Electric Power Systems Research 56 (2000) 121 – 127

www.elsevier.com/locate/epsr

A third order model for the doubly-fed induction machine

Andrés Feijóo *, José Cidrás, Camilo Carrillo
Departamento de Enxeñerı́a Eléctrica, Uni6ersidade de Vigo, E.T.S.E.I.M., Lagoas-Marcosende, 36280 Vigo, Spain

Received 15 September 1999; received in revised form 17 February 2000; accepted 6 March 2000

Abstract

Induction generators are generally simulated by means of a well-known model described by Brereton et al. [1], based on the
induction motor equations derived by Stanley [2]. In this model the possibility of opening the rotor circuit in order to inject a
voltage source is not taken into account, although there are other models where it is dealt with [3]. This paper presents an
alternative way of obtaining the mentioned model and introduces the possibility of modeling voltage sources in the rotor circuit,
which can be very useful when simulating some generating schemes, such as variable speed asynchronous wind turbines. © 2000
Elsevier Science S.A. All rights reserved.

Keywords: Induction machine models; Stator; Rotor

1. Introduction 2. The third order models of the induction machine

Induction generators are the most commonly used
electric generators in industry, and they are particularly
common in some renewable energy applications such as
wind turbines (WTs). Several models have been pro-
posed to simulate them. The model proposed by Brere-
ton et al. is of great use when induction generators
must be dynamically simulated. This model is a very
valid one when the machine works with its rotor short-
circuited, that is, the rotor voltage having a value of 0.
The trend for future years seems to be the use of
variable-speed WTs, in order to achieve an optimiza-
tion of the generated power profiles. Several solutions
have been proposed, such as the use of synchronous
generators linked to the electrical network through
electronic devices, and doubly-fed induction generators,
which this paper is concerned with. The simulation of
doubly-fed induction generators has been dealt with in
many papers, and generally involves handling a set of
differential equations related to the rotor and stator
currents, fluxes and voltages. The purpose of this paper
is to obtain a set of more simplified equations than
those generally used.
* Corresponding author. Tel.: +34-986-812600, fax: + 34-986-
812173.
E-mail addresses: afeijoo@uvigo.es (A. Feijóo), jcidras@uvigo.es
(J. Cidrás).
First, the matrices ks and kr [4,5] will be assumed,
which allows one to formulate a change of variables
transforming them from the three-phase variables sys-
tem to an arbitrary reference frame. In this case the
new reference frame is rotating at synchronous speed.
Both matrices can be seen in Appendix A.
The following notation will be employed:
“ Ir123 is a vector whose components are the per phase
rotor currents (I Tr123 = (Ir1Ir2Ir3)), and Irqd0 (I Trqd0 =
(IrqIrdIr0)) is a vector whose components are the rotor
currents seen from the dq0 axes. The superindex T is
used for notating transpose matrices.
“ Isabc and Isqd0 are the stator currents in both reference
frames.
“ Vr123 and Vrqd0 are the rotor voltages.
“ Vsabc and Vsqd0 are the stator voltages.
“ 8r123 and 8rqd0 are the rotor fluxes.
“ 8sabc and 8sqd0 are the stator fluxes.
The following equations express the relationship be-
tween both reference frames:
Vrqd0 = krVr123 [ Vr123 = k − 1
r Vrqd0 (1a)
Vsqd0 = ksVsabc [ Vsabc = k − 1
s Vsqd0 (1b)
Irqd0 = krIr123 [ Ir123 = k − 1
r Irqd0 (1c)
Isqd0 = ksIsabc [ Isabc = k − 1
s Isqd0 (1d)

0378-7796/00/$ - see front matter © 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 3 7 8 - 7 7 9 6 ( 0 0 ) 0 0 1 0 3 - 6
122 A. Feijóo et al. / Electric Power Systems Research 56 (2000) 121–127

