Torque and Reactive Power Control of a Doubly-Fed Induction Machine by

Position Sensorless Scheme

Wei Cheng Longya Xu
Student Member, IEEE Senior Member, IEEE

The Ohio State University

Department of Electrical Engineering
2015 Neil Avenue
Columbus, Ohio 43210

L. Introduction scheme is proposed. Compared to the other position

sensorless schemes for doubly fed machine, &e proposed
Doubly fed wound rotor induction machine with control method uses only the rotor voltages and currents
vector control is very attractive to the high performance, as the feed back signals, which substantially reduces the
variable speed drive and generating applications[ 1-31. costs and enhances relibility of the position sensorless
In variable speed drive application, the so-called slip control scheme of the doubly fed AC machine[8,9].
power recovering scheme is a common practice where the Independent control of torque and reactive power is also
power due to the rotor slip belowlabove synchronous possible, and the doubly fed system can achieve unity or
speed is recovered to/supplied from the power source, leading power factor, in addition to variable speed
resulting a high efficiency variable speed system. In operation.
variable speed generating application, the doubly fed
induction machine is most suited for variable-speed
constant-frequency generating systems in which the
speed of the prime mover is allowed to vary within a A . Configuration of Doubly Fed Induction Machine
certain range (sub- and super synchronous speed), but the System
output electrical power is always maintained at a
constant frequency. In such a case, the doubly fed
The configuration of a variable speed doubly
induction machine is actually in the dual mode of
fed induction machine system is shown in Fig. 1.
variable speed drive system.

The fundamental feature of the doubly fed

induction machine system, including the drive and
generating systems, is that the power processed by the
power converter is only a small fraction of the total
system power. Therefore, for a very large rating system
it is possible to use a high-frequency switching PWM
BiWional
Power-Flow
Convcttcr
-
-
A

converter to achieve high performance, such as fast

dynamic response, low harmonic distortion, high
efficiency, etc., without cost penalty. However, a high
resolution rotor position sensor is generally needed for Fig. 1 Configuration of Doubly fed Induction Machine
the proper operation of the doubly fed induction machine system
systems.
The major component of the system is a wound rotor
Research has been very active in recent years to induction machine which needs to be excited at both the
eliminate position sensors in field orientation stator and rotor terminals. It is common practice that the
controlled cage rotor induction machine and the results stator winding is connected to the AC power supply
are promising[4-61. In a cage rotor induction machine, directly, while the rotor winding is connected to the
the rotor current is passively induced and the rotor field power supply through a variable frequency power
is subsequently defined by the induced current. Yet, for converter. To obtain sub- and sup-synchronous speed
the doubly excited induction machine the rotor field is operation, the power converter of the rotor winding must
explicitly defined by the external excitation source be able to handle slip power in both directions [2] using
which requires synchronization with respect to the stator a PWM voltage source inverter. Among the three power
field. Therefore, without using a shaft position sensor, flow ports, i.e. the stator terminals. the rotor terminals,
vector or field orientation control of a doubly fed and the rotor shaft, the rotor terminals act as the energy
induction machine is more complicated than that of a regulating port, balancing the power flow of the entire
singly-excited cage rotor machine. system. For example, when the shaft speed is below
synchronous speed, a small portion of power is
subtracted from the system through the rotor terminals
In this paper, a novel control strategy to so that the power input to the system through the stator
realize torque and reactive power control of a doubly winding is balanced by the power output from both
excited induction machine using position sensorless

0-7803-1993-1194 $4.00 0 1994 IEEE

496
mechanical shaft and the rotor circuit. Similarly, when C . Reactive Power Control Through Rotor Currents
the shaft speed is above synchronous speed, a small
portion of power is added to the system through the rotor The reactive power of the doubly fed induction
circuit. While the example is given in terms of motor machine system can be controlled through the rotor
operation, it is straightforward to prove that for current vector to realize unity or. leading power factor
generator operation, the rotor circuit functions in a operation. Assuming that the required reactive power to
similar manner. the stator is Q,which can be expressed in the d-q-0
reference frame as
B. Basic Equations and Operation Principles

The operation principle of the variable speed The reactive power Qs can be referred to the rotor side and
doubly fed induction machine can be conveniently expressed in terms of rotor quantities as
analyzed by the classical rotating field theory with the
well-known d-q-0 transformation. By inspection, the
stator and rotor equations are written as follows in the
matrix form.

