You are on page 1of 8

NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 1

Performance Analysis of a Brushless


Double Fed Cage Induction Generator
F. Rncos1, R.Carlson2, A. M. Oliveira2, P. Kuo-Peng2, N Sadowski2

1
WEG S.A., C.P. 3000, 89250-900 Jaragu do Sul SC BRAZIL fredemar@weg.com.br
2
GRUCAD-UFSC, C.P. 476, 88040-900 Florianpolis SC BRAZIL - rcarlson@ieee.org

AbstractThis paper analyses design and performance This paper focuses firstly on machine operation and main
aspects of a brushless double fed cage induction generator as design aspects and secondly on steady state and dynamic
an economic and technical alternative to the classical wound analytical and numerical models that enable the efficient
rotor induction generator used in wind power generation. It
focuses firstly on the machine operational characteristics and
prediction of the DFTSCIG performance. Experimental
main design criteria and secondly on steady state and dynamic results are presented and discussed.
analytical and numerical models for efficient prediction of
machine behavior. The analytical dynamic model is obtained
by transforming the equations written in machine variables II. OPERATION
into equations written in an arbitrary reference frame. In the The DFTSCIG is an induction machine with a main three-
finite element approach the electrical machine is modeled in a phase winding with 2pp poles directly connected to the
2D domain, using the Maxwell equations to formulate the field electrical grid, and a three-phase auxiliary winding with 2pa
behavior and the FE method to discretize the domain of study.
A prototype was built and tested to validate the simulation
poles connected to the electrical grid through a vector-
models and to verify the design criteria. controlled converter (Fig. 1). The electrical connections
shown in Fig. 1 allow the control of the torque, the speed
Index Termsbrushless, double-fed, induction generator, and the power factor of the main winding by the inverter
rotor cage. connected to the auxiliary winding [2].

I. INTRODUCTION

T HE increasing interest on wind power generation directs


the study and development of several alternatives of
brushless electrical generators that operate at variable
speeds. One of these alternatives consists of a Doubly Fed
Twin stator Squirrel Cage three-phase Induction Generator
[DFTSCIG] for the reason that its performance when it is
double supplied presents certain features of practical
interest. By using an appropriate drive, it is possible to
control the induction machine to operate as a generator as
well as a motor at almost any speed. This is especially
convenient when a variable speed constant frequency is
required, as is the case of wind turbines. This machine has
been studied up to these days only in small power ranges,
not allowing conclusions about its ability to properly
operate at larger power range systems as required in a
modern Wind Power Station [1], [2]. Thus, to evaluate its
capabilities it is important to use pertinent analytical and
numerical models to aide in the machine design and to have
a better insight on its peculiar characteristics mainly in what
concerns the rotor cage [3], [4]. Fig. 1 Grid connection of the DFTSBIG
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 2

A special rotor cage configuration is designed where When the frequency induced in the auxiliary winding fa is
nested loops are included in the rotor circuit to minimize null, the machine is running at its natural synchronous speed
harmonic effects (Fig. 2). The advantage of this system over fsn, which according to Eq. (4) is:
other double-fed induction generator configurations is the
fp
fact that it is compact and brushless. f sn = (5)
p p + pa

Fig. 3 shows the variation of the frequency of the induced


currents in the cage and in the auxiliary winding as a
function of the mechanical speed of the machine. In this
figure, fsp represents the synchronous rotating speed of the
main stator winding and fsa the synchronous rotating speed
of the auxiliary stator winding, which are given by
fp fp
f sp = and f sa = .
pp pa

Fig. 2 Nested cage of the DFTSBIG

However, in wind generation it is not yet used because


this machine was not deeply studied at large sizes.
The operation of this machine depends greatly on the
rotor construction with a special cage. The number Npg of
rotor bars, which are responsible for creating the poles of
the cage, is given by:
N pg = p p + pa . (1)

