and the axial displacement of the blades
1 V 2 ð1 aÞ2
caused by flapping and tower bending yb, respectively. Its outputs are dF r ¼ ρ ðC L sin ϕ  C D cos ϕÞcdr ð27Þ
the aerodynamic torque Tr tangential force Fr, the thrust force FT and
2 sin 2 ϕ
the aerodynamic moment.
1 V 2 ð1  aÞ2   
Wind turbine aerodynamic models are used to relate wind dM x ¼ ρ C m c þ yG C L cos ϕ þ C D sin ϕ  ðzG  za Þ
inflow conditions to the loads applied to the turbine. The sub-
2 sin 2 ϕ
 
sequent analysis develops the most common aerodynamics theory  C L sin ϕ  C D cos ϕ cdr ð28Þ

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 11

Fig. 26. The complete model of the mechanical subsystem of WECS.

Fig. 27. Velocities at rotor plane (a), center of gravity coordinates and the aerodynamic center (b).

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
12 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

4. Calculate new values of ai and a0i using the following equations:


!1
4 sin ϕi
2
ai ¼ 1þ   ð35Þ
σ i C Li cos ϕi þ C Di sin ϕi
0

!1
4 sin ϕi cos ϕi
a0i ¼ 1þ   ð36Þ
σ 0i C Li sin ϕi  C Di cos ϕi

Fig. 28. Lift, drag and pitching moment coefficients for a NACA.


In this process iterations are done until the values of induction
Vð1  aÞ
ϕ ¼ tan  1 ð29Þ factors and inflow angle converge to their final values, and then we
Ωr ð1 þa Þ
0
can calculate the axial force and the tangential force for each
where c is the chord length of the blade element, the wind section along the blade by:
inflow angle ϕ is the angle between the local flow direction and
1 V 2 ð1  ai Þ2  
the rotor plane (Eq. (29)), parameters involved in these expres- F Ti ¼ ρ C Li cos ϕi þ C Di sin ϕi ci li ð37Þ
sions are graphically represented in Fig. 27(a). ð yG ; zG Þ are the 2 sin ϕi
2

center of gravity coordinates, za the aerodynamic centers coordi-


nate, as shown in Fig. 27(b). 1 V 2 ð1  aÞ2  
F ri ¼ ρ C Li sin ϕi  C Di cos ϕi ci li ð38Þ
CL, CD and Cm are lift, drag and pitching moment dimensionless 2 sin 2 ϕi
coefficients respectively as functions of the angle of attack α. Lift,
drag and pitching moment coefficients for a NACA 4415 airfoil are 1 V 2 ð1  aÞ2   
shown in Fig. 28. This graph indicates that for low values of inci- M xi ¼ ρ C mi ci þ yGi C Li cos ϕi þ C Di sin ϕi  ðzGi  zai Þ
2 sin 2 ϕi
dence angle, the airfoil successfully produces a large amount of lift  
with little drag.  C Li sin ϕi  C Di cos ϕi ci li ð39Þ
Eqs. (25) and (27) are used to calculate the axial induction:
The fundamental aerodynamic theory used by the Bond Graph
!1
4 sin ϕ
2 model is presented in this paper. A MGY-element of bond graph is
a¼ 1þ 0 ð30Þ used to implement Eqs. (32)–(39), with wind flow (MSf source)
σ ðC L cos ϕ þ C D sin ϕÞ
being transformed into aerodynamic forces (Se source), as shown