if the reactive power consumed by the stator leakage

inductance is very small and neglected.
After the d-q-0 transformation, the equations, in
the synchronous reference frame, are In a doubly fed induction machine the airgap
flux is determined largely by the stator voltage which
(3) has a constant magnitude and frequency. Hence, it is
reasonable to further assume that the airgap flux is
constant. According to Eqs. (11) and (12), if the rotor
current idr is controlled, the reactive power or the power
where we is the speed of the synchronous reference frame, factor can be controlled. Again, for reactive power
and w j s the rotor speed. control, the rotor speed and the rotor position must be
known for the rotor current vector projection. (see
Appendix)
The torque equation becomes

Lu. Estimation of T o r a u e Anele and Rotoy

wherein &,,,= &,& -AbLts SDeed
and $dm =&dr --iqdr Ltr
As discussed above, in order to have precise
control over the torque and reactive power (or power
In Eq. (5),6 is the spatial angle between the airgap flux
factor) of the doubly fed induction machine, we need to
and rotor current vectors (named torque angle in this
paper) seen in the synchronous reference frame. With maintain a controllable torque angle, 6, or a controllable
the alignment of the airgap flux to the d-axis of the rotor current vector, liqdri. For the doubly fed induction
reference frame, we will have machine system using a high resolution shaft encoder,
A+,,=&+,, and hm=O. (7) the rotor position is constantly sensed to ensure that the
required slip frequency, CO,, satisfies the equation w, = we-
The torque equation is reduced to
0,. The orientation of the rotor current vector is realized
Te --?E
2 2 Ihdml Iiqrl (8)
by the final tuning of the slip frequency so that the
In order for rotor current vector in command currents i and idr are projected orthogonal
synchronization with the stator current vector, the 4,.
and parallel to the airgap tlux respectively. In the
frequency of the rotor current, a,,must satisfy the slip following sections, we will discuss how i,, and idr are
frequency constraint projected to the right position without the use of a costly
w, = oe-w, (9) position sensor.
Furthermore, for field orientation control, 6 must be
maintained at go', or equivalently, the current vector has A. Estimation of Torque Angle
to maintain orthogonal to the airgap flux. For a general
vector control, the rotor current vector is controlled with It is extremely important to realize that the
a predetermined angle with respect to the airgap flux. It electromagnetic torque can be viewed as the result of the
is because of Eq. (9) and the field orientation or vector interaction between rotor current and airgap flux,
control that the rotor speed and position must be regardless of the seleciion of the reference frame. That
measured (or estimated) instantaneously so that the rotor is, Equations
current vector can be correctly positioned.
497
 A

cos$=
^h m i
2r + Lxm i x r (20)
km i r
Taking the inverse of sin6 and cos6 terms, the torque
angle 6 is uniquely determined. Similar to the
computation of sine term, the needed inputs for Eq. (20)
are rotor side variables which are readily available and no
are equivalent. In Eqs. (13-14), we have used subscripts additional sensing is needed.
of d-q, and x-y to differentiate the selection of the
reference frames tied to the synchronous and rotor circuit B . Control of Torque Angle
frame respectively. Since Eqs. (13-14) are equivalent,
observation of the magnitudes of flux and current, as well It is interesting to note that by using the rotor
as the electromagnetic torque, are independent of the voltages and currents, the torque angle 6 is estimated
reference frame selection. One can estimate the toraue without any transformation. Nevertheless, the estimated
angle in one reference frame. and then use the estimated A
results in other reference frames. Furthermore, the airgap angle 6 can be used in any other reference frames. Using
flux can be estimated from either the stator, rotor or the estimated angle of ^s
as the feedback variable, it is
combined circuit. Therefore, one can always choose the possible to force the actual 6 to follow the desired value.
most convenient circuit, that of the stator or rotor, in the
The scheme of the torque angle estimation and control is
most convenient reference frame, the rotor or
summarized as shown in Fig. 2. In the control block of
synchronous frame, to compute the magnitudes of the
Fig. 2. a voltage controlled oscillator (VCO) with high
flux and current, and the torque angle 6. gain is used to generate a desired slip frequency so as to
converge the actual torque angle to the desired
In the position sensorless control strategy
proposed in this paper, the rotor currents and voltages
are used as the basic input variables to compute the
torque angle, because the rotor variables are already
available for the purpose of current regulation.
Assuming that the reference frame is tied to the rotor
axis, the torque angle and the flux magnitude together
with the current magnitude can be estimated by the
measured rotor voltages and currents according to the
following equations: A

