W. Sauer. Dynamic Modeling of Doubly Fed Induction
-
Upload
pacha-mami -
Category
Documents
-
view
218 -
download
0
Transcript of W. Sauer. Dynamic Modeling of Doubly Fed Induction
-
8/12/2019 W. Sauer. Dynamic Modeling of Doubly Fed Induction
1/6
Dynamic modeling of wind power generation
Hector A. Pulgar-Painemal and Peter W. Sauer.Department of Electrical and Computer Engineering
University of Illinois at Urbana-Champaign, USA
E-mail: [email protected] and [email protected]
AbstractThis paper presents a dynamic model appropriatefor power system analysis. This article shows modeling assump-tions, derivation of a third order model for a doubly-fed inductiongenerator and its controller models. Due to the detail level,it can be used as a tutorial for students and engineers thatare new in this area. A four-bus system with one synchronousmachine and one wind turbine is used to perform a small signalstability analysis. No considerable difference is observed betweenthe modal behavior of our test system and a small system with
just one synchronous generator.Index TermsPower System Dynamics, Wind Power Modeling
I. INTRODUCTION
The energy consumption over the world has been sus-
tainedly increased in the last decades due to the rate of growth
in world gross domestic productthe main driver of energy
demand. It is expected that the electricity demand will increase
at a rate of 2.6% per year during the period 2004-2030.Additionally, global energy-related carbon-dioxide emissions,
a major cause of global warming, are expected to increase by
1.7% per year during the same periodreaching 40.4 109
tonnes in 2030. Unfortunately, power generation is projected
to contribute almost50% of that increased emission [1]. Thus,the power generation sector is under scrutiny and has to be
expanded to fulfill the high-energy demand scenario but takinginto account environmental effects such as global warming.
Due to the notorious impact on environmental problems and
the depletion of fossil fuels, renewable energy sources have
been considered appealing to face the forthcoming energy sce-
nario. Hydraulic, wind, solar, biomass and geothermal sources
are the most common alternative generation systems. Hy-
draulic systems are attractive due to their robustness, reliability
and high rated power levels. However, the main drawbacks are
the scarce available locations and the negative impact on the
local ecosystem by flooding extensive areas. At present, among
the other alternatives, wind generation systems are the most
qualified to produce electricity in power systems. Although
being irregular in their electricity production, wind farms areable to provide energy: (a) without the risk of depletion of
their primary energy source, and (b) being able to comply with
operations standards. Additionally, wind generation systems
have the fastest payback period [2], less than a year; the lowest
project gestation period, due to modular concept; and low
operation-maintenance costs [3]. These characteristics make
wind power attractive for mass production which is reflected
by the increase of the worldwide installed wind-power capacity
over the last years [4].
In the 1990s, typical wind power turbines were characteri-
zed by a fixed-speed operation. Basically, they consisted of the
coupling of a wind turbine, a gearbox and an induction ma-
chine directly connected to the grid. Additionally, they used a
soft starter to energize the machine and a bank of capacitors to
compensate the machine power-reactive absorption. Although
being simple, reliable and robust, the fixed-speed wind turbines
were inefficient and power fluctuations were transmitted to the
network due to wind speed fluctuations [5].
In the mid-1990s, variable-speed wind power turbines gave
an impulse to the wind power industry. A better turbinecontrol is able to reduce power fluctuations. Besides, optimal
power extraction from the wind was possible by operating the
turbine at optimal speed. Among the different configurations
of variable-speed wind turbines, the doubly-fed induction gen-
erator (DFIG), at present, is the most used in the development
of new wind farm projects. This configuration consists of the
coupling of a turbine, a gearbox and an induction machine
doubly connected to the griddirectly connected from the
stator circuits and indirectly connected from the rotor circuits
by using converters. Its main drawbacks are the use of slip
rings and protection in case of grid disturbances [6]. The
control can be done (a) by controlling the voltage applied
to the rotor circuits, (b) by adjusting the pitch angle of theturbine bladesangle of incidence of the blade and the wind
direction, and (c) by designing aerodynamically the turbine
blades to stall when the wind speed exceeds its limit [7].