1 in Fig. 29.
4 sin ϕ cos ϕ
a0 ¼ 1þ ð31Þ Modulated inputs to MGY elements of Fig. 29 are the pitch
σ 0 ðC L sin ϕ  C D cos ϕÞ angle β, the angular velocity Ωr, and the wind speed V and the
Solution for a given blade cannot be found directly from the axial displacement of the blades caused by flapping and tower
equations but an iterative method is required to calculate the axial bending. In order to simulate the blade model, Eqs. (32)–(39) are
induction factor, the tangential induction factor, angles of attack integrated within each MGY element; which means that their
and thrust coefficients for each section along the span of the blade. traditionally constitutive relation is changed.
It is composed of the following steps: The non-dimensional power coefficient Cp represents the
fraction of available power in the wind that is extracted by the
1. Estimate the initial value of the axial induction factor ai and the
tangential induction a0i relative to the ith blade element.
An efficient technique (among others) is to assume that the
inflow angle is small: sin ϕi  ϕi . The tangential induction a0i
and the drag coefficient CDi are null. The lift coefficient is
C Li ¼ 2παi , where the angle of attack is αi ¼ αi  β i , with βi being
the sum of the twist angle βti depending on the position of the
center of gravity of the ith section and the pitch angle β which is
an input control variable. After some rearranging, it yields to:

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1   Fig. 29. Bond Graph model of the aerodynamic subsystem.
2 þ πλri σ 0i  4  4πλri σ 0i þ πλri σ 0i 8βi þ πσ 0i
2
a¼ ð32Þ
4

With λri ¼ ΩVr ri being the local speed ratio and σ 0i ¼ 2πciri being the
local solidity, calculated using chord values ci depending on ri
(Appendix A).
2. Use the initial value of ai and a0i to calculate ϕi.


V ð1  ai Þ
ϕi ¼ tan  1 ð33Þ
Ωr r i ð1 þ ai Þ
0

3. Calculate incidence angle αi and then CLi and CDi using look-up
tables (Fig. 20).

αi ¼ ϕi  βi ð34Þ Fig. 30. Curve of Cp vs λ.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 13

Fig. 31. Bond Graph model and sub-model of the induction machine.

turbine. It can be rewritten as: induction machine. Therefore, for an efficient control of the
2 mechanical-magnetic–electric energy, the induction machine has
C p ¼ 4að1  aÞ ð40Þ
to be accurately modeled [27].
The theoretical maximum power coefficient from an idealized Induction machines have been addressed in many publications.
rotor Cpmax, known as Betz limit, can be found by setting the There are two general frameworks models: one using a Park
derivative of Cp with respect to a equal to zero, which leads to reference frame [28] and the other using the natural reference
a ¼ 1=3 then C pmax ¼ 0:596 which corresponds to the maximum frame (three sinusoidal waveforms) [29]. An induction machine
possible efficiency for an idealized wind turbine of 59.6%. Fig. 30 Bond Graph model in the natural reference frame is sketched in
shows the values of Cp, the simulation program calculates Cp as: Fig. 31.
The assumptions considered for the induction machine model
F r RΩr
Cp ¼ ð41Þ are:
1
ρπ R2 V 3
2
 Magnetic hysteresis and magnetic saturation effects are
From the figure the maximum value of Cp is 0.42, having a tip neglected.
speed of 7; the theoretical maximum value of Cp is 0.596, known  The stator windings are sine-wave-distributed throughout the
as Betz limit, at around a tip speed of 8. In practical designs, the airgap.
maximum achievable Cp is below 0.5. The difference to standard  The stator slots cause no appreciable variation of the rotor
values is acceptable and it is better than the value shown in [15] inductances with respect to the rotor position.
(around 0.33). This simulation confirms that dynamical model of
the blade þ hub þ main bearing adequately performs. Fig. 31 explicitly shows the three phases of stator and rotor.
The phase 'a' stator input voltages is represented by a modulated
2.4. Electrical subsystem effort source (MSe ¼Vsa). Coil resistance translating the losses of
phase a is represented by R-field, Rsa. The effort source and the R-
Energy is basically produced in mechanical form and consumed field are connected to one junction. The stator phase 'a' input
in electrical form. The induction generators are largely the most current can be measured at this junction. The same is done for
popular electric machines in WECS industry to convert mechanical phases 'b' and 'c'.
power to electric power through the medium of magnetics. Thus, a The rotor energy port is similar to the stator energy port with a
multi-domain energy exchange in an induction machine alw- different set of parameter for elements. The non-idealities of the
ays takes place. This increases the complexity of modeling an core loss are represented by the R-field, Rri . The rotor input