2 2
ir = - \ l i x m + i,
Fig. 2 Estimation and Control of Torque Angle
Thus, according to Eq. 14, the sine of the torque angle is:
one. It is in this way that the torque and reactive power
control of the doubly fed induction machine is properly
achieved. The torque angle control scheme can be best
A A
In Eqs. (15) through (19), 6, &,A.,
A
and I\hm
are used to illustrated by the phasor diagram shown in Fig. 3. Note
that as soon as an error in the torque angle is detected by
indicate that they are the estimated values from the
estimator. Note, however, that sin% alone is not the estimator, for example s<S* as shown in the phasor
A diagram, the VCO based controller will immediately slow
sufficient to uniquely determine the value of 6, since the down the slip frequency so that e, is reduced. As a result,
inverse of a sine function is dual-valued. Interpreting the torque angle increases. To make the current
this observation to the machine physics, we can say that regulators in the synchronous reference frame function
the information contained in the electromagnetic torque properly, i. e. to force iqr to increase and idr to decrease,
or active power can not uniquely determine the torque it is a wise choice to transform the measured rotor current
angle. It is evident that additional information is to the synchronous reference frame by
necessary. In effect, by computing a quantity related to
the instantaneous reactive power, cos 6 can be found by idr = ixrcos8, + iyr sine,
the following equation:
iqr = -ixrsine, + iyr cos8,
and then use them as the feedback signals.

498
 feed forward signal as shown in Fig. 2 may be helpful to
compensate t h e temporarily lose of the synchronization.

The principles and control scheme of the

position sensorless controlled doubly fed induction
machine described in Fig. 4 is implemented by a
computer program and then the performance is
investigated. The specifications of the doubly fed
induction machine used in the computer simulation are
shown in Table 1.

"7
Fig. 3 Phasor Diagram of Torque Angle Control
For the doubly fed induction machine under
investigation, the stator is fed from a three phase AC
upp ply at 60 Hz from the utility lines. The rotor current
is provided by a CRPWM power converter. First, the
C . Estimation and Control of Rolor Speed
estimation scheme of the torque angle is checked and the
results are shown in Figs. 5(a) through 5 ( f ) . Fig 5(a)
Integrating the torque angle estimation and shows the torque angle command, 6*; (b) the actual rotor
control schemes into the system, the rotor speed can be
estimated conveniently. Recall that we have forced the speed; (c) the estimated sin& (d) the estimated cos6 ; (e)
torque angle to the command value continuously. This the estimated torque angle; and (f) the estimated rotor
simply implies that the rotor current vector is speed. Comparing Figs. 5(a) with 5 ( e ) , it is evident that
synchronized with respect to the stator field, satisfying the torque angle is accurately estimated. The rotor is
= CO,-CO, Consequently,
the slip frequency constraint CO, accelerated steadily, covering both the sub- and sup-
synchronous speed zones. Similar results are shown in
the rotor speed can be estimated by using the slip
the same figures when the machine is decelerated. As
= w,-w,, where W, and CO,
frequency constraint equation CO, shown in Figs. 5(b) and 5(Q, the estimated motor speed
are known quantities. is very close to the actual one.

6y ; : !
Adding the rotor speed estimation and control to
the torque angle control loop, the block diagram of the
overall system is obtained as shown in Fig. 4
1W

-50

.1W

(4
1.5,

Fig. 4 Block Diagram of the Proposed Controller

The portion enclosed in the box can be implemented by

software and executed in a digital signal processor
conveniently.

Note that the rotor current vector is not always

synchronized with respect to the stator field or airgap
field. The momentary lose of synchronization between
the rotor current and airgap field is particularly true, and
is actually necessary for the changes of the torque and/or
reactive power command. In such a case, an error will
occur in rotor speed estimation. To correct this error, a (6)

499
 power factor, which are clearly achieved by the position
sensorless control method. In addition, iqr is not affected
indicating a decoupled reactive power control from the
- -50
torque control.

02 04 06 08 1 12 14 16
200

-200

4 ob1 o h ob3 oh obs obs o d

I
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
(0nw-4

Fig. 5 Simulation Results of Torque Angle and Speed

Estimation

Fig. 6 shows the result of the closed loop speed

control. As discussed previously, in the power angle
controller, a slip frequency command is generated to Fig. 7 Decoupled Control of Reactive Power
synchronize the rotor current vector with respect to the
stator field. This implies that the correctly generated -ntal Resm
slip frequency can be used readily for rotor speed
estimation. That is, or=we -aswhere we and w, are
To substantiate the theoretical analysis and
known variables. With the virtually overlapped two computer simulation, the control algorithm has been
traces, Fig. 6(a) shows the command and the actual rotor realized by a general purpose Motorola DSP56000
speed; Fig. 6(b) shows the estimated rotor speed which is system and tested on a doubly fed variable speed drive,
used as the feedback variable; and Fig. 6(c) shows the

1 "
with the machine specifications the same as those of the
rotor phase current. simulated one, implemented at The Ohio State
University. Only four A/D channels are used on the rotor
circuit for torque angle estimation and control.