8rqd0 =kr8r123 [ 8r123 =k −

r 8rqd0
1

 
(1e) Á dk − 1
Â
r Ád8rqd Â
8sqd0 =ks8sabc [ 8sabc =k −1
8sqd0 (1f ) Ãkr dt 0 Ã
8rqd à Ã
s
+Ã −1 Ã + Ã dt Ã
à 0 dk 8 à 0 Ã
ks s à sqd Ä Å
where the matrices kr and ks are given in Appendix A, Ä dt Å
where both reference frames can also be seen.
(4)
In order to obtain the model, the following equa-
tions, valid for the rotor and stator electrical circuits, The rotor and stator fluxes can be written as func-
must be taken into account: tions of their currents, as follows:

      
Vr123
=
r 0 Ir123
+
d 8r123
(2)
  
8rqd L
= rr
Lrs  
Irqd
(5)
Vsabc 0 R Isabc dt 8sabc 8sqd Lrs Lss Isqd

where:

 
where
lrr 0
ÁRR 0 Lrr = (6a)
0 Â ÁRs 0 0Â 0 lrr
à à à Ã
r = Ã 0 RR 0 Ã R= Ã 0 0 Ã,
 
and Rs
à à à Ã
Ä0 0 RR Å Ä0 0 Rs Å lss 0
Lss = (6b)
0 lss

 
RR is the rotor resistance and Rs the stator resistance.
The machine is assumed to be balanced. Taking this lrs 0
assumption into account, the component ‘0’ can be Lrs = (6c)
0 lrs
neglected. So, from now on only the ‘d’ and ‘q’ compo-
nents are employed. and:
Changing the reference frame by means of the “ lrr = l2 + lrs, and l2 is the leakage inductance in a
matrices phase of the rotor.

   
“ lss = l1 + lrs, and l1 is the stator leakage inductance.
kr 0 kr 0 −1 “ lrs is the magnetizing inductance.
and ,
0 ks 0 ks From Eq. (5) the following expression for the rotor
currents can be obtained:
the following is obtained:
Irqd = L − 1
rr (8rqd − LrsIsqd ) (7)

    
Vrqd
=
r 0 Irqd
Now the rotor short-circuit constant is defined as
Vsqd 0 R Isqd
T%0 = lrr/r. So from Eq. (4) the following equation can

+

kr 0 d   k−
r
1
0  
8rqd
(3)
be obtained:

0 ks dt 0 k− 1
8sqd d8rqd dk − 1 1
s
= − kr r 8rqd − (8rqd − LrsIsqd ) (8)
dt dt T%0
On the other hand, from Eqs. (5) and (7) the stator
flux can also be written as:
2.1. Model for the induction machine with
8sqd = LrsIrqd + LssIsqd
short-circuited rotor
rr 8rqd + (Lss − LrsL rr Lrs)Isqd
= LrsL − 1 −1
(9)
In the case of an induction generator with short-cir-
cuited rotor, which is a common solution in constant and also from Eqs. (4) and (9) the next equation for the
speed WTs , the previous equations become the follow- stator voltage is valid:
ing, by neglecting the stator transients d8sqd /dt = 0:
Vsqd = RIsqd

    
0
=
r 0 Irqd + ks
dk −
dt
s
1

rr 8rqd + (Lss − LrsL rr Lrs)Isqd )

(LrsL − 1 −1

Vsqd 0 R Isqd
(10)
 A. Feijóo et al. / Electric Power Systems Research 56 (2000) 121–127 123

For the sake of simplicity the following variables and

Now, defining the following complex numbers: V( s =
constants are defined, 8%rqd =LrsL − rr 8rqd, L%= Lss −
1
Vsq + jVsd, E( % =E%q + jE%d and I( s = Isq + jIsd, the equa-
LrsL −
rr
1
L rs and L = L rs With regard to this and to Eq.
tions describing the model are:
(8) the following equations can be written:
dE( %
 
1
= −jsvsE( %− (E( %− j(X−X%)I( s) (16a)
d 8’rq dt T%0
dt 8’rd V( s − E( % =(Rs + jX%)I( s (16b)

=−
 0 − svs  
8%rq These equations are the same as those obtained by
Brereton et al. [1], and are generally accepted and
svs 0 8%rd
widely used for transient stability studies, where induc-