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
14 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

0

1
voltages are implemented by an effort source equal to zero. The IC-   2π 2π
B cos θ cos θ  cos θ þ
port-field denotes the self ðLs; LrÞ and mutual (Lm) inductances B 3 3 C
B

C
C
between the stator and rotor, which are symbolized by: B 2π   2π C
LM ¼ LSR B cos θ þ cos θ cos θ  C ð45Þ
B 3 3 C
B

C
! ! ! B 2π 2π   C
ϕa;b;cs LS LS ia;b;cs @ cos θ  cos θ þ cos θ A
¼ ð42Þ 3 3
ϕa;b;cr LTM LR ia;b;cr
The IC-port field is used because Eq. (45) contains sine and
where cosine relations, which are necessary to calculate the rotor posi-
0 1 tion θm (Eq. (46)). This means that the speed machine is integrated
LIS þ LMS
1
 LMS
1
 LMS in order to calculate θm.
B 2 2 C
B C At the shaft level, the developed electromagnetic torque
B 1 1 C
LS ¼ B
B  2 LMS LIS þ LMS  LMS C
C ð43Þ (Eq. (46)) as a function of the stator, rotor currents and the angle
B 2 C between them is represented in the IC-port output. The electro-
@ 1 1 A
 LMS  LMS LIS þ LMS magnetic torque Te is consumed at the 1-junction by the inertial
2 2
moment Jind, frictional and mechanical torque T (which, in this
case, will be provided by output gearbox stage). The information
0 1
1 1 flow at this junction, which gives the rotor speed value, is trans-
LIR þ LMR  LMR  LMR
B 2 2 C mitted to the Axis Rotor for calculating the instantaneous angle of
B C
B 1 1 C the rotor windings. Eq. (47) shows the mechanical part expression:
LR ¼ B
B  2 LMR LIR þLMR  LMR C
C ð44Þ
B 2 C 0 1
@ A ira
1
 12 LMR  LMR LIR þ LMR d B C
2 T e ¼ ðisa isb isc Þ ðLM ðθm ÞÞ33 @ irb Ap ð46Þ
dθm
irc

d
J ind Ωm ¼ T e  T ð47Þ
dt
A simulation is conducted to check out the models behavior. In
order to allow a direct electrical network connection, an RL ele-
ment, used as a load, is connected in series between the generator
and the three-phase power network. Appendix A contains the
complete set of parameters.
Stator currents curves and speed rotor are shown in Fig. 32. The
goal is to show that the model works as expected. In the next
section, it will be connected as a generator having a negative
torque.

2.5. Pitching subsystem

Fig. 32. Stator currents forms and speed rotor simulation. In WECS, one of the operational problems is the variability and
discontinuity of wind. In most cases, wind speed can fluctuate
rapidly. Hence, quality of produced energy becomes an important
problem. Several control techniques have been designed to
improve the quality of power generated from wind turbines. Pitch
control is the most efficient and popular power control method,
especially for variable-speed wind turbines. It is a useful method
for power control above the rated wind speed by changing the
pitch angle of the rotor blades. The actuator that drives the blades
around their longitudinal axes is a hydraulic or electromechanical
device. Higher flexibility of these devices enabled the imple-
mentation of efficient and reliable model and control strategies for
power or speed limitation. The pitch actuator is a nonlinear servo
Fig. 33. Subsystems of the pitching system. that generally rotates all the blades or part of them.