The fundamental frequency of the air-gap induction wave Fig. 3 Cage and auxiliary winding induced current frequency.
generated by the main winding induces a current density in
the cage, that in turn creates a new induction wave in the III. DESIGN CRITERIA
air-gap with a frequency fg calculated by: Physically the DFTSCIG consists of two three-phase
f g = f p p p fm (2) windings sharing the same stator magnetic core. To avoid
the magnetic coupling between these windings, the number
where fp is the main winding frequency and fm is the shaft of poles of the main winding 2pp and of the auxiliary
mechanical frequency, both in hertz. winding 2pa must have a Maximum Common Divisor which
This air-gap induction wave due to the special cage divides the two numbers of poles giving as a result an odd
induces in the auxiliary winding a current density with a number for one of them and an even number for the other.
negative phase sequence with frequency fa given by: To avoid also the unbalanced electromagnetic pull, the
difference between the two numbers of pole pairs must obey
[ ( ) ]
f a = f p p p + pa f m (3) the relation [3], [4]:
where fa is the induced frequency in the auxiliary winding, p p pa > 1 . (6)
in hertz.
Therefore, the mechanical frequency of the shaft of the The main winding generates a set of induction harmonic
machine in hertz can be calculated by: waves in the air-gap of the machine with the following
numbers of pole pairs:
f p + fa
fm = . (4)
p p + pa gp
p = p p 1 + M p (7)
c p
Equation (4) shows that it is possible to control the speed
of the DFTSCIG by changing the frequency of the imposed where Mp is the number of the phase belts per pole pair of the
voltage on the auxiliary winding [2]. main winding; cp is the fractional part of the main winding
The frequency converter connected to the auxiliary and g p = 0;1;2;3;4;... assumes integer values from
winding, as shown in Fig. 1, not only imposes the
frequency, but also controls the amplitude and phase of the to +.
voltage imposed on the auxiliary winding allowing in this The auxiliary winding is able to generate air-gap
way a complete control of the DFTSCIG. harmonic induction waves with the following number of
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 3

pole pairs:

g
a = pa 1 + M a a (8)
ca

where Ma is the number of the phase belts per pole pair of the
auxiliary winding; ca is the fractional part of the auxiliary
winding and g a = 0;1;2;3;4;... assumes integer values
from to +.
The main and auxiliary windings are electromagnetically
coupled only through the rotor cage, and a good
performance is obtained when the Npg rotor bars produce Npg
poles, that couple the main and auxiliary windings
producing additive torques.
The currents induced in the cage by the induction waves
of p pole pairs generated by the main winding are capable
to generate induction waves in the air-gap with g pole pairs
given by:

g = p + g g N pg (9)
Fig. 4 Cage bars with loops.
where gg=0;1;2;3;4assume integer values from Consequently, we consider q2 loops per pole of the rotor
to +. cage knowing that the distribution factor which acts on the
Thus, to guarantee the electromagnetic coupling between harmonic g induced in the cage is:
the two stator windings and the cage, it is necessary to
impose that the number of pole pairs of the induction
sin g q2 N 2
harmonics generated by the cage g will be the same as the 1 2
f dg g = (14)
number of pole pairs of the harmonics of the auxiliary q2 N2
sin g
winding a, that is: 2

g =a . (10) where N2 is the angle between the rotor bars.


Concluding, to maximize the generator performance, the
Moreover, to produce addictive torques, it is necessary number of poles of the main winding must be greater then
that the currents induced in the cage by the main winding the auxiliary winding, mainly for the following reasons:
induction fundamental wave p=pp will be capable to a) The electromagnetic torque of an induction machine is
generate an induction wave in the air-gap, with a number of proportional to its number of poles and in the case of
pole pairs given by: the DFTSCIG the shaft speed is the same for both
armature windings, hence the mechanical power on
g = a = pa . (11) the shaft is proportional to the number of poles;
b) When the machine operates in steady-state as a motor
Taking these considerations in Eq. (9), it results: or as a generator around the natural synchronous
pa = p p + g g N pg . (12) speed, as shown in Fig. 3, the voltage and frequency
induced in the cage and consequently in the auxiliary
Using the condition given in Eq. (12), we obtain the winding are smaller, and therefore require less power
largest number of bars of the cage making gg= 1, therefore: from the converter;
c) The auxiliary winding with smaller number of poles
N pg = p p + pa . (13) presents smaller impedance. As will be seen later, this
impedance is viewed as in series with the main
Equation (13) gives us a rule of how to choose the winding. So, its reduction improves the performance
number of rotor cage bars, an important machine design of the machine in terms of power factor and
feature as previously depicted in the operation context efficiency.
(Eq. 1).
As shown in Eq. (13) the number of rotor bars is small. IV. ANALYTICAL APPROACH
Consequently, the harmonic content of the air-gap induction
wave is high. A. Dynamic Model
To minimize this problem each pole of the cage may be The analytical dynamic model is obtained by
constructed not only with one bar but with several loops, as transforming the equations written in machine variables into
shown in Fig 4. The objective of these loops is the reduction equations written in an arbitrary reference frame. Fig. 5
of this harmonic content by acting as a distributed rotor shows the reference frame used in this machine analysis.
winding. The stator circuit is considered fixed to the stationary
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 4