Due to the importance of wind power generation on the
current and future worldwide energetic scenario, dynamic
models are required for teaching and training purposes. One
major problem is that dynamic models of commercial wind
power turbines contain proprietary information and require
confidentiality agreements between the company and the user
[8]. In the literature, there are several articles dealing with
different issues related to wind power generators. However, a
rather general description of the dynamic model can be found
and a complete dynamic model appropriate for power systemanalysis is not directly available. Thus, this paper presents
a detailed dynamic model for wind power generation which
is suitable for power system analysis. Modeling assumptions,
derivation of a third order model for the DFIG and its
controllers are described through out the article which can be
used as a tutorial for students and engineers that are new in
this area. Using a four-bus test system with one synchronous
generator and a wind turbine, the modal behavior is similar
to a small system with just one synchronous generator. The
-
8/12/2019 W. Sauer. Dynamic Modeling of Doubly Fed Induction
2/6
paper is structured as follow: in Section II a description of the
turbine, induction machine generator and its controllers are
presented; in Section III a four-bus system is used to perform
a small signal stability analysis; and in Section IV conclusions
are presented.
I I . DYNAMIC MODEL
In Figure 1, the scheme of a variable-speed wind turbine
based on doubly-fed induction generator is presented. Typi-
cally, the DC/ACconverter on the grid side is controlled tohave a constant Vdc and a unity power factor. The AC/DCconverter on the rotor side is usually controlled to have (a)
optimal power extraction from the wind and (b) a specified
reactive power at the generator terminal. Note that this con-
verter provides sinusoidal three-phase voltages at the slip rotor
frequency.
A. General considerations
A gearbox required to increase the angular speed from the
wind turbine to the induction-machine shaft is shown in Figure
1. It is characterized by its gearbox ratio, k, which is the ratiobetween the wind turbine speed wt and the machine shaftspeedms. Assuming a machine ofp poles, these speeds andthe electrical rotor speed of the machine, r, are related as
r =p
2ms=
p
2kwt (1)
Another important definition is the tip speed ratio, , whichis the ratio between the speed of a blade tip vtip [
ms
] and thewind speed vwind [
ms
]. IfR is the radius of the turbine, thetip speed ratio is
= vtipvwind
=wtR
vwind=
2k
p
rR
vwind(2)
Additionally, a crowbaran electronic deviceis some-times considered to protect the machine and converters from
grid disturbances. When the protection is activated, the
machine-rotor phases are short-circuited protecting them from
over-voltages and protecting the rotor-side converter from
over-currents. This crowbar protection is activated whenever
the rotor-current limit is surpassed [9]. Note that there are wind
turbines equipped withfault ride-throughcapability which can
withstand a fault during some cycles without using the crowbar
protection [6].
B. Assumptions
Ideal converters are considered and the dc voltage between
converters is assumed to be constant. Therefore, the conversionis lossless and the converters on the rotor- and grid-side are
modelled as a controlled-voltage source and a controlled-
current source, respectively. By power balance, the active-
power absorbed by the grid-side converter is the same than
the power injected by the rotor-side converter. Using this
active-power and the grid voltage, the current on the grid-side
converter is calculated. Other premises can be found in [10].
A scheme of the wind-power generation model is depicted in
Figure 2.
Fig. 1. Wind power generator scheme
Fig. 2. Schematic of the wind turbine generator model
C. Wind turbine model
The wind turbine model basically represents the relation
between the mechanical power extracted from the turbine,
Pwind, and the wind speed as [11]
Pwind=
1
2AwtCp(, )v
3
wind [W] (3)
where is the air density [ kgm3
],Awt is the wind turbine sweptarea [m2], vwind is the wind speed [
ms
] and Cp is the powercoefficient. Cp is dimensionless and depends on both the tipspeed ratio, , and the pitch angle, [degrees]. Using a fixedpitch angle, typical power curves in function of the wind speed
and the electrical rotor speed is depicted in Figure 3.