Fig. 34. Bond Graph model of the pitching system.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 15

Many wind turbines are equipped with controlled electro- sketched. The mechanical subsystem includes all mechanical
mechanical actuators to adjust the rotor blades. Taking the components of the pitching system: the inertia of rotor of the
dynamic characteristics of the pitching system into consideration electric motor is represented by I-element ð J rotor Þ, the friction at the
is crucial, for defining the control strategy and for designing the rotor by R-element ðBrotor Þ the transmission element (gearbox) is
pitch drive components. represented by TF-element (TF). The rotor blade inertia is repre-
The pitching system of a wind turbine can be divided into four sented by I-element ð J blade Þ.
subsystems (Fig. 33) [30]. The consideration of the interaction The aerodynamics effort is represented by Se-element that
between these systems is essential in order to achieve a correct represents the aerodynamic moment applied for each blade ele-
load calculation for the components of the total system [31]. ments for which the angle of attack α is calculated by iteration. The
The bond graph of the pitching system is depicted in Fig. 34. used airfoil provides the lift, drag and pitch moment coefficients
where the different functional parts of the pitching system are C L ; C D ; C M respectively. Besides, the pitching moment is determined
using equations that were introduced in Aerodynamic Subsystem
section.
The electrical subsystem represents the electric motors beha-
vior. There are many possibilities to model the electric motor with
different levels of complexity and accuracy. The detailed method
depends on the motor type. In our model, we use a DC motor for
controlling the pitch angle. The Bond Graph model is depicted in
the electrical part of Fig. 34, where the coil resistance, featuring
the rotor losses, is represented by an R-element ðRind Þ and the
motor inductance by an I-element ðLind Þ. Fig. 35 shows the Bond
Fig. 35. Bond Graph sub-model of the pitching system.
Graph sub-model.

Fig. 36. Complete wind turbine model.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
16 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 40. Power curves of the model.

Fig. 37. Axial force, tangential force and pitching moment.

Fig. 41. Curve Cp and λ.

Fig. 38. Responses for a constant wind.

Fig. 42. Gearbox vibration and deflection blades and tower.

Fig. 39. Stator currents of generator.


blade element. Mechanical subsystems output is the mechanical
power, which represents the electrical subsystems input. The
power network connects the output of the electrical subsystem. In
2.6. Complete system
order to ensure a power exchange, a load is connected between
The individual subsystems presented in the previous sections generator and power network. Pitching subsystems output is
are combined in a global system as shown in Fig. 36. The three connected with the pitching moment model of the mechanical
inputs of the aerodynamic subsystem are connected by the wind subsystem, and its input is the control law. The different control
sources, the hub speed and the pitch angle, the latter representing laws (i.e. pitch, speed and break control) are not dealt with in this
the primitive of the rotation speed at the pitching subsystems paper. A more detailed model will be presented in a subsequent
output. The three output forces of the aerodynamic subsystem paper. Actually, some considerations are required to allow the
symbolize the mechanical subsystems' inputs applied on each wind turbine to work without pitch control. The assumptions are:

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 17

Fig. 45. Stator currents of generator.

Fig. 43. Axial force, tangential force and pitching moment.

Fig. 46. Power curves of model.

Fig. 47. Curve Cp and λ.

Fig. 44. Responses for a variable wind.

 Wind flow never exceeds its nominal values. This allows the
wind turbine to operate without pitch control.
 A load is connected between the generator and the power
network, this yields a power exchange.

3. Simulation and discussion

Besides calculating various parameters of the wind turbine, two


different simulations were carried out. A constant and variable
wind profile is considered in the time domain in order to analyze
the interaction between the different subsystems of the wind
turbine. The constant wind simulation scenario is: a constant wind Fig. 48. Gearbox vibration and deflection blades and tower.
flow of 5 m=s is applied at the simulation start; then, at t ¼ 10 s, a
wind ramp is applied. This wind ramp increases from 5 m=s up to variable wind profile. Fig. 37 depicts the three aerodynamic forces
13 m=s, for t ¼ 10 s through t ¼ 50 s. This value is maintained until applied in each one of the blade elements' center of gravity, from
the simulation ends (no control laws are used for the simulation). t ¼ 50 s to t ¼ 100 s. The tangential aerodynamic forces applied for
The following figures show the simulation scenario of the each element are F t1 ¼ 1:45 kN; F t2 ¼ 1:5 kN; F t3 ¼ 1:2 kN. And each
whole wind turbine conversion system for the constant and elements' center of gravity position is r 1 ¼ 3:9 m; r 2 ¼ 11:7 m; r 3 ¼

