You are on page 1of 9

122 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 36, NO.

1, JANUARY/FEBRUARY 2000

A Modified Direct Torque Control for Induction


Motor Sensorless Drive
Cristian Lascu, Ion Boldea, Fellow, IEEE, and Frede Blaabjerg, Senior Member, IEEE

Abstract—Direct torque control (DTC) is known to produce Several solutions with modified DTC are presented in the lit-
quick and robust response in ac drives. However, during steady erature. Due to its simple structure, DTC can be easily integrated
state, notable torque, flux, and current pulsations occur. They are with an artificial intelligence control strategy. The fuzzy logic
reflected in speed estimation, speed response, and also in increased
acoustical noise. This paper introduces a new direct torque and solution for flux and torque control is shown in [3].
flux control based on space-vector modulation (DTC-SVM) for A different approach is to combine the voltage vector selec-
induction motor sensorless drives. It is able to reduce the acous- tion with an adequate pulsewidth modulation (PWM) strategy in
tical noise, the torque, flux, current, and speed pulsations during order to obtain a smooth operation. The closed-loop stator flux
steady state. The DTC transient merits are preserved, while predictive control, open-loop torque control using space-vector
better quality steady-state performance is produced in sensorless
implementation for a wide speed range. The flux and torque modulation (SVM) implementation is shown in [4]. The SVM
estimator is presented and an improved voltage–current model is a performant open-loop vector modulation strategy [5].
speed observer is introduced. The proposed control topologies, This paper introduces a new direct torque and flux control
simulations, implementation data, and test results with DTC based on SVM (DTC-SVM) for IM sensorless drives. It imple-
and DTC-SVM are given and discussed. It is concluded that the ments closed-loop digital control for both flux and torque in a
proposed control topology produces better results for steady-state
operation than the classical DTC. similar manner as DTC, but the voltage is produced by an SVM
unit. This way, the DTC transient performance and robustness
Index Terms—Direct torque control, estimators, sensorless. are preserved and the steady-state torque ripple is reduced. Ad-
ditionally, the switching frequency is constant and totally con-
I. INTRODUCTION trollable.
Another important issue for a sensorless drive is the flux,

R ESEARCH interest in induction motor (IM) sensorless


drives has grown significantly over the past few years due
to some of their advantages, such as mechanical robustness,
torque, and speed estimation. Both open-loop and closed-loop
speed and position estimators are widely analyzed in the litera-
ture. The most promising speed observers seem to be the adap-
simple construction, and maintenance. Present efforts are de-
tive ones, either with linear or nonlinear structures [6], [7]. How-
voted to improve the sensorless operation, especially for low
ever, the low-speed range estimation still remains an unsolved
speed and to develop robust control strategies.
problem. This is not the case for flux and torque observers which
Since its introduction in 1985, the direct torque control
are able to generate accurate estimation for the whole speed
(DTC) [1] (or direct self control (DSC) [2]) principle was
range [8]–[10]. An improved voltage–current model speed ob-
widely used for IM drives with fast dynamics. Despite its sim-
server based on a model reference adaptive controller (MRAC)
plicity, DTC is able to produce very fast torque and flux control
structure is proposed herewith.
and, if the torque and flux are correctly estimated, is robust
The paper presents the complete sensorless solution based on
with respect to motor parameters and perturbations. However,
a DTC-SVM strategy. The proposed control topologies, digital
during steady-state operation, notable torque, flux, and current
simulations, implementation data, and test results with DTC and
pulsations occur. They are reflected in speed estimation and in
DTC-SVM are given and discussed.
increased acoustical noise.