Considering manufacturer data and optimization techniques,
a power coefficient function has been derived for a variable-
speed wind turbine [12]. Using an intermediate parameter i
Cp(i, ) = 0.22
116
i 0.4 5
e
12.5i (4)
wherei =
1
+ 0.08
0.035
3 + 1
1
(5)
A torque expression is required to model the motion of the
rotatory massturbine, gearbox and machine shaft. Assume
that the power transmission from the wind turbine to the
machine shaft is lossless. Then, the torque in [Nm] at themachine shaft is
TM=Pwindms
= k
2R3Cp(, )v
2wind [Nm] (6)
-
8/12/2019 W. Sauer. Dynamic Modeling of Doubly Fed Induction
3/6
Fig. 3. Extracted power from the wind turbine for various wind speeds
Use the per-unit base of the induction-machine to define power
base, Sb, voltage base, Vb, and electrical speed base, b. Thetorque base can be calculated as Tb = Sb
p2b
. More details
about per unit system in electrical machines can be found in
[13][15]. As in the next sections all expressions are in per
unit, the torque equation in per-unit becomes
Tm=TMTb
=1
2
R2bSbr
Cp(, )v3wind [pu] (7)
D. DFIG model
A DFIG is simply a wound rotor induction machine where
the stator- and rotor-circuits are energized. Both stator and
rotor windings participate in the electromechanical energy
conversion. Consequently, aDFIGmodel is basically the same
than the model of an induction machine in which the rotor
voltages are supplied by an electric source, i.e., rotor side
AC/DC converter. This configuration allows the machine tooperate in a wide speed range.
The induction machine model requires the use of reference
frame theory. Basically, a model using a-b-c stator- and rotor-
phases is referred to a particular reference frame with twoorthogonal axis, namely, quadrature axis (qaxis) and directaxis (daxis). A stationary reference frame, a rotor referenceframe or a synchronously rotating reference frame can be
used [15]. The last is adopted in this article. Note that a
generator convention is used, i.e., the stator and rotor currents
are positive when they are leaving and entering the machine,
respectively. Likewise, a frame in which the q axis leadsthe d axis is considered.
1) Fifth order model: The stator and rotor circuits as well
as the equation of motion of the rotatory mass define the
following set of differential equations
1
s
dqsdt =Vqs+ RsIqs ds (8)
1
s
ddsdt
=Vds+ RsIds+ qs (9)
1
s
dqrdt
=Vqr RrIqr (s r)
sdr (10)
1
s
ddrdt
=Vdr RrIdr+(s r)
sqr (11)
2HDs
drdt
=Tm Te (12)
whereV,I,R and correspond to the voltages[pu], currents[pu], resistances[pu]and flux linkages[pu], respectively.Te=drIqr qrIdr is the electrical torque at the machine shaft[pu]. All variables and parameters are in per unit except r,s and HDD stands for DFIG.
The stator and rotor flux equations define the following set
of algebraic equations
qs = XsIqs + XmIqr ds= XsIds+ XmIdr (13)qr = XmIqs+ XrIqr dr= XmIds+ XrIdr (14)
whereXm is the mutual reactance between the stator and therotor, Xs = Xs + Xm is the stator reactance and Xr =Xr+ Xm is the rotor reactance. Xs and Xr are the statorand rotor leakage-reactance, respectively. All reactances are in
per unit.2) Third order model: In dynamics simulations, it has been
observed that stator dynamics are faster than rotor dynamics
[15][17]. In order to visualize it, solve for Idr and Iqr fromequation (14)
Iqr =Xm
XrIqs+
qr
XrIdr=
Xm
XrIds+
dr
Xr(15)
Multiply equations (10)(11) by XmRr
and replace Iqr andIdrto obtain
XrsRr
dXmXr
dr
dt =
XmRr
Vdr XmXr
dr X2mXr
Ids
+(s r)
s
XrRr
XmXr
qr
(16)
XrsRr
dXmXr
qr
dt =
XmRr
Vqr XmXr
qr X2mXr
Iqs
(s r)
s
XrRr
XmXr
dr
(17)
Define
T0 = XrsRr
Xs= Xs X2mXr
(18)
EqD =XmXr
dr E
dD= XmXr
qr (19)
whereT0 is the transient open-circuit time constant,X
s is the
transient reactance,EqD and E
dDare the quadrature and direct
axis rotor-voltages, respectively. For large machinesT0 1s
,
thus, stator dynamics are faster than rotor dynamics [16], [17].