−
1   
8’rq
−
l− l% 0  
Isq
(11)
tion machines are present. The representation of the
machine as an electrical circuit can be seen in Fig. 1.
T%0 8’rd 0 l− l% Isd 2.2. The proposed model for the doubly-fed induction
machine
And with the same variables and taking Eq. (10) into
account: The case of the doubly-fed induction machine is

  
Vsq
−
0 − vs     
8%rq
=
R 0 Isq
(12)
similar to the previous one, with the difference that in
Eq. (4), the rotor voltage is different from 0. The same
Vsd vs 0 8%rd 0 R Isd assumption is made, which means that the stator tran-
sients can be neglected. The following is valid when
where the values of l and l% are derived in Appendix B assuming waves of the fundamental frequency only. If
and the expression of slip has been taken into account this assumption were not made, the derivation of the
(s = vs −V/vs). model would be more complex, due to the difficulty of
Now, applying the next change of variables: dealing with harmonic currents when dynamic pro-

E%qd =
 0 −vs 
8%rqd (13)
cesses are being analyzed. With regard to this, it can be
said that the rotor current harmonics, when reflected to
vs 0 the stator, induce distortion harmonic currents in the
stator windings. These currents have relatively low
the equations can be finally expressed as: magnitudes, and are not so serious if the rotor is

 
d E%q
connected to a high-frequency PWM converter. This
issue is discussed in [6].

    
dt E%d

  
Vrqd r 0 Irqd
=
0 − svs E%q Vsqd 0 R Isqd
=−
svs 0 E%d

 
Á dk − Â
    
1
k
r
0 Ã Ád8rqd Â
à r dt 8rqd à Ã
1 E%q 0 − (X −X%) Isq +Ã + Ã dt Ã
− − (14) −1 Ã
à 0 dk 8 à 0 Ã
T%0 E%d X −X% 0 Isd ks s à sqd Ä Å
Ä dt Å
       
Vsq E%
− q =
R 0 Isq
+
0 − X%  
Isq (17)
Vsd E%d 0 R Isd X% 0 Isd
In this case, by performing the following change of
(15)
variables:

V%rqd = LrsL − 1
rr Vrqd (18)

from Eq. (4) the rotor flux equations become:

    
d 8%rq
=
V%rq
−
0 − svs  
8%rq
dt 8%rd V%rd svs 0 8%rd

−
1   
8%rq
−
l− l% 0  
Isq
(19)
Fig. 1. Induction machine dynamic model. T%0 8%rd 0 l− l% Isd
124 A. Feijóo et al. / Electric Power Systems Research 56 (2000) 121–127

which coincides with V( r = Vrq + jVrd.

Also by doing dE( %/dt = 0, the expression of the
voltage source E( % in steady-state can be obtained and
expressed as follows:
Xm
E( %= (V( %r − RRI( r) (24)
s(X2 + Xm)
Fig. 2. Doubly-fed induction machine steady-state model. which, obviously, in the case that the machine is not
doubly-fed, can be written as:
so, including Eq. (13) the following system is obtained:

       
Xm
E( %= − RRI( r (25)
d E%q 0 −vs V%rq 0 −svs E%q s(X2 + Xm)
= −
dt E%d vs 0 V%rd svs 0 E%d

−
1  
E%q
3. Electromechanical equations and rotor currents

T%0 E%d The mechanical power, Pm, electrical power, PE, and

  
slip, s, are related by means of the electromechanical
0 − (X −X%) Isq [7]:
− (20)
X− X% 0 Isd Pm ds
−PE = − 2H (26)
1− s dt
and from Eq. (3):