Paper IPCSD 99–46, presented at the 1998 Industry Applications Society An- II. PROPOSED SENSORLESS IM DRIVE
nual Meeting, St. Louis, MO, October 12–16, and approved for publication in
the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Industrial Drives The proposed sensorless IM drive block diagram is shown in
Committee of the IEEE Industry Applications Society. This work was supported Fig. 1. It operates with constant rotor flux, direct stator flux,
by the Danfoss Professor Programme and the Institute of Energy Technology,
Aalborg University, Aalborg East, Denmark. Manuscript submitted for review
and torque control. The speed controller is a classical propor-
October 15, 1998 and released for publication August 23, 1999. tional-integral-derivative (PID) regulator, which produces the
C. Lascu is with the Department of Electrical Machines and Drives, reference torque. Only the dc-link voltage and two line currents
University Politehnica of Timisoara, RO-1900 Timisoara, Romania (e-mail:
cristi@et.utt.ro).
are measured. The IM model is
I. Boldea is with the Department of Electrical Machines and Drives, (1)
University Politehnica of Timisoara, RO-1900 Timisoara, Romania (e-mail:
boldea@lselinux.utt.ro).
F. Blaabjerg is with the Institute of Energy Technology, Aalborg University,
DK-9220 Aalborg East, Denmark (e-mail: fbl@iet.auc.dk).
Publisher Item Identifier S 0093-9994(00)00036-0. (2)
0093–9994/00$10.00 © 2000 IEEE
LASCU et al.: MODIFIED DTC FOR IM SENSORLESS DRIVE 123

Fig. 1. The DTC-SVM sensorless ac drive.

For rotor flux coordinates, the rotor flux components are

(3) (7)

(4) (8)

where is the stator voltage, , are the stator and rotor cur- The output of the open-loop current model (superscript “ ”)
rents, resepctively, are the stator and rotor flux, respec- is the stator flux calculated in stator coordinates
tively, are the motor parameters, is the
reference frame speed (arbitrary), is the rotor speed, and (9)
the derivation operator.
where is the estimated rotor flux from (7) and (8) in a sta-
The electromagnetic torque is
tionary reference frame (see Fig. 2).
(5) The voltage model is based on (1) and uses the stator voltage
with the number of pole pairs. and current measurement. For the stator reference frame, the
The stator flux and torque closed-loop control is achieved by stator flux is simply
the DTC-SVM unit. In order to reduce the torque and flux pulsa-
tions and, implicitly, the current harmonics content, in contrast (10)
to the standard DTC, we do use decoupled PI flux and torque In order to correct the value of estimated stator flux, to com-
controllers and SVM. pensate for the errors associated with pure integrator and stator
resistance measurement (estimation) at low speed and to pro-
III. FLUX AND SPEED ESTIMATOR vide a wide speed range operation for the entire observer, the
The estimator calculates the stator flux , the rotor flux , voltage model is adapted through a PI compensator
the electromagnetic torque , and the rotor speed . It is
(11)
based on the induction motor equations (1)–(5). The inputs of
the state estimator are the stator voltage and current space The coefficients and may be calculated such that, at
vectors. They are referred to a stationary reference frame. zero frequency, the current model stands alone, while at high
The flux estimator is a full-order wide-speed-range stator frequency the voltage model prevails (12)
and rotor flux observer [8], [9] (see Fig. 2). It contains two
(12)
models—the open-loop current model which is supposed to
produce an accurate value, especially for low-speed operation, Values such as –5 rad/s and 20–30 rad/s for the two
and the adaptive voltage model for wide speed range operation. poles of the closed-loop flux estimator transfer function
The rotor flux current model estimator (6) is deduced from are practical for a smooth transition between the two models.
(2) and (4) in a rotor flux reference frame ( , subscript The rotor flux is calculated in a stator reference frame
“ ”) using the measured stator current
(13)
(6)
The detailed parameter sensitivity analysis of this observer can
where is the rotor time constant. be found in [9].
124 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 36, NO. 1, JANUARY/FEBRUARY 2000

Fig. 2. The flux estimator for the DTC-SVM drive.

Fig. 3. The MRAC speed estimator.

The speed estimator has the structure of a model reference Fig. 4. The DTC-SVM controller.
adaptive controller (MRAC) [6], [7]. In order to achieve a wide
speed range, an improved solution, which uses the full-order
flux estimator, is proposed (see Fig. 3). IV. DTC-SVM
The reference model is the rotor flux estimator presented so
far (13). It is supposed to operate accurately for a wide fre-
The proposed topology of the DTC-SVM is shown in Fig. 4.
quency band (1–100 Hz). The adaptive model is a current model
The controller contains two PI regulators—one for flux and one
based on (2) for a stationary reference frame ( , super-
for torque—and an SVM unit. It receives as inputs the stator flux
script “ ”)
and torque errors and generates the inverter’s command signals.
(14) The – components of the reference voltage vector in a stator
flux reference frame are
The rotor speed is calculated and corrected by a PI adaptation
(17)
mechanism