Using a zero-order integral manifold to represent the stator
dynamics, the reduced-order machine model becomes
Vqs = RsIqs X
sIds+ E
qD (20)
Vds= RsIds+ XsIqs+ EdD (21)
T0dEqDdt
=EqD + (Xs X
s)Ids
+ T0
s
XmXr
Vdr (s r)E
dD
(22)
T0dEdDdt
= (EdD (Xs X
s)Iqs)
+ T0
s
XmXr
Vqr+ (s r)E
qD
(23)
-
8/12/2019 W. Sauer. Dynamic Modeling of Doubly Fed Induction
4/6
Fig. 4. Equivalent circuit of a DFIG connected to a grid
To obtain the stator algebraic equations (20)(21), use equa-
tions (13), (15) and (19). Rotor flux equations can be stated
in terms of the new variables EqD and E
dD as,
Idr=EqDXm
+XmXr
Ids Iqr = EdDXm
+XmXr
Iqs (24)
The third-order model of theDFIG is completely defined byequations (20)(24) plus the equation of motion (Equation 12).
Notice that multiplying Equation (21) by ej2 and adding
Equation (20), a phasor machine representation to calculate
stator algebraic-variables, given E
qD andE
dD, is obtained.
EqD jE
dD= (Rs+jX
s)(Iqs jIds) + Vqs jVds (25)
3) Machine and grid connection: Consider aDFIGsend-ing active power, through a short-length transmission line, to
an infinite busthe line and the infinite bus may represent
the Thevenins equivalent of a more complex grid. In order to
integrate the generator model in power system modeling, it is
necessary to relate the synchronously rotating reference-frame
variables to a phasor. Applying the proper transformation to
the machine stator voltage, it is obtained that the terminal-
voltage phasor at phase-a is VD = VDejD =Vq jVd [15].Consequently, the machine model to be used in power system
modeling is defined by equation (25) (see Figure 4). The grid-
side converter is represented by a controlled-current source. As
this converter does not absorb reactive power, the controlled-
currentIGC GC stands for grid-side converter is,
IGC=Protor
V
D
=VqrIqr+ VdrIdr
VDejD (26)
4) Decoupled control of active and reactive power:Assume
that thed-axis is oriented along the stator flux axis, i.e., s=ds withqs = 0. Besides, neglectRs and use equations (8)(9) in steady-state to get Vds= qs = 0and Vqs = ds= VD.Consider the flux equation (13) to obtain
Iqs =XmXs
Iqr Ids=XmXs
Idr VDXs
(27)
Then, the complex power leaving the generators stator is
P+jQ = (VdsIds+ VqsIqs) +j(VqsIds VdsIqs) (28)
=
XmXs
VsIqr
+j
VD
XmIdr VDXs
(29)
It turns out that the control of active and reactive power can
be performed independently varyingIqr andIdr, respectively.