Torque angle estimation is tested f i s t and the

results are shown in Fig. 8. In the test, the machine is
hmnd and actual r m r .p.d
0 under vector control using a rotor position encoder. It
0 02 04 00 OB 1 12 14 16 18 can be observed that when the accelerating torque is
(a)
applied the estimated torque angle is about -90' and when
the decelerating torque is applied the torque angle is
about go', both are very well matched to the actual toque
: l -0 5
?
angles. The phase current is shown as the third trace,
clearly indicating the correspondence of the magnitude
versus that of the phase current. It should be pointed out,
E.turuM I o b l s p u d
0 however, that when the torque is very small. during the
0 02 04 OB OB 1 12 I4 16 18 interval of constant speed and low phase current, the
(b)
torque angle observation curve has many ripples. This is
because the input signals to the torque angle estimator
are too weak and the estimation is sensitive to the input
errors.

Fig. 9 shows the experimental results using the

estimated torque angle and speed as the feedback signals
for torque and speed regulation. In this case, the doubly
fed machine is truly under the position sensorless
control. As indicated by the testing results, the torque
Fig. 6 Simulation Results of Speed Control angle and speed estimation are applied to the torque
control and speed regulation successfully. It is
The third simulation investigates the reactive interesting to compare the phase current of this figure to
power control of the doubly fed induction machine by the that of the previous one. Note that the phase current
position sensorless control scheme, and the results are magnitude at light load with position sensorless control
shown in Fig. 7. Two levels of & are applied to change the is much larger than that with position sensor. This is

500
because, as describe above, the torque angle observation
at light load is easily subject to the errors for the low In experimental testing, the torque angle
signal/noise ratio. estimation algorithm is tested over a wide speed range
covering both sub- and super synchronous speed and the
results are shown in Fig. 11. It can.be seen that when the
rotor crossing synchronous speed, a big dip appears on
Or (600 rpm/div)
the torque angle estimation curve, showing a difficulty of
estimating torque angle at very low frequency. This is
actually a dual problem met in the singly fed induction
machine when the rotor speed is very low, or
6 (18o'/div) equivalently, the excitation frequency approaches zero.
While the torque angle estimation at low frequency itself
constitutes a major research topic and is out of the scope
in, (20A/div) of this paper, more advanced control algorithm does
show the light that the problem could be overcome as
time (OSddiv) described in [lo].

or(14001pm/div)
Fig. 8 Testing Results of Torque Angle Estimation

1 . 8 r I , . I / , ,
6 (ISO'/div)

or(1000rpm/div)
A
or(1000rpmldiv) l
9
'

in, (2ONdiv)
6 (1807div)
time (OSs/div)

in, (20Ndiv)
Fig. 11 Testing Results of Torque Angle Estimation
time (2s/div) When Rotor Crossing Synchronous Speed

VT. C o n c l u s i o u
Fig. 9 Testing Results of Speed Control without
Position sensor A position sensorless scheme is proposed for
the doubly fed induction machine system using only the
Fig. 10 is the experimental results of decoupled rotor side variables as the measured signals. The
reactive power control using position sensorless principles associated with this particular position
scheme. The power factor of the stator circuit is sensorless control are presented and discussed. The
evidently boosted to almost unity when a step theoretical results are verified by computer simulation
magnetizing current idr is applied while the torque is not and lab experiments. The following conclusions are
affected. reached:

I i I I I t ' .
I !
!
: !
!
::
:
1). The torque angle 6 can be estimated satisfactorily by
the method suggested except when the rotor frequency is
very low. The angle observed contains information of
the rotor mechanical speed and angle, together with
vas (6Ov/div)
those of the rotor excitation;
10 in, (lOA/div)
2 ) . For torque angle estimation in position sensorless
scheme, information related to both torque (active
power) and reactive power must be used to uniquely
idr (10A/div) determine the torque angle. The estimation is
independent to the selection of reference frames, and to
'9' the rotor or stator circuits.
time (lOms/div)
3). Torque angle control technique can be used to replace
rotor position sensor for high performance vector
Fig. 10 Reactive Power Control control of doubly-excited AC machines. Independent