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
18 Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎

19:5 m. Actually, the rotor torque of the three blades must be This present work highlights the fact that all these parts are
P
T t ¼ 3 3i ¼ 1 F ti r i ¼ 139; 815 kN m, which is verified as shown in considered in one integrated model. This model deeply describes
Fig. 38 where the torque value from t ¼ 50 s to t ¼ 100 s is the actual behavior of the turbine in that it takes into account the
T t ¼ 140:067 kN m. The system undergoes some oscillations in the most important dynamics phenomena encountered in the system.
beginning of the simulation due to the dynamic behaviors of the We have tried to emphasize that the BGA provides a better
system in the transient phase. Fig. 38 shows also the torque at high understanding of what actually happens in the system.
shaft, from t ¼ 50 s to t ¼ 100 s. The high shaft and low torque are After simulation for both constant and variable wind speed
respectively, T t ¼ 140:067 kN m, and T g ¼ 2:334 kN m. With these consideration, good results have been obtained with the proposed
two values, it was proved that a relationship between the high and model, but comparisons with real measurements are not available
the low shafts confirms the gearbox reduction ratio of 60. at this stage. Research works by other authors are under progress
A 490 V generator is used as an input of the electrical sub- in this direction.
system. Fig. 39 shows the three-phase stator currents. The present model can be used to determine the loads of the wind
Wind turbine converts wind power into mechanical power, turbine drive train during the simulation process, and to calculate the
and into electrical power, and then distributed to the load or to deflection of the blade and tower and then develop the law control
the power network. Fig. 40 shows the aerodynamic, mechanical algorithms, which would minimize the probability of blade tower
and electrical power in the wind turbine model. The power of vibrations. The model can also be used to provide information on the
the main shaft and the high speed shaft approximately match gearbox vibration, as well as help in the design the components of the
ð740 kWÞ; the generators active power ð733 kWÞ is close to this pitching system. Moreover, it is possible to use this model for the
value. This difference is due to the lost power in the generator. system dynamic behaviors optimization. It was also efficiently used
Fig. 41 also shows the conventional values for Cp and λ. Fig. 42 for testing new control strategies of wind turbines.
sketches the blades, tower deflections, and the gearbox vibra- Only a few publications involving all the stages of a wind tur-
tion. The maximum deflection of the tower and the blades are bine are available in the open literature. In this context, the model
respectively 0:2 m and 0:45 m, The latter (whose value is 0:45 m) proposed here is an important starting point for the analysis of
takes into account the deflection of the tower, because it is this complex system.
assumed that the top of the tower and the bottom of the blade Based on the results of this paper, interesting future prospective
in the axial extension are rigidly connected. Therefore, the research will include performing control design using bond graph
actual deflection blade is 0:25 m (namely 0:45 0:2 ¼ 0:25 m). and possibly constructing a complete wind turbine model.
The vibration amplitude of the gearbox lies between 7 0:004 m.
These values are acceptable while the proposed wind profile is
around its rated value. In the case of the high wind speed, a Appendix A
control system is needed to keep the structure system and the
gearbox vibration around its rated values. See Tables A1 and A2.
The second simulation is carried out using a variable wind
speed. Figs. 43–48 show the simulation responses for the gen- Table A1
erator selected variables. The torque values C p ; λ and omega are Blade structure data.
taken to their average values. It is revealed that responses of
Parameters Section 1 Section 2 Section 3
each stage change with wind speed. In contrast to a constant
wind input, currents and power responses also reflect their E, Young modulus of blade material (N/ 1.71  109 1.71  109 1.71  109
variability with wind profile. This allows the behavior verifica- m2)
G, Coulomb modulus of blade material 6.58  109 6.58  109 6.58  109
tion of the proposed model.
(N/m2)
This contribution was based on real data for a 750 kW wind Ix, The second moments of area of axis 7.7  10  5 5  10  4 6.2  10  5
turbine, all parameters of the system are reported in Appendix A. x (m4)
Iy, The second moments of area of axis 3.85  10  3 2.5  10  3 3.1  10  4
y (m4)
J, Second moment polar (m4) 4.62  10  5 5.2  10  5 2.6  10  5
4. Conclusions and futures L, Element length (m) 7.8 7.8 7.8
M, Mass of section (kg) 928 560 207
Jt, Inertial mass of section (kg m2) 26 13.78 1.66
Due to interactions between aerodynamic, mechanical and
μ, Structural damping coefficient 0.01 0.01 0.01
electrical subsystems, a mechatronic model of a wind turbine
Blade aerodynamic conversion
generating system using the Bond Graph Approach is proposed to
r, Position on the center of gravity (m) 3.9 11.7 19.5
analyze the dynamic behavior of a wind turbine. The closed con- c, Cord (m) 1.56 1.459 0.8315
nection between the dynamics of the rotor blade and the pitch βt, Twist angle (deg) 7.65 4.53 2.72
actuator requires a certain complexity of the mechanical model. By ρ, Air density (kg/m3) 1.225 1.225 1.225
za, Coordinate of the aerodynamic 0.2652 0.25 0.14
means of the presented modeling method, a correct load calcula-
center (m)
tion of the different parts of the system can be carried out. We are yG, Coordinates of the center of gravity 0.5 0.47 0.226
not looking to validate a specific turbine system, but we aim to (m)
show a suitable way to model it. zG, Coordinates of the center of gravity 0.11 0.102 0.06
(m)
Some contributions have used the bond graph implementa-
tion, but they involve only one part of the wind turbine, e.g. the Tower
E, Young modulus of tower material
gearbox in [20] or blade in [16]. Others [32,33] have modeled
(N/m2)
the principal part of the wind turbine but the models are not I, The second moments of area (m4) 1.474 0.414 0.826
enough detailed. This gap was seemingly filled in [19] but the L, Element length (m) 22.33 22.33 22.33
axial extension deformation of the blade, the deflection of the M, Element length (kg) 56.214  103 38.581  103 23.901  103
Jt, Inertial mass of section (kg.m2) 1.99  105 1.064  105 5  104
tower, the vibration of the gearbox and the pitching system
μ, Structural damping coefficient 0.01 0.01 0.01
were not considered.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i
Z. Khaouch et al. / ISA Transactions ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 19