Fig. 5. Power reference for the rotor speed controller
Fig. 6. Rotor speed controller
E. Controllers
1) Rotor speed controller: The rotor speed controller is
designed to extract maximum power from the wind turbine.As stated in Section II-C, the power extraction depends on
both the wind speed, which is uncontrollable, and the tip
speed ratio, which is controllable. Actually, depends onr as stated in equation (2). Therefore, controlling r wecan move along the power curve for a given wind speed to
maximize the power. Joining the maximum-power points for
every given wind speed of figure 3, we can obtain a one-
to-one correspondence between the optimal power and the
rotor speed. Additionally, cut-in and maximum speed due
to converter ratings [18] must be taken into account. The
cut-in point corresponds to the minimum speedtypically
0.7 ratedrequired to produce power. If the speed is
below this point, theDFIGis shutdown. The maximum speedpoint corresponds to the maximum power that the DFIGcan produce. If the speed is above the maximumtypically
1.2 ratedthe turbine aerodynamic must be modified toreduce the power extracted from the wind turbine. To this
end, pitch angle control must be performed. It turns out that
the power reference is piece-wise defined as in Figure 5. A
tracking control capable to follow Pref is used. Based onequation (29), a PI controller with an internal Iqr-controlloop is considered [19] (Figure 6).
2) Pitch angle controller:The pitch angle control modifies
the aerodynamic efficiency of the turbine in order to limit
the power production to the rated power. Note that the pitch-
angle time-constant is quite high depending on both the sizeof the turbine blades and economical/practical limitations of
the blade drives [10]. A P I controller with an internal pitch-angle control loop is considered [19] (see figure 7). ref ischosen such that the pitch-angle is around zero when the speed
is located at the optimal tracking curve and it is positive and
less than 90o when the maximum speed is binding.
3) Reactive power controller: According to the Federal
Energy Regulatory Commission, USA, [20] large wind plants
seeking to interconnect to the grid need to maintain a power
-
8/12/2019 W. Sauer. Dynamic Modeling of Doubly Fed Induction
5/6
Fig. 7. Pitch angle controller
Fig. 8. Reactive power controller
factor within the range of 0.95 leading to 0.95 lagging,
measured at the high voltage side of the substation trans-
formers. Although, this requirement is mandatory only in
the case that a power factor out of this range jeopardizes the
security and reliability of the system, it makes a precedent
for control capability requirements. In this article, a reactive-
power tracking control is used (see Figure 8). Based onequation (29), a PI controller with an internal Idr-currentcontrol loop is considered [19].
III. TEST SYSTEM SIMULATION
Consider the four-bus system of Figure 9. Assume an expo-
nential model for the load asPL= P0VkpL andQL= Q0V
kqL .
The synchronous generator (SG) is represented by a two-axis
model [14] and the transient reactance at both quadrature
and direct axis are equal, i.e., Xd = X
q. An IEEE Type 1
Exciter and a linear speed governor without droop [21] are
considered. Besides, assume that the wind speed is such that
the rotor speed is above and below of the cut-in and maximum
speed, respectively. Consequently, the rotor speed controller isoperating in the optimal tracking curve and no pitch control is
required. The optimal curve is defined as Pref=C3r [pu].The system is modelled by a set of differential algebraic
equations (DAE) of the Hessenberg index-1 form [22]. Ingeneral, this set ofDAE can be written as dx
dt = f(x,y,),
0 = g(x,y,) where x Rnd1 is the vector of differentialvariables, y Rna1 is the vector of algebraic variables and Rnp1 is the vector of parametersnd, na and np arethe number of differential variables, algebraic variables and
parameters, respectively. gy = g(x,y,)
y is nonsingular along
the solution of the DAE.In order to perform a small signal stability analysis, a linear
approximation of the set ofDAEhave to be obtained. Then,
Fig. 9. Four-bus test system
Fig. 10. Trajectory of dominant eigenvalues withvwind = 10 [m/s] and0.32 P0 1
Fig. 11. Trajectory of dominant eigenvalues withvwind = 12 [m/s] and0.32 P0 1
around an equilibrium point,
x= Ax + By (30)
0 = Cx + Dy (31)
Using Krons reduction, the systems equations can be reduced
tox= (ABD1C)x= Asysx. Eigenvalues ofAsysdetermine the stability of the operating point. It is desired to
study the system stability as the load is increased. Then, the
eigenvalues ofAsys are calculated at every equilibrium point.Note that if all eigenvalues are located on the left half of the
complex plane, i.e., all eigenvalues have a negative real part,
then the equilibrium point is stable. If some eigenvalue has
a positive real part, then the equilibrium point is unstable.