reference frame p1 and all machine variables (rotor and respectively. The indices 1 and 2 are related to the stator and
auxiliary winding parameters) are referred to the main stator rotor, respectively. The matrix parameters are Rp1, Ra1, the
winding. In Fig. 5 vectors f represent the machine voltages main and auxiliary windings resistances, respectively; Rp2,
or currents. The rotor circuit rotates with an angular speed Ra2, the rotor resistances reflected to the main windings;
of 2 electric rad/s; q and d axes rotate with an angular _qd0, the flux linkage; u_qd0, the voltages; i_qd0, the currents;
speed of qd0 electric rad/s and the angular displacement qd0, the reference axis speed; p2, the rotor speed and a1,
between the rotor circuit and the arbitrary q axis is 2. the auxiliary winding speed.
The system of differential equations in (16) is solved by
the fourth order Runge-Kutta method and, as a result, the
dynamic behavior of the machine is obtained.
B. Steady-State Model
The steady-state behavior is obtained using the machine
equivalent circuit considering the two stator windings
cascaded plus connected (Fig. 6). With this model it is
possible to analyze the machine operating at steady-state
both as a motor and as a generator for any load condition
with inductive and capacitive power factors.

Fig. 6. DFTSBIGs equivalent circuit.


Fig. 5 Arbitrary Reference Frame.
In Fig. 6, the sub-indices p and a are related to the main
As known, the stator circuits of the auxiliary winding are
and the auxiliary windings, respectively. The equivalent
physically fixed in the stator (stationary), but in order to &
consider the cascade effect in our dynamic model, we are circuit impedances are Z& , the stator impedance, Z _ c 2 ,
_ c1
forced to admit that their axes rotate with an angular speed &
the rotor impedance, Z _ cfe1 , the stator iron impedance,
a1 electric rad/s that represents the angular speed of the
Z& _ cfe 2 , the rotor iron impedance and Z& _ cm , the
stator circuits of the auxiliary winding and is given by:
magnetizing impedance. Z& cad is an external impedance
a1 = ( p p + pa ) m (15) connected in series with the auxiliary winding, representing
the static converter.
where m represents the mechanical angular speed of the Fig. 7 displays the curves of torque of the DFTSCIG, in
rotor; pp and pa are the number of poles of the main winding steady-state, obtained by the equivalent circuit model shown
and of the auxiliary winding, respectively. in Fig. 10. The blue curve represents the torque developed
By transforming the equation system to the arbitrary by the auxiliary winding (8 poles), the green curve the
reference frame, we obtain the following set of equations: torque of the main winding (12 poles), and the red curve
[ ]
u pqd 01 R p1[ ] [0] [0] [i pqd 01 ]
represents the total torque, that is, the sum of the main and
auxiliary windings torque, proving the desired addictive

[0] = [0]

[Rp 2 + Ra 2 ] [0] [ipqd 02 ] + behavior of the torque.

[ ]
aqd 01 [0]
u [0] [Ra 1 ] [iaqd
01 ]

[
qd 0 ] [0] [0] [ pqd 01 ]

[0]

[qd 0 p 2 ] [0] [ pqd 02 ] + (16)

[qd 0 a1 ] [aqd 01 ]
Torque (p.u.)