Table A2

Hub and main shaft. Generator

Jhub, Hub inertia (kg m2) 5000 Vn, Voltage of electrical network (V) 460
Dhub, Main bearing (N s/m) 1000 f, Frequency network (Hz) 50
Kms, Main shaft stiffness (N/m) 3.67  107 Rs, Stator resistance (Ω) 4.92  10  4
Dms, Main shaft bearing (N s/m) 200 Rr, Rotor resistance (Ω) 2.7  10  4

High speed shaft p, Number of poles 2


Khs, High shaft speed stiffness (N/m) 107 Lls, Self-inductance of stator (H) 4.66  10  5
Dhs, High shaft bearing (N s/m) 54.5 Lm, Mutual inductance (H) 1.99  10  3

Load Jind, Rotor inertia (kg m2) 18.7


R, Charge resistance (Ω) 20
L, Charge inductance (H) 0.1

Gearbox
Js, Sun inertia (kg m2) 3.2 Krg, Mesh stiffness between ring and hub (N/m) 29.2  109
Jr, Ring inertia (kg m2) 144.2 Drg, Mesh damping between ring and hub (N s/m) 106
Jc, Carrier inertia (kg m2) 59.1 Kst1, Mesh stiffness of first shaft of parallel gearbox (N/m) 5.28  105
Jp, Planet inertia (kg m2) 3.2 Dst1, Mesh damping of first shaft of parallel gearbox (N s/m) 5.8  102
J1, Parallel gear 1 inertia (kg m2) 11.4 Kst2, Mesh stiffness of second shaft of parallel gearbox (N/m) 7.3  105
J2, Parallel gear 2 inertia (kg m2) 3.2 Dst2, Mesh damping of second shaft of parallel gearbox (N s/m) 8  102
J3, Parallel gear 3 inertia (kg m2) 8.16 K1, Mesh stiffness between parallel gear 1 and 2 (N/m) 13.5  109
J4, Parallel gear 4 inertia (kg m2) 3 D1, Mesh damping between parallel gear 1 and 2 (N/m) 6.23  103
Zs, Number of planetary teeth 21 K2, Mesh stiffness between parallel gear 3 and 4 (N/m) 9.8  109
Zr, Number of ring teeth 99 D2, Mesh damping between parallel gear 3 and 4 (N s/m) 4.67  103
Zp, Number of sun teeth 39 Ksp, Mesh stiffness between sun and planet (N/m) 16.9  106
Z1, Number of parallel gear 1 teeth 70 Dsp, Mesh damping between sun and planet (N s/m) 4.65  103
Z2, Number of parallel gear 2 teeth 20 Krp, Mesh stiffness between ring and planet (N/m) 19.2  109
Z3, Number of parallel gear 3 teeth 51 Drp, Mesh damping between ring and planet (N s/m) 2.24  104
Z4, Number of parallel gear 4 teeth 17

References [16] Lee C-W, Yun J-S. Dynamic analysis of flexible rotors subject to torque and
force. J Sound Vib 1996;192:439–52.
[17] Agarwal S, Chalal L, Dauphin-Tanguy G, Guillaud X. Bond graph model of wind
[1] Breedveld Peter C. Port-based modeling of mechatronic systems. Math Com- turbine blade. In: Bond graph model: theory and Practice. Vienna: MathMod;
put Simul 2004;66:99–127. 2012.
[2] Buur J. A theoretical approach to mechatronics design [Ph.D. dissertation];1990. [18] Spruce Christopher J, Turner Judith K. Tower Vibration Control of Active Stall
[3] Assareh Ehsanolah, Biglari Mojtaba. A novel approach to capture the max- Wind Turbines, IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY,
imum power from variable speed wind turbines using PI controller,: RBF VOL. 21, NO. 4, July 2013, http://dx.doi.org/10.1109/TCST.2013.2261298.
neural network and GSA evolutionary algorithm. Renew Sustain Energy Rev [19] GWEC, Global Wind Statistics 2012. Global Wind Energy Council; February
2015;51:1023–37. 2013.
[4] Poultangari I, Shahnazi R, Sheikhan M. RBF neural network based PI pitch [20] Coudert N, Dauphin-Tanguy G, Rault A. Mechatronic design of an automatic
controller for a class of 5-MW wind turbines using particle swarm optimiza- gear box using bond graphs. In: Proceedings of the IEEE systems man and
tion algorithm. ISA Trans 2012;51:641–8. cybernetics conference, 1993.
[5] Enge-Resenblatt Olaf, Schneider Petre. Modelica wind turbine models with [21] Deur Joko, Ivanovic Vladimir, Assadian Francis, Kuang Ming, Tseng Eric H,
structural charges related to different operating modes. Fraunhofer Institute Hrovat D. Bond graph modeling of automotive transmissions and drivelines.
Integrated Circuits, Branch Lab Design Automation. Zeunerstraβe 38, 01069 In: Proceedings of 7th Vienna International Conference on Mathematical
Dresden, Germany. The Modelica Association; 2008. p. 610–9. Modelling (MATHMOD), Vienna, Austria, 2012.
[6] Chen Qiong-zhong, D'Ambrosio Flavio, Defourny Michel, Brüls Olivier. Inte- [22] Sanchez R, Medina A. Wind turbine model simulation: a bond graph approach.
grated control and structural analysis of DFIG wind turbines using a mono- Simul Model Pract Theory 2014;41:28–45.
lithic approach, European Wind Energy Conference & Exhibition, 2012. [23] Chaari F, Fakhfakh T, Haddar M. Dynamic analysis of a planetary gear failure
[7] Joselin Herbert GM, Iniyan S, Amutha D. A review of technical issues on the caused by tooth pitting and cracking. J Fail Anal Prev 2006;6:73–8.
development of wind farms. Renew Sustain Energy Rev 2014;32:619–41. [24] Chaari F, Baccar W, Abbes MS, Haddar M. Effect of spalling or tooth breakage
[8] Njiri Jackson G, Söffker Dirk. State-of-the-art in wind turbine control: trends on gearmesh stiffness and dynamic response of a onestage spur gear trans-
and challenges. Renew Sustain Energy Rev 2016;60:377–93. mission. Eur J Mech—A/Solids 2008;27:691–705.