The point at which a complex pair of eigenvalues crosses the
imaginary axis to the right half plane, while the other remain
-
8/12/2019 W. Sauer. Dynamic Modeling of Doubly Fed Induction
6/6
on the left half plane, is called a Hopf Bifurcation (HB)point. The system becomes critically stable. The eigenvalues
trajectories are shown in Figures 10 and 11 for a wind speed
of 10 [m/s] and 12 [m/s], respectively. The load is variedfrom P0 = 0.32 [pu] to P0 = 1 [pu].
The incidence of eigenvalues on state variables can be
estimated by using participation factors [14]. For the two
presented cases, the pathway of eigenvalues are notably differ-
ent. This difference may be related to parameter sensibilities
on system stability [23] due to the increase of generated
power of the DFIG. Note that the eigenvalues associatedto r, and Pm do not move when load and wind speedare varied. The eigenvalues that cross the imaginary axis are
associated to states (Eq,E
d,Efd ,VR) and (E
q ,Efd,VR) whenthe wind speed is 10 [m/s] and 12 [m/s], respectively. TheHB points for these two cases happen at P0 = 0.941 [pu]andP0 = 0.971 [pu], respectively. This does not considerablydiffer from the modal behavior of a system with a conventional
synchronous generator [14], [24]. For all wind speeds and
loading simulated, up to the HB point, eigenvalues associated
to theDFIG state-variables are stable. From these results, itseems that the dynamic of the wind-power generator does nothave a serious impact on the system behavior. Probably the
major interaction between the machines is the interchange of
active-power due to wind variations. Thus, as future work, a
reduced-order equivalent model that captures the active-power
dynamic of the wind-power generator is going to be derived.
IV. CONCLUSION
A complete dynamic model appropriate for power system
analysis is presented. This article contains several details such
as modeling assumptions, derivation of a third order model for
theDFIGand its controller descriptions, thus, it can be used as
a tutorial for students and engineers that are new in this area.A four-bus system with one DFIG and one SG is considered
to perform a small signal stability analysis. The results reveal
that the state-variables associated to the unstable modes around
theHB point are those associated to the SG voltage controller.
Based on the eigenvalue pathways, the modal behavior does
not considerably differ from the modal behavior of a system
with only one SG.
APPENDIX
Parameters are in per unit unless otherwise is specified.
Synchronous machine
Xd = 2.2, Xq = 1.76, Xd = 0.2, Xq = 0.2, Td0 = 8 [s],Tq0 = 1 [s], H= 10 [s], KE= 1, TE= 0.7 [s], KF = 0.03,TF = 1 [s], KA = 200, TA = 0.04 [s], TG = 5 [s],Vref= 1.0078
Network
R1 = 0.03 and X1 = 0.10 (Line 1); R2 = 0.10 andX2 = 0.10 (Line 2); pv = 0 and qv = 0 (Load Parameters),XT = 0.07 (Transformer)
Doubly-fed induction generator
Xm = 3.5092, Xs = 3.5547, Xr = 3.5859, s =120 [rad/s], Rs = 0.01015, Rr = 0.0088, H = 2 [s],p = 4, = 1.225 [kg/m3], R = 15 [m], Sb = 1 [MVA],C = 3.2397 109 [s3/rad3], k = 1/45, KP1 = KP2 =KP3 = KP4 = 1, KI1 = KI2 = KI3 = KI4 = 5, Qref= 0
REFERENCES
[1] World Energy Outlook 2006. International Energy Agency, 2006.[2] L. Schleisner, Life cycle assessment of a wind farm and related
externalities, Renewable Energy, vol. 20, no. 3, pp. 279288, 2000.[3] N. Golait, R.M. Moharil, P.S. Kulkarni, Wind electric power in the
world and perspectives of its development in india, Renewable andSustainable Energy Reviews, vol. 13, pp. 233247, 2009.