         
where H is the inertia constant, proportional to the
Vsq E% R 0 Isq 0 − X% Isq moment of inertia (J), and defined as the relationship
− q = + between the energy of the machine at synchronous
Vsd E%d 0 R Isd X% 0 Isd
speed and its rated power.
(21) The term PE can be calculated as a function of E( % and
I( s as follows:
Finally, the previous equations can be written in a
simpler way as follows, by defining the complex num- PE = − Re{E( %I( *}
s (27)
ber V( %r =V%rq +jV%rd : where Re denotes the real part of a complex number.
dE( % 1 When the steady-state is reached, the value of PE so
= jvsV( %r −jsvsE( %− (E( % −j(X −X%)I( s) (22a) calculated, in the case of the rotor cascade angle having
dt T%0
a value of p, coincides with the following expression:
V( s − E( %= (Rs + jX%)I( s (22b)
RR V%rIr
This model is valid for the doubly-fed induction PE = − I 2r − (28)
s s
machine, and the electrical circuit of Fig. 1 can repre-
sent the machine. So, the case of the induction machine and in the case of the machine with short-circuited
having short-circuited rotor is a special case of the rotor, with:
general case, considering the doubly-fed induction ma- RR
chine as the general case. PE = − I 2r (29)
s
2.3. Steady-state If the angle is different from p or 0 the following
expression should be written:
In order to obtain the steady-state models, it may be RR V¦r Ircosf
assumed that dE( %/dt =0. By doing this in the general PE = − I 2r + (30)
s s
model, the steady-state circuit can be given for the
induction machine. This circuit is represented in Fig. 2, where f is the angle between rotor voltage and current.
and the reference of the rotor current is chosen to be in The following observation must be made, that con-
concordance to that chosen for the dynamic model. cern the way of calculating the rotor current. There are
The equivalence between steady-state and dynamic some differences between steady-state and dynamic
model is achieved taking the following into account (see models. For the steady-state model, according to the
Appendix C): references given in Fig. 2, the equation is as follows:
X2 + Xm (Rs + j(X1 + Xm))V( ¦r − jsXmV(
V( ¦r = ( r
V% (23) I( r = (31)
Xm (Rs + j(X1 + Xm))(RR + js(X2 + Xm))+sX 2m
 A. Feijóo et al. / Electric Power Systems Research 56 (2000) 121–127 125

As for the dynamic model, from Eq. (7), taking into 4. Simulations
account the fact that 8%rqd =LrsL −
rr 8rqd and also Eq.
1

(13) the rotor current can be calculated as: Simulations have been carried out with the model of
an induction generator corresponding to a WT having
E( % Xm
I( r = − I( s (32) the parameters given in Appendix D. It is a horizontal
jXm X2 + Xm axis three-blade asynchronous WT.
which only coincides with that given by Eq. (31) when For the simulation, the mechanical power is calcu-
the machine reaches the steady-state. In spite of this lated from the wind speed according to the equation
observation, the rotor current in the dynamic model
must never be calculated during the simulation. Pm = 1/2rAU 3cp [8], where r is the density of air (1.225
kg m − 3), A is the area swept by the rotor, U is the wind
speed in m s − 1 and cp is the power coefficient. The
power coefficient depends on the tip speed ratio, l,
which is the relationship between the speed of the
blades and the wind speed. Generally, the power coeffi-
cient can be expressed as a polynomial, cp =  ail i. The
machine simulated here has the features described in
Appendix D.
The control system was not simulated, as it is not the
purpose of the paper to study how it works. So, the
simulation consists of the starting of the machine and
the injection in the rotor circuit, 4 s later, of a voltage
source with a value of 0.01 p.u., and an angle of p
radians with respect to the rotor current shown in Fig.
2. As a consequence, the evolution of the machine is
Fig. 3. Evolution of slip during starting and changing rotor voltage.
towards a new starting point.
In the simulation, the constant voltage injected in the
rotor is defined as proportional to the stator voltage, V%r
(kV). The electronic devices allow control not only of
the rms value but also of the angle of the rotor voltage.
From the point of view of the machine, this can be
simulated as a voltage source with a magnitude and an
angle.
In Fig. 3 the evolution of the slip can be seen. In Fig.
4 the evolution of mechanical and electrical power can
be seen. The change in the value of the mechanical
power is due to the fact that, with the change of the
slip, the tip speed ratio reaches another value and,
consequently, the power coefficient and the electrical
Fig. 4. Evolution of real and mechanical power during starting and power do the same. The steady-state, real power-slip
changing rotor voltage. curves are shown in Fig. 5. In this curve the evolution
of electrical power and slip can also be seen. It can be
observed here, that when a new steady-state is reached,
the results obtained with the dynamic model are coinci-
dent with those obtained with the steady-state one.