[0] [0]

[ ]
pqd 01
d
dt
[ ]
pqd 02
Total

[ ]
Main winding
Auxiliary winding
aqd 01

In the above matrix system, the sub-indices p and a are Speed (p.u.)
related to the main winding and the auxiliary winding, Fig. 7. Steady-state torque-speed curves.
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 5

In the point of 1 pu rotating speed, the three torque pass Q is a matrix associated with the flux linkage in thin
by zero indicating that the machine is in its natural conductors; R and L are diagonal matrices representing,
synchronous speed fsn as shown in Fig. 3. respectively, the resistances and end-windings inductances
At 1.667 pu rotating speed, again the three torque pass by of the thin conductors; X is the vector of the static converter
zero. In this point we have the synchronous rotation of the state variables; E is the vector of the independent electric
main stator winding fsp shown in Fig. 3. sources; G1 to G6 are matrices that depend on the static
In Fig. 7 it is also possible to observe that in the speed converter topology; Q is a matrix associated with flux
interval from 0 to 1pu the machine behaves as motor, the linkage in thick conductors; 1 is the identity matrix; Im is the
three torque being positives. From 1 pu to 1.667 pu speed vector of the thick conductors currents; R is a diagonal
the machine behaves first as generator (negative torque) matrix representing thick conductors resistances; Em is the
until the torque of the main stator winding becomes positive vector of the independent voltage sources; C1 and C2 are
again. Then the total torque also becomes positive and the matrices that depend of the thick conductors connection.
machine behaves again as a motor. For speeds above The Moving Band technique is chosen to consider the
1.667 pu the three torque are negative again, and the rotor movement during the field calculation. So, at each
machine works as a generator one more time. time step, the complete magneto-dynamic system is updated
Then, the DFTSCIG can work perfectly as motor or as and solved and the resulting magnetic vector potentials are
generator, when controlled by the static converter, as shown used to calculate the instantaneous electromagnetic torque
in Fig. 1, in a speed range of 30% around the 1 pu natural via the Maxwell Stress Tensor. The new angular speed and
synchronous rotation. rotor displacement are evaluated using the mechanical
oscillation equation.
V. FINITE ELEMENT APPROACH
In the finite element (FE) approach the electrical machine
is modeled in a 2D domain, using the Maxwell equations to
formulate the field behavior and the FE method to discretize
the domain (Fig. 8). The movement is taken into account by
means of the Moving Band technique, the Maxwell stress
tensor and the mechanical oscillation equation.
The formulation uses the magnetic vector potential A as
unknown, the Galerkin method to obtain the set of equations
to be solved numerically, the Euler recurrence method to
discretize the temporal derivatives and the Newton-Raphson
method to consider the non-linear characteristic of magnetic
materials. As thin conductors (stator windings) and thick Fig. 8. Finite element domain (6290 elements and 3555 nodes).
conductors (rotor bars) must be taken into account, the
equation of a magneto-dynamic system is obtained [5]. VI. COMPARISON RESULTS BETWEEN ANALYTICAL AND
Moreover, thin conductors that can be fed by external NUMERICAL METHODS
electric circuits, as static converters, form the machine The finite element analysis provides a good insight into
windings and multiple and different combinations of thick machine phenomena as iron core saturation, eddy currents
conductors connections can also be taken into account. The distribution, losses and so on. But, depending on the
final FE magneto-dynamic system is: problem characteristics it may take a few hours for one
simulation. On the other hand, analytical methods are faster
dA(t )
M ( )A(t ) + N PI f (t ) P' U m (t ) = D due to their reduced number of variables and, if its
dt
parameters are well determined, they provides good results
dA(t ) d
Q + R + L - G 6 I f (t ) - G 4 X(t ) = G 5 E(t ) becoming a natural tool of machine project and analysis. By
dt dt working with these two methods together we can exploit its
d (17)
G 3 I f (t ) + G 1 - X(t ) = G 2 E(t ) complementary potentials, as an example, using the FE
dt approach to determine machines inductances to be included
d in the analytical model.
Q' A(t ) 1U m (t ) + R' I m (t ) = 0
dt An example of the information the FE method can
C1U m (t ) + C 2 I m (t ) =Em provide for a visual evaluation is shown in Fig. 9 and 10,
that is, respectively, the magnetic induction and induced
where If is the vector of the electric currents in the thin currents in the rotor bars distribution obtained in the FE
conductors; Um is the matrix of the electric voltages in the simulation of the DFTSBIG acceleration process at t = 0.6s.
thick conductors; M is a matrix related to the iron core We can evaluate from Fig. 9 the level of saturation of the
permeability; N is a matrix related to the conductors electric magnetic core and to identify points of excessive saturation,
conductivity; P is a matrix that relates the current in the thin for example.
conductors elements and its nodes; P is a matrix that relates Fig. 11 to 14 shows some results obtained in the
the voltage in the thick conductors elements and its nodes; simulation of the DFTSBIG acceleration process using the
D is a vector related to permanent magnet contributions; two previously presented methods, the analytical and the
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 6

finite elements one. The main stator winding is considered


Y-connected and fed directly by three-phase balanced
voltage sources 760 VRMS 60 Hz. The auxiliary winding is
considered Y-connected but, for the results presented in this
paper, with its external terminals short-circuited. To later