501
control of the reactive power can also be achieved by the IEEE/PES Winter Meeting, Feb., 1994, New
position sensorless control scheme. York, to appear on IEEE Transactions on Energy
Co nv ersio n.
4). To use the position sensorless method for rotor speed
estimation and control, the estimation is subject to an Table 1: Specifications of the Induction Machine
error introduced by the very low operating frequency or
whenever the load torque is very light. Voltage 220 volts Current 14.8 amps
Rated speed 1725 rpm R s = 1.11 ohms
It should be pointed out that although the Rr = 0.3 Ohms Xm = 18 ohms
position sensorless method presented in this paper is xes = xir = 0.77 ohms
applied to the doubly fed induction machine, the
mechanism described is applicable to the other doubly Appendix:
excited machines including conventional synchronous
machine, P M machine, synchronous reluctance machine,
and the newly developed doubly excited brushless Reactive Power Flow of the Doubly Fed Induction
reluctance machine. In the future work, more advanced Machine
algorithm will be included in the torque angle estimation Reactive power flow in the doubly fed induction
to account for parameter variation at low frequency. machine is derived in this appendix. Note that the basic
Further, intelligent control approach, such as fuzzy logic equations of a doubly fed induction machine are
control, can be added to the slip frequency generator to expressed in the synchronous reference frame as
enhance robustness and simplify the controller. dhd
hs
v k = rs ids - CO, +
dt
References

S . R. Doradla, S . Shakravorty, and K. E. Hole, "A

New Slip Recovery Scheme with Improved Supply A6 = Lcs ik + A b (A-3)
Power Factor," IEEE Trans. on Power Electronics,
vol. 2, No. 2, April, 1988 kqs = 4 s 6s
+ hgm (A-4)
L. Xu and Yifan Tang, "A Novel Variable Speed If the d-axis of the reference frame is aligned with the
Constant Frequency Wind-Power Generating airgap flux, the next two equations follow.
System Using Doubly-Excited Brushless Aq, = 0 = L,,, (iqs+ i,,) (A-5)
Reluctance Machine", IEEE/IAS Annual Meeting,
Oct. Houston, 1992. = hl1= Lm (ids + idr) (A4
L. Refoufi, P. Pillay and M. R. Harris, "A
Chopper-Controlled Slip Energy Recovery The reactive power flow into the stator winding is
Induction Motor Drive." I E E E P E S Summer 3
Meeting, Seattle, 1992 =2
Qs (vqs ids - vds i q d (A-7)
Hirokazu Tajima and Yoichi Hori, "Speed Substituting Equations (A-5) and (A-6) into Eqs. (A-1)
Sensorless Field Orientation Control of the through (A-4). and making use of the results in Eq. (A-7),
Induction Machine," IEEE/IAS Transactions, Vol. we have
29, No. 1, Jan/Feb, pp 175-180 3 3
T. Ohtani, N. Takata and K. Tanata, "Indirect =?W, h,ids + ? a e L t S (ih2 + 42) (A-8)
Vector Control Without Shaft Encoder," IEEE/iAS
Annual Meeting Record, 1989, pp 500-507 Eq. (A-8) can be expressed purely in terms of rotor side
F. Zheng amd T. Fukao, "Robust Speed variables
Identification for Speed Sensorless Vector
Control of Induction Motors," IEEE/IAS Annual
Meeting, 1993, pp 419-426 %=?We -(Ls h,2 - h,idr + Les (id: + 6
: - 2 h,&,) ) (A-9)
Ll?
L. Xu and Y. Tang, "Stator Field Oriented Control Furthermore, the reactive power consumed by the leakage
of Doubly-Excited Induction Machine in Wind inductance is very small, and can be neglected.
Power Generating System", 35th Mid West Therefore,
Symposium on Circuit and System, Washington
D. C., Aug. 1992. (A-IO)
Y. Liao and C. Sun, "A Low Cost, Robust
Position Sensorless Control Scheme for Doubly- If the magnitude of the airgap flux h, is stable, which is
Fed Reluctance Motor Drives," IEEE/IAS Annual true for the doubly fed induction machine in steady state,
Meeting 1993, pp 437-444 then the desired amount of reactive power flow into the
A. B. Plunkett- and F. G. Turnbull "Load- stator can be controlled by controlling as indicated by
Commutated Inverter/Synchronous Motor Drive the Equation.
without a Shaft Position Sensor", IEEE/IAS
Transaction, Vol. IA-15 No. 1, Jan/Feb. 1979, pp
63-71
Y. Tang and L. Xu, "Fuzzy Logic Application for
Intelligent Control of a Variable Speed Drive",

502