5. Conclusions

A model is presented in order to make it easier to

Fig. 5. Real power against slip curves for the doubly-fed induction
machine in dynamic and steady-state.
dynamically simulate doubly-fed induction machines.
Simulations are presented to prove that the model is
adequate from the point of view of steady-state. The
advantage of the model is that it allows one to deal
with the machine with only one differential equation in
the electrical part.
126 A. Feijóo et al. / Electric Power Systems Research 56 (2000) 121–127

Á Â

   
à cos vst − sin vst 1Ã
à 2p 2p Ã
k− 1
= Ãcos vst− 3 − sin vst− 1Ã

   
s 3 (36)
à Ã
Ãcos v t+ 2p − sin vst+
2p
1Ã
Ä s
3 3 Å
Á Â

   
à cos Dvt −sin Dvt 1Ã
à 2p 2p Ã
k− 1
= Ãcos Dvt − 3 − sin Dvt − 1Ã

   
r 3
Fig. 6. Stator, rotor and dq reference frames.
à Ã
Ãcos Dvt + 2p − sin Dvt +
2p
1Ã
Ä 3 3 Å
One of the assumptions made to derive the model is (37)
the fact that the harmonics of electrical magnitudes are
neglected. This assumption is generally accepted in Á Â

   
à − sin vst − cos vst 0Ã
many applications involving electrical machines and
−1
dk s à 2p 2p Ã
electronic devices. Nevertheless, the authors think that =vs à −sin vst− 3 − cos vst− 0Ã

   
dt 3
it would be interesting to study a model able to deal à Ã
with harmonics. Ã − sin v t+ 2p − cos vst+
2p
0Ã
Ä s
3 3 Å
(38)
Appendix A. Matrices used for the transformation of the
−1
reference frames dk r
dt
In the matrices the following notation and the refer- Á Â

   
ence frames shown in Fig. 6 are used: Ã − sin Dvt − cos Dvt 0Ã
à 2p 2p Ã
Dv = vs − V (33) = Dv à −sin Dvt − 3 − cos Dvt − 0Ã

   
3
à Ã
where: Ã − sin Dvt + 2p 2p
“ vs is the synchronous speed.
− cos Dvt + 0Ã
Ä 3 3 Å
“ V = pVR, with p being the number of pole pairs and
VR the rotor speed. Á Â

   
à − sin Dvt − cos Dvt 0Ã
à 2p 2p Ã
2 = svs à − sin Dvt − 3 − cos Dvt − 0Ã

   
ks = 3
3 Ã Ã

    Ã − sin Dvt + 2p 2p
Á Â − cos Dvt + 0Ã
à cos vst 2p 2p Ã Ä 3 3 Å
cos vst − cos vst +

   
à 3 3 à (39)
à 2p 2p Ã
à −sin vst − sin vst − −sin vst + à Á 0 vs 0 Â
à 3 3 à dk − 1 à Ã
= Ã − vs 0 0 Ã
s
ks (40)
à 1 1 1 à dt à Ã
à 2 2 2 Ã Ä 0 0 0Å
Ä Å
(34) Á 0 Dv 0Â Á 0 svs 0 Â
dk − 1 Ã Ã Ã Ã
= Ã − Dv 0 Ã = Ã − svs 0 0 Ã
r
2 kr 0
kr = dt à à à Ã
3 Ä 0 0 0Å Ä 0 0 0Å
Á
à cos Dvt 
cos Dvt −
2p  
cos Dvt +
2p  Â
Ã
(41)

   
à 3 3 Ã
à 2p 2p Ã
à −sin Dvt −sin Dvt − −sin Dvt + Ã
à 3 3 à Appendix B. Definition of variables
à 1 1 1 Ã
à 2 2 2 Ã
Ä Å
(35) L%= Lss − LrsL − 1
rr Lrs
 A. Feijóo et al. / Electric Power Systems Research 56 (2000) 121–127 127