Torque (p.u.)
compare the simulation results whit experimental ones we
chose to consider the case where an external constant motor
torque equal to 2000 Nm is imposed to the shaft by an
electrodynamometer during the entire simulation interval,
allowing the machine to surpass the synchronous speed.

Time (s)
(a)
4

Torque (p.u.)
0

-1

-2

-3

Fig. 9. Field distribution (T) at t = 0.6 s: FE simulation. -4


0 0,2 0,4 0,6 0,8 1 1,2
Time (s)

(b)
Fig. 11. Electromagnetic torque over the entire speed range: (a) analytical
model, (b) FEM.
Current (p.u.)

Phase a
Phase b
Phase c

Time (s)

(a)

8
Fig. 10. Rotor bars induced current distribution (A/m2) at t = 0.6 s.
6
The analytical and FEM simulation results shown in
4
Fig. 11 for the electromagnetic torque compare very well in
amplitude and time. We can identify in these dynamic 2
Current (p.u.)

results the instants the rotor passes through the natural 0


synchronous speed (t 0.35s) and the main winding
-2
synchronous speed (t 0.7s). These characteristic points of
the DFTSCIG operation were already identified in Fig. 7 for -4
Phase a
the steady-state regime. -6 Phase b
Phase c
In Fig. 12 we can observe the behavior of the main -8
winding currents during the acceleration of the machine 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Time (s)
rotor. It is important to notice the significant reduction of
these currents at the points corresponding to that two (b)
characteristic speeds. Fig. 12. Main Winding Current: (a) analytical model, (b) FEM.
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 7

Fig. 13 shows the induced currents in the short-circuited method, the rotor bars are directly considered in the model,
three-phase auxiliary winding. In it we can observe the being its real dimensions and electric connections modeled.
frequency variation illustrated in Fig. 3 (red line). The Therefore, it gives, as a result, the real current at each of the
frequency of the induced currents in the auxiliary winding rotor bars, the currents waveform in time and its density
decreases with the rotor speed, being zero when the machine distribution (Fig. 10).
reaches the synchronous natural speed (t 0.35s). After that
moment, it increases linearly with the speed.
Also, two points of minimum current are present, the first
one at the synchronous speed (t 0.35s) and the second one
at the main winding synchronous speed (t 0.7s), as
observed in the main winding currents.

Current (p.u.)
Phase a
Current (p.u.)

Phase b
Phase c

Time (s)
(a)