(15)
(18)
applied on the error between the two models (13) and(14) From (1), for a stator flux reference frame ( —the
(16) stator flux speed, ), the voltage vector components are
The dynamic analysis of a rather similar speed estimator [6], (19)
[7] proves that the achievable bandwidth with which the actual (20)
speed can be tracked is only limited by noise considerations.
However, very-low-speed and fast dynamic operation remain an and the electromagnetic torque is
unsolved problem. (21)
LASCU et al.: MODIFIED DTC FOR IM SENSORLESS DRIVE 125

If the stator flux is constant, it is evident that the torque can be


controlled by the imaginary component —the torque com-
ponent—of the voltage vector
(22)
The stator flux speed is calculated in a stationary refer-
ence frame from two successive estimations of the stator flux
and as
Fig. 5. The SVM voltage vector timing.

(23) where is the stator flux position.


The precision of the calculation is not so important, since a PI The duty cycles and for each active vector are the
regulator is present on the torque channel. It corrects the torque solutions of the complex equation
even if the last term in (22) is erroneously estimated.
The flux control is accomplished by modifying the real com- (30)
ponent —the flux component—of the voltage vector.
For each sampling period , one can approximate the
voltage as
(31)
(24)

At high speed, the voltage drop can be neglected and (32)


the voltage becomes proportional with the flux change and
with the switching frequency 1/ . At low speed, the term where is the dc-link voltage.
is not negligible. The current–flux relations are rather compli- The duty cycle for the zero vector is the remaining time inside
cated (in stator flux coordinates) the switching period
(33)
(25) The vector sequence and timing during one switching period
is
(26)
where
The sequence guarantees that each transistor inside the in-
(27)
verter switches once and only once during the SVM switching
It is evident that a cross coupling is present in terms of period. A strict control of the switching frequency can be
and currents. The simplest way to realize the decoupling achieved by this approach. Fig. 5 shows the command signals
is to add the term at the output of the flux regulator in for the inverter when the vectors and and
the same manner as the speed dependent term was added to zero vectors and are applied.
the torque controller output. However, the computation of the A situation that must be considered appears when the
voltage drop term requires a time-consuming stator flux coor- control requirements overpass the voltage capability of the in-
dinate transformation. Instead of it, a PI controller was used on verter—the reference voltage is too high. The PI control method
the flux channel. does not guaranty for six-pulse operation. The adopted solution
The SVM unit produces the inverter control signals. It re- is to switch to the classical DTC when the PI controllers
ceives the reference voltages (17) and (18) in a stator flux refer- saturate. If the torque or flux is “far from target,” the respective
ence frame. The SVM principle [5] is based on the switching be- error is big positive or negative and the forward–backward
tween two adjacent active vectors and a zero vector during one DTC strategy is applied. A single voltage vector is applied
switching period. The reference voltage vector defined by during the whole switching period. It assures the target will be
its length (28) and angle (29) in a stator reference frame can reached quickly. If the torque and flux are “close to target,”
be produced by adding two adjacent active vectors the errors are small and now the SVM strategy based on PI
and and, if necessary, a zero vector controllers is enabled instead of applying a zero vector as a
or classical DTC would apply.
The “saturation point” for PI regulators is considered at
(28) . The voltage amplitude control becomes ineffective
for a reference voltage higher than , but the voltage angle
control is still effective. This observation permits to choose
the switching point from SVM to DTC at a relatively high
(29)
voltage—up to where the PI antiwindup becomes active.
126 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 36, NO. 1, JANUARY/FEBRUARY 2000

Fig. 6. The classical DTC controller.

Fig. 9. The estimated speed and torque with detuned estimator when R =
: R (!
04 ; M ) and when R = 1:8R (! ; M )—simulation
results.

Fig. 7. The real and estimated speed (! , ! ) and the real and estimated
torque (M , M ) with the tuned estimator—simulation results.

Fig. 10. The estimated speed and torque with detuned estimator when T =
: T (!
04 ; M ) and when T = 1:4T (! ; M )—simulation
results.

Fig. 8. The estimated speed and torque with detuned estimator when R =
0:4R (! ; M ) and when R = 1:4R (! ; M )—simulation Fig. 11. The experimental setup.
results.