[4] World Wind Energy Report 2008. World Wind Energy Association, 2008.[5] Z. Saad-Saoud, N. Jenkins,, Simple wind farm dynamic model, IEE
Proceedings: Generation, Transmission and Distribution, vol. 142, no. 5,pp. 545548, 1995.
[6] T. Ackermann, Wind power in power systems. John Wiley and Sons,Ltd., 2005.
[7] R. Scherer, Blade design aspects, Renewable Energy, vol. 16,pp. 12721277, 1999.
[8] R. Zavadil, N. Miller, A. Ellis, E. Muljadi, E. Camm, B. Kirby, Queuingup: Interconnecting wind generation into the power system,IEEE Powerand Energy Magazine, vol. 5, no. 6, pp. 4758, 2007.
[9] M.V.A. Nunes, J.A. Pecas Lopes, H.H. Zurn, U.H. Bezerra, R.G.Almeida, Influence of the variable-speed wind generators in transientstability margin of the conventional generators integrated in electricalgrids,IEEE Transactions on Energy Conversion, vol. 19, no. 4, pp. 692701, 2004.
[10] J.G. Slootweg, Wind power modelling and impact on power systemdynamics. 2003. Ph.D. Thesis, Delft University of Technology,Netherlands.
[11] S. Heier, Grid integration of wind energy conversion systems. JohnWiley and Sons Ltd., 1998.
[12] J.G. Slootweg, H. Polinder, W.L. Kling, Dynamic modeling of a windturbine with doubly fed induction generator, Proceeding of IEEE PES,Summer meeting, vol. 1, pp. 664649, 2001.
[13] P. Kundur, Power system stability and control. McGraw-Hill, 1994.[14] P. W. Sauer and M. A. Pai, Power systems dynamics and stability.
Prentice Hall, Upper Saddle River, NJ, 1998.[15] P.C. Krause, O. Wasynczuk, S.D. Sudhoff, Analysis of Electric Machin-
ery and Drive Systems. Wiley-IEEE Press, 2nd ed., 2002.[16] S. Ahmed-Zaid, M. Taleb, Structural modeling of small and large
induction machines using integral manifolds, IEEE Transactions onEnergy Conversion, vol. 6, no. 3, pp. 529535, 1991.
[17] E. Drennan, S. Ahmed-Zaid, P.W. Sauer, Invariant manifolds andstart-up dynamics of induction machines, Proceeding of the 21th
Annual North American Power Symposium, pp. 129138, October 1989.University of Missouri-Rolla by IEEE Power Engineering Society.
[18] J. Ekanayake, L. Holdsworth, N. Jenkins, Control de DFIG windturbines, IEE Power Engineer, vol. 17, no. 1, pp. 2832, 2003.
[19] R.G. de Almeida, J.A. Pecas Lopes, Participation of doubly fed induc-tion wind generators in system frequency regulation,IEEE Transactionson Power Systems, vol. 22, no. 3, pp. 944950, 2007.
[20] Federal Energy Regulatory Commission, USA, Standard interconnec-tion agreements for wind energy and other alternative technologies,http://www.ferc.gov.
[21] W.D. Rosehart, C.A. Canizares, Bifurcation analysis of various power
system models,Electrical Power and Energy Systems, vol. 21, pp. 171182, 1999.
[22] Brian Fabien,Analytical System Dynamics: Modeling and Simulation.Springer, 1st ed., 2008.
[23] P. Goncalves, Behaviour modes, pathways and overall trajectories:eigenvector and eigenvalue analysis of dynamic systems, System Dy-namic Review, vol. 25, no. 1, pp. 3562, 2009.
[24] H. Pulgar-Painemal, P.W. Sauer, Bifurcations and loadability issuesin power systems, IEEE Powertech Conference, 2009. Bucharest,Romania.