4000
Phase a
Phase b
Phase c 3000

2000
Time (s)
1000
(a)
Current (A)

4 0

3 -1000

2 -2000

-3000 External loop


1 Intermediate loop
Current (p.u.)

Internal loop
0 -4000
0 0,2 0,4 0,6 0,8 1
Time (s)
-1

-2 (b)
Phase a Fig. 14 Rotor cage currents: (a) analytical, (b) FE.
-3 Phase b
Phase c
-4
0 0,2 0,4 0,6 0,8 1
VII. 7 TEST RESULTS
Time (s)
The prototype here analyzed is a 100kW 440V/760V
(b) 60Hz, for 12 poles the main winding and 8 poles for the
Fig. 13 Auxiliary winding current: (a) analytical model, (b) FEM. auxiliary winding.
The measured data we have chosen to present here is the
Fig. 14 shows the induced currents in the rotor bars, it is machine torque x speed curve (Fig. 15).
analogous to Fig. 13, i. e., we can identify two instants The test was performed by applying a negative torque
when the induced currents are minimum, one at the (motor torque) to the DFTSCIG shaft, this imposed torque
synchronous natural speed (t 0.35s) and another at the being enough to drag it up to approximately two times and a
main winding synchronous speed (t 0.7s). Also here we half the DFTSCIG natural synchronous speed. It was used a
can observe the frequency variation depicted in Fig. 3 (bleu Rotary Torque Sensor (RTS) inserted between the
line). Differently from the auxiliary winding, in the rotor DFTSCIG shaft and the dynamometer and its signals have
bars the frequency of the induced currents passes by zero at been recorded directly by an analog plotter.
the main winding synchronous speed. Comparing the experimental results with the simulated
A main difference between the analytical and the FE ones (Fig. 15), we observe that the peaks or motor and
approach is evidenced in Fig. 14. In the analytical method, generator measured torque are smaller than the calculated
an equivalent three-phase balanced winding obeying a ones. This difference is partly attributed to the fact that the
sinusoidal distribution on the rotor surface represents the dynamometer accelerates very rapidly and the plotter pen is
rotor bars. So, as a result, this method gives us three-phase unable to follow this acceleration. Another reason is the fact
currents equivalent to the real rotor currents. In the FE that losses were not considered in the simulations.
NORDIC WIND POWER CONFERENCE, 1-2 MARCH, 2004, CHALMERS UNIVERSITY OF TECHNOLOGY 8

4 [2] R. Li, R. Spe, A. K. Wallace and G. C. Alexander. (1994,


FEM simulation July/August). Synchronous Drive Performance of Brushless Doubly-
3 Measured Fed Motors. IEEE Trans. on Ind. Appl., v. 30, n. 4, pp. 963-970.
Analytical simulation [3] S. Williamson, A.C. Ferreira and A.K. Wallace. (1997, March).
2 Generalized Theory of Brushless Doubly-Fed Machine. Part 1:
Analysis. IEE Proc. Elect. Power Appl., Vol 144, No 2, pp. 111-122.
1 [4] S. Williamson, A.C. Ferreira, A.K. Wallace. (1997, March).
Torque (p.u.)

Generalised Theory of Brushless Doubly-Fed Machine . Part 2:


0
Model verification and Performance. IEE Proc. Elect. Power Appl.,
v. 144, n. 2, pp. 123-129.
-1
[5] A. M. Oliveira, P. Kuo-Peng, N. Sadowski, F. Rncos, R. Carlson, P.
-2
Dular. (2003, July). Finite Element Analysis of a Double Winding
Induction Motor with a Special Rotor Bars Topology. Proc. of
-3 COMPUMAG2003, Saratoga Springs, NY USA.

-4
0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 2,2
Speed (p.u.)

Fig. 15. Dynamic Torque - Speed characteristics.

VIII. 8 CONCLUSION
A discussion about operational characteristics and design
criteria of a DFTSCIG machine was presented as well as its
analytical and finite elements models. These simulation
models analytical and FEM have been validated by
experimental results obtained from the specially built
prototype. The theoretical analysis has presented good
results when compared with the experimental ones and also
the performance analysis, in what concerns the torque was
quite satisfactory confirming the initial statements that the
DFTSCIG may be an advantageous alternative to generate
electric energy from wind power.
The main difficulty encountered during the design phase
is related to the high harmonic content generated in the air-
gap flux due to the small number of rotor cage bars. These
harmonics create high leakage reactances. The loops
introduced in the rotor cage help minimize this harmonic
content, but showed not enough in our prototype. Therefore,
additional studies are necessary in order to improve the
rotor cage design.
Another important aspect is that windmill speed is quite
low and then the DFTSCIG should have a large number of
poles in the main winding. However this increases the
leakage reactances of the generator affecting its power
factor and efficiency. The power factor can be easily
corrected through the auxiliary winding by the static
converter, but the efficiency can only be improved by
changing the machine design.
The results presented in this paper, although very
important in the machine comprehension and development
of design tools, are still preliminary. Continuous efforts will
be engaged to improve the DFTSCIG performance.

ACKNOWLEDGMENT
Authors wish to thank WEG S.A. for the construction of
the prototype and for the use of the testing facilities.

REFERENCES
[1] Y. Liao. (1996, August) Design of a Brushless Doubly-Fed Induction
Motor for Adjustable Speed Drive Applications. Conference Record
of the Thirty-First IEEE/IAS Annual Meeting, Volume 2, pp. 850
855.

You might also like