The DTC strategy can be simply expressed: each sampling


The proposed strategy was called DTC-SVM because it re- period the adequate voltage vector is selected in order to rapidly
alizes the direct torque and flux voltage control combined with decrease, in the same time, the torque and flux errors.
SVM and uses DTC when the errors are large. The two methods The convenient voltage vector is selected in accordance with
are compatible since DTC is a high-gain voltage control. The the signals produced by two hysteresis comparators and the
classical DTC topology is presented in Fig. 6. stator flux vector position.
LASCU et al.: MODIFIED DTC FOR IM SENSORLESS DRIVE 127

Fig. 12. DTC-SVM—1 Hz (30 rpm) no load steady state—experimental Fig. 13. Classical DTC—1 Hz (30 rpm) no load steady state—experimental
results. results.
128 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 36, NO. 1, JANUARY/FEBRUARY 2000

Fig. 14. DTC-SVM no load starting transients—experimental results. Fig. 15. Classical DTC no load starting transients—experimental results.

V. SIMULATION RESULTS
inverter. The digital control system contains a digital signal
The simulation results with DTC-SVM are presented next. processor (DSP) (ADSP-21062) and a microcontroller (SAB
The induction motor used for experiments and simulations has 80C167).
the rated values kW, Hz, V, The DSP performs the calculations and the microcontroller
N m, pole pairs and the parameters , produces the PWM signals. The sampling time was 125 s and
, H, H, and H. The the switching frequency 8 kHz. Deadtime compensation was in-
simulation shows the estimator parameter sensitivityanalysis. cluded. Both DTC-SVM and classical DTC sensorless strategies
To determine the estimator performance, it is open-loop op- were implemented.
erated and the motor speed is controlled using the real speed. A The design of the two PI controllers is based on (22) and (24).
step reference speed from 0 to 50 Hz is applied at and a The torque controller gain should equal, at least, the first term
step from 50 to 1 Hz is applied at s. in (22): . The values 2–4 denote a high-gain
Fig. 7 shows the real and estimated speed and torque with torque controller, but are necessary to realize a fast torque re-
tuned estimator. A correct estimation can be observed. Fig. 8 sponse. A very high gain will produce operation similar to the
shows the estimated speed and torque when the stator resis- DTC. For the flux controller, the gain is smaller than
tance used for estimation is under and overestimated ( the switching frequency kHz, but the overall system’s
and ). Fig. 9 shows the same speed and torque when stability is improved, even if the flux controller is not a very fast
the rotor resistance for estimation is under and overestimated one. The integrator term in both controllers introduces a unitary
( and ). Figs. 8 and 9 indicate that the discrete pole and compensatesforthe cross-coupling errors.
system is relatively robust with respect to and . The controllers’ parameters used for experiments are the fol-
Fig. 10 shows the speed and torque when the rotor time con- lowing.
stant is under and overestimated ( and ). The
• The PI compensator for the flux estimator in Fig. 2 uses the
rotor time constant has a significant influence since the cur-
values and calculated for a transition
rent model is used in both flux and speed estimators. Very large between the two models around 10 Hz.
errors in estimation produce instabilities. • The PI speed estimator in Fig. 3 uses and
determined as in [7]. Higher gains produce
VI. EXPRIMENTAL RESULTS instability.
The experimental setup of the DTC-SVM system is shown • The PI torque and flux controllers in Fig. 4 use
in Fig. 11. The inverter is a 7-kVA industrial voltage-source , and .
LASCU et al.: MODIFIED DTC FOR IM SENSORLESS DRIVE 129

Fig. 16. DTC-SVM speed and torque transients zoom during no load
acceleration from 5–50 Hz—experimental results.

Fig. 18. DTC-SVM speed reversal transients (from 25 Hz to −25


Hz)—experimental results.

Comparative experimental results with low-speed no-load


operation are presented first. Fig. 12 shows the estimated
speed, torque, stator, and rotor flux, and the measured current
for steady-state 1–Hz DTC-SVM operation. Fig. 13 shows the
estimated speed, torque, stator, and rotor flux for steady-state
1–Hz DTC operation. An improved operation in terms of
high-frequency ripple can be noticed with DTC-SVM.
Fig. 17. Classical DTC speed and torque transients zoom during no load
The no-load starting transient performance is presented in
acceleration from 5–50 Hz—experimental results. Fig. 14—estimated speed and torque—for DTC-SVM and in
130 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 36, NO. 1, JANUARY/FEBRUARY 2000

Fig. 15—the same quantities—for DTC. Again, the torque ripple [8] P. L. Jansen, R. D. Lorenz, