net/publication/317576660
CITATIONS READS
0 498
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Iurii Storozhenko on 14 June 2017.
I.Storozhenko, Q.Sun
Abstract: The objective of this research is to develop and design a numerical model of
wind turbines (WT) capable of simulating WT dynamic behavior under healthy and faulty states.
Existing mathematical models cannot be used for an accurate representation of the processes
occurring inside the WT doubly-fed induction generators (DFIG) and their influence on other
components of the WT. This paper provides a modeling methodology for simulating the WT
dynamic behavior by combing the aerodynamics and mechanical models by FAST (Fatigue,
Aerodynamics, Structures, and Turbulence) with the finite element model of the DFIG using
Matlab/Simulink. The adequacy of the developed DFIG Finite element model is validated by the
simulation of turn-to-turn fault of the generator stator phase winding.
Keywords: wind turbine, doubly fed induction generator, numerical modeling, finite
element modeling.
1. Introduction:
The development of WT technologies has experienced a tremendous growth during the last
three decades. A lot of different configurations of WT topologies were presented, however one of
the most popular topology on the market is the horizontal axis wind turbine (HAWT) with DFIG.
Nearly 85% of WT under operation at present time are based on this topology [1]. This is due to
the high level of reliability, increased wind energy production, limited extent faults ride-through
capability and some other advantages.
Because of the continuous development in the wind energy field, the level of availability
of WT at present time has reached 98% [2]. Still, failures can occur during operation. Unexpected
failures of WT can result in: underproduction of electrical energy; instability of the power grid
system, injuries or fatal outcome for operational or maintenance personnel. To increase reliability
and reduce downtime, new intelligent diagnostic techniques and condition monitoring systems
should be developed. The key element to progress in this field is to develop accurate numerical
models of WT. Numerical models can be used for the simulation of WT dynamics behavior and
understanding transient loads and their couplings across the drivetrain components.
Numerical models can be used not only to study the nonlinear interactions between
different WT subsystems and parameters, and their influence on WT dynamic behavior, but also
for the investigation of WT operation under healthy and faulty states, which is very important for
1
enhancing the reliability of existing turbines and developing new diagnostic methods and
algorithms.
This paper demonstrates a methodology for developing WT operational model that can
simulate the actual operation of WT under healthy and faulty states. The approach of this study is
to show the design and the construction of a numerical simulator of a WT by using three simulation
tools: Matlab/Simulink, Comsol Multiphysics and FAST (Fatigue, Aerodynamics, Structures and
Turbulence) an aeroelastic simulation code.
2
Fig.2. WT subsystems and their interactions [4].
In order to develop a WT model, we must clearly understand the physical nature of the
processes occurring inside each of the presented subsystems. The identification of the assumptions
and limitation for each subsystem is also very important.
3
FAST is an aero-elastic code developed initially by Oregon State University. It is one of
the most advanced design codes for a HAWT [5]. To simulate a turbine dynamic behavior the Fast
uses a combination of modal and multibody dynamics formulations [21].
Let us have a look on the modeling process of each subsystems of a WT in more details.
3.2. Aerodynamic and mechanical drivetrain subsystems:
A detailed representation of WT aerodynamics can be implemented by algorithms built-in
in FAST. The model of a WT in FAST mainly includes tower, blades, shaft, furl, and support
platform. The aerodynamic model is based on the unsteady blade-element momentum theory and
covers an implementation of rotational sampling, turbulent wind fields and a tower model for the
structural dynamics.
A mechanical drivetrain in FAST is represented by a two-mass model, which is usually
enough for the analysis of the dynamic behavior of WT. A two-mass model consists: the generator
(with gearbox) mass and the hub with blades (wind wheel) mass. The connections between masses
is represented by springs defined by damping and stiffness coefficients. The two-mass model of
the mechanical drivetrain is presented in Fig.3.
Depending on the drive train dynamics to be analyzed, more detailed drivetrain models can
be developed in Matlab/Simulink. The detailed description of the existing mechanical drivetrain
models, and analysis of their influence on dynamics of WT can be found in [18].
4
different electromagnetic properties. It makes the process of mathematical modeling of DFIG very
complicated and not straightforward.
During the last several decades, many different methods and techniques for the numerical
modeling of the rotating electrical machines have been developed [17]. Most methods can be
implemented for the dynamics analysis. Others can only be used for steady-state simulations. In
general, all modeling methods can be classified to four broad categories: circuit modeling, state-
space modeling, finite element modeling and methods based on the combination of mentioned
modeling methods. All methods have their limitations and assumptions, which greatly influence
the accuracy of the model generated output and the possibility to simulate fault conditions. The
comparison of different modeling techniques and their capability to simulate fault conditions can
be found in [17].
The analytical models of DFIG presented in the literature are based on the 5th or reduced
rd
3 order machine models. This simplified state-space models of DFIG usually does not consider
nonlinearities, for example, the magnetic saturation effect or skin effect presented in the generator.
Thus, the simulated DFIG quantities do not reflect the real processes accurately. As an example,
it was shown in [6] that using the 3rd order model may result in too low transient currents, which
may lead to inaccurate results. Another problem is that this model assumes constant lumped circuit
parameters. The simulation of a certain fault in such models is very difficult or even impossible.
One of the best way to study the electromagnetic phenomenon presented inside the DFIG,
considering aforementioned non-linearity’s, is to use the finite element method (FEM). The
detailed description of this method with it advantages and drawbacks with respect to the simulation
of DFIG modeling will be covered latter on.
5
1 Az 1 Az A
z J z (1)
x x y y t
Where is a relative magnetic permeability, Az axial component of magnetic vector
potential, electrical conductivity of the materials, J z exciting current density of the
windings.
The DFIG of interest i.e. the reference model is a 250 kW generator running of a 2300
Vrms line-to-line, 50 Hz, 3-phase supply. It is a 4-pole machine running at the synchronous speed
of o 1500 RPM. The stator winding configuration of the machine is shown in Fig. 5. There are 36
stator slots and 24 rotor slots.
The rating of the DFIG is presented in Table 1:
Table 1
The geometrical structure and finite element mesh description of the analyzed DFIG is
presented in Fig.4. The model was meshing with triangular elements. In case of symmetrical
structure of the machine, without faults, the computational time of the model can be reduced by
using the geometrical and magnetic symmetry. In this case one can model only one-fourth of the
machine by using anti-periodic boundary conditions [22].
6
The structure of the windings of the stator and rotor have significant influence on the space
harmonics content of the air-gap magnetic permeance due to slotting effect, which results in torque
pulsations [19]. In order to minimize this effect, the stator and rotor windings have to be
distributed, and shortened. The stator winding layout for the 4-pole DFIG with two-layer stator
winding with shortening are presented in Fig.5.
In the Fig.6, Va ,b ,c is the phase voltage, Ra ,b,c is the sum of coil side and end-winding
resistances, Lend is the end-winding inductance, Vemf is the induced electromotive force (EMF).
There are several possible ways to calculate the end-coil inductance [9]:
it can be calculated with an approximated formula;
it can be evaluated from the 3D model of the electrical machine windings;
is can be measured on the real generator;
it can be provided by the manufacturer.
7
The resistance and inductance of the end-windings Rend and Lend respectively, can be
calculated with empirical formulas given by Boldea [10]:
N 2 lend Cu
Rend
d 2Cu (2)
4
lend 2 y 0,02 (3)
2 0 N 2 (4)
Lend 0,34(lend 0,64 y)
p
Where y coil pitch, p number of pole pairs, number of parallel wires, N effective
turns per phase, dCu copper wire diameter, 0 the permeability of free space, Cu electrical
conductivity of copper, lend length of the end winding.
We should mention that the evaluation of inductance will not take into account the mutual
inductance between the end-winding parts of windings coils.
From Fig.6 based on the Kirchhoff’s circuit law and Faraday’s law of induction we can
write the voltage equation per each phase:
dI d
Va I a Ra Lend a (5)
dt dt
Where I a phase current, flux linkage inductance, Lend end winding inductance
outside the core region, which represents a part of flux linkage which is not included in .
The third term of the right-hand side of eq. (5) represents the back-EMF, which can be
expressed [11]:
Nl dA
Vemf m z d slot (6)
As s dt
Where As cross sectional area of stator (rotor) slot, lm an effective length of the coils in the
core region.
Equation (6) can be rewritten as:
n N Az n
Nn Az
Vemf lm
n
d d (7)
n 1 Asn w t n 1 Asn w t
n n
Where N n number of turns in coil, n , n number of the positively and negatively
oriented coil sides respectively, nw and nw are respectively, the cross-sectional areas of the
“go” and “return” side of the phase conductors.
By substituting (7) into (5) we will get the final equation:
dI n N Az n
N Az
Va I a Ra Lend a lm n d n d (8)
dt n 1 Asn nw t n 1 Asn w t
n
Equation (8) describes the DFIG finite element model considering the voltage excitation
of the windings.
8
4.3. The mechanical equation for the DFIG:
The electromagnetic torque of the DFIG can be calculated by several methods: Maxwell
stress tensor method, Coulomb’s method, Magnetic co energy derivation and Arkkio’s method.
The most commonly used approach is to use the Maxwell stress tensor. However, this method is
very sensitive to the air-gap mesh quality. The electromagnetic torque developed by DFIG was
simulated using the Arkkio method [12]. This method is more stable and less sensitive to the type
and number of meshing elements, and represents the volume integration of the magnetic flux
densities product. The mathematical representation of this method is expressed by
lm
0 (rs r0 ) S
Te r Br Bt ds (9)
where, Te electromagnetic torque, rs the inner radius of the stator, r0 the outer radius
of the rotor, Br radial component of the magnetic flux, Bt tangential component of the flux,
S surface of the enclosed area.
The movement in the DFIG is determined by the next equations
1
m (Te TL )dt (10)
J
m dt (11)
where J moment of inertia, TL load torque, m the angular speed of the generator,
angular position of the rotor.
9
Fig.7. Simplified representation of WT model.
10
Fig.8. Stator currents of the DFIG model without faults.
Fig.9. Stator currents under turn-to-turn fault in Phase A of the DFIG model.
Conclusion:
In this paper the methodology for the dynamic modeling of a WT is presented. The
proposed combination of Comsol Multiphysics software with FAST is quite unique, and allows
the study of the dynamic behavior of a WT based on the DFIG topology with higher level of
accuracy. The advantages of using the finite element model of the DFIG for WT simulation was
covered. The development of the DFIG finite element model is presented in details. To make the
11
modeling result of the DFIG more consistent, skewing effect of the rotor slots can be incorporated
into the DFIG finite element model.
This is a study still under development and further conclusions for this work will be the
subject of another publication.
References
1. Rechesteiner, R. (2008). Wind power in context- A clean revolution in the energy sec‐
tor. online, http://www.energywatchgroup.org/fileadmin/global/pdf/ 2009-
01_Wind_Power_Report.pdf.
2. MWEP Jahresauswertung 1996-2000, ISET, Kassel, Germany, ISBN 3-9805896- 6-8/5-
X/n-n.
3. https://www.mathworks.com/matlabcentral/fileexchange/57590-active-power-control-of-
grid-connected-dfig-15-wind-with-energy-system-by-indraneel-saki
4. Singh, M., Muljadi, E., Jonkman, J., Gevorgian, V., Girsang, I., & Dhupia, J.
(2014). Simulation for Wind Turbine Generators--With FAST and MATLAB-Simulink
Modules (No. NREL/TP-5D00-59195). National Renewable Energy Laboratory (NREL),
Golden, CO..
5. http://wind.nrel.gov/public/bjonkman/TestPage/FAST.pdf
6. Akhmatov, V. (2002, April). Modelling of variable-speed wind turbines with doubly-fed
induction generators in short-term stability investigations. In 3rd International Workshop
on Transmis-sion Networks for Offshore Wind Farms (p. 23).
7. Ferkova, Z. (2015). Comparison between 2D and 3D modelling of induction machine using
finite element method. Advances in Electrical and Electronic Engineering, 13(2), 120.
8. Dalcali, A., & Akbaba, M. (2016). Comparison of 2D and 3D magnetic field analysis of
single-phase shaded pole induction motors. Engineering Science and Technology, an
International Journal, 19(1), 1-7.
9. Konwar, R. S., Kalita, K., Banerjee, A., & Khoo, W. K. S. (2013). Electromagnetic analysis
of a bridge configured winding cage induction machine using finite element
method. Progress In Electromagnetics Research B, 48, 347-373.
10. Boldea, I., & Nasar, S. A. (2010). The induction machine handbook. CRC press.
11. Bastos, J. P. A., & Sadowski, N. (2003). Electromagnetic modeling by finite element
methods. CRC press.
12. Arkkio, A. (1987). Analysis of induction motors based on the numerical solution of the
magnetic field and circuit equations. Helsinki University of Technology.
13. https://www.comsol.com/video/form-assembly-rotating-machinery-models
14. Developing S-Function, http://files.matlabsite.com/docs/books/matlab-
docs/simulink_sfunctions_r2015a.pdf
15. Kanerva, S. (2005). Simulation of electrical machines, circuits and control systems using
finite element method and system simulator (Doctoral dissertation, Helsinki University of
Technology).
16. Singh, M., Muljadi, E., Jonkman, J., Gevorgian, V., Girsang, I., & Dhupia, J.
(2014). Simulation for Wind Turbine Generators--With FAST and MATLAB-Simulink
Modules (No. NREL/TP-5D00-59195). National Renewable Energy Laboratory (NREL),
Golden, CO.
12
17. Singh, A., Grant, B., DeFour, R., Sharma, C., & Bahadoorsingh, S. (2016). A review of
induction motor fault modeling. Electric Power Systems Research, 133, 191-197.
18. Muyeen, S. M., Tamura, J., & Murata, T. (2008). Stability augmentation of a grid-
connected wind farm. Springer Science & Business Media.
19. Neto, L. M., Camacho, J. R., Salerno, C. H., & Alvarenga, B. P. (1999). Analysis of a
three-phase induction machine including time and space harmonic effects: the A, B, C
reference frame. IEEE transactions on energy conversion, 14(1), 80-85.
20. Multiphysics, C. O. M. S. O. L. (2012). COMSOL multiphysics user guide (Version 4.3
a). COMSOL, AB, 39-40.
21. Prowell, I., Elgamal, A. W. M., & Jonkman, J. M. (2010). FAST simulation of wind
turbine seismic response. National Renewable Energy Laboratory.
22. Bastos, J. P. A., & Sadowski, N. (2003). Electromagnetic modeling by finite element
methods. CRC press.
13