[9] Hoffmann MJ, Reuss Ramsay R, Gregorek GM, Effects of grit roughness and [25] Liu Xiong, Chen Yan, Ye Zhiquan. Wind turbine computing model of aero-
pitch oscillations on the NACA 4415 airfoil. The Ohio State University dynamic performance with horizontal axis. J Sol Beijing 2005;26:792–800.
Columbus, Ohio; July 1996. [26] Hansen MOL, Sorensen JN, Voutsinas S, Sorensen N, Madsen HA. State of the
[10] Margolis Donald. Bond Graph Modelling of Engineering Systems: Theory, art in wind turbine aerodynamics and aeroelasticity. Progress Aerosp Sci
Applications and Software Support; 2011. 2006;42:285–330.
[11] Merzouki Rochdi, Samantaray Arun Kumar, Pathak Pushparaj Mani, Ould [27] Delaleau E, Louis JP, Ortega R. Modeling and control of induction motors. Int J
Bouamama Belkacem. Intelligent mechatronic systems: modeling, control and Appl Math Comput Sci 2001;11(1):105–29.
diagnosis. 2013. [28] Sahm D. A two-axis,: bond graph model of the dynamics of synchronous
[12] Bianchi Fernando D, De Battista Hernn, Mantz Ricardo J. Wind turbine control electrical machines. J Frankl Inst 1979;308(3):205–18.
systems: principles, modelling and gain scheduling design. 2007. [29] Dauphin-Tanguy G. Les Bond Graphs. Herms Science Editor, Paris, 2000.
[13] Ponta Fernando L, Otero Alejandro D, Lago Lucas I, Rajan Anurag. Effects of [30] Schlecht Berthold, Mtauweg Samer, Rosenlcher Thomas. Dynamic design of an
rotor deformation in wind-turbine performance: the dynamic rotor defor- electromechanical pitching system for wind turbines by means of multi-body
mation blade element momentum model (DRDeBEM). Renew Energy simulation. Institute of Machine Elements and Machine Design, Dresden
2016;92:157–70. University of Technology, Customer Application, SIMPACK News 2011.
[14] Jung Sungmoon, Kim Sung-Ryul, Patil Atul, Hung Le Chi. Effect of monopile [31] Mtauweg S. Dynamische Analyse des Pitchsystems der Windenergieanlage GE
foundation modeling on the structural response of a 5-MW offshore wind 1.5 MW. TU-Dresden: unverffentlichter Bericht; 2010.
turbine tower. Ocean Eng 2015;109:479–88. [32] Bakka Tore, Reza Karimi Hamid. Bond graph modeling and simulation of wind
[15] Shipley DE, Miller MS, Robinson MC, Luttges MW, Simms DA. Evidence that turbine systems. J Mech Sci Technol 2013;27(6):1843–52.
aerodynamic effects, including dynamic stall, dictate HAWT structural loads [33] Bakka Tore, Karimi Hamid Reza. Wind turbine modeling using the bond graph.
and power generation in highly transient time frames. In: WINDPOWER, In: IEEE Multi-conference on systems and control, 2011. p. 28–30.
AWEA, Minneapolis, 1994. p. 615–26.

Please cite this article as: Khaouch Z, et al. Mechatronic modeling of a 750 kW fixed-speed wind energy conversion system using the
Bond Graph Approach. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.07.009i

You might also like