     
Á1 Â stator currents in both models. So, taking the steady-
0Ã
lss 0 lrs 0 Ãlrr lrs 0 state model into account, the rms stator current can be
= − Ã Ã expressed, according to the references of Fig. 2, as:
0 lss 0 lrs à 1à 0 lrs
0
Ä lrr Å
(RR + js(X2 + Xm))V( − jXmV( ¦r
I( s = (48)
Álrrlss −l 2
rs  (Rs + j(X1 + Xm))(RR + js(X2 + Xm))+sX 2m
à l 0 Ã
=Ã rr
2Ã (42)
à lrrlss − l rs à On the other hand, with regard to the dynamic
0
Ä lrr Å model, by making dE( %/dt =0 in Eq. (22a) in order to
get a steady-state equivalent model, and taking into
Taking into account the fact that lrr =l2 +lrs and account Eq. (22b) for substituting the value of E( % in
lss = l1 +lrs the following equation can be written: steady-state as a function of V( and I( s, the expression of
the rms value for the stator current can be written as:
lrrlss − l 2rs l2lrs
l’ = = l1 + (43) (RR + js(X2 + Xm))V( − j(X2 + Xm)V( %r
lrr l2 +lrs I( s = (49)
(Rs + j(X1 + Xm))(RR + js(X2 + Xm))+sX 2m

 
and:
l2lrs XX As both expressions must coincide, Eq. (23) must be
X%= l%vs = l1 + vs =X1 + 2 m (44)
l2 + lrs X2 +Xm satisfied.
On the other hand:

L= Lss =
 lss 0  
l +l
= 1 rs
0  (45)
Appendix D. Parameters of the induction generator
used in the simulations
0 lss 0 l1 +lrs

So, the inductance l can be defined as: Following the notation given in Fig. 2, the machine
used in the simulations has the following parameters,
l= lss =l1 +lrs (46) RR = 0.00612 p.u., Rs = 0.00571 p.u., X1 = 0.06390 p.u.,
and the reactance X as: X2 = 0.18781 p.u., Xm = 2.78000 p.u., H= 3.05 p.u.,
gear box ratio, 1:44.38, rotor diameter, 15.2 m, rated
X = lvs = (l1 +lrs)vs =X1 +Xm (47) power, 350 kW, and rated voltage, 660 V
The power coefficient curve for this machine is given
by the equation cp = 10
i = 0 ail , where the coefficients ai
i

are given in Table 1

Appendix C. Equivalence of voltage sources in the
dynamic and steady-state model
References
The equivalence of voltage sources representing the
rotor cascade can be deduced from the expressions of [1] D.S. Brereton, D.G. Lewis, C.C. Young, Representation of induc-
.
tion motor loads during power system stability studies, AIEE
Trans. 76 (1957) 451 – 461.
Table 1 [2] H.C. Stanley, An analysis of the induction machine, AIEE Trans.
Constants of the polynomial cp = f(l) for the machine employed in 57 (1938) 751 – 755.
the simulations [3] P.C. Krause, O. Wasynczuk, M.S. Hildebrandt, Reference frame
analysis of a slip energy recovery system, IEEE Trans. Energy
Coefficient Value Conversion 3 (2) (1988) 404 – 408.
[4] P.C. Krause, Analysis of electric machinery, McGraw-Hill, New
a0 0.0914344959 York, 1986.
a1 −0.486804621 [5] M. Pavella, P.G. Murthy, Transient Stability of Power Systems,
a2 0.944258742 Wiley, New York, 1994.
a3 −0.909776507 [6] Z.M. Salameh, L.F. Kazda, Analysis of the double output induc-
a4 0.488200324 tion generator using direct three-phase model. Part II: harmonic
a5 −0.153325541 analysis, IEEE Trans. Energy Conversion 12 (2) (1987) 182–182.
a6 0.0295642442 [7] A.E. Feijóo, Influencia de los parques eólicos en la seguridad
a7 −0.0035602243 estacionaria y calidad de onda de redes eléctricas de gran dimen-
a8 0.000261703947 sión, Ph.D. thesis (in Spanish), Escola Técnica Superior de Enx-
a9 −1.07606521e-05 eñeiros Industriais, Universidade de Vigo, Spain, 1997.
a10 1.8992284e-07 [8] L.L. Freris, Wind energy conversion systems, Prentice Hall, New
York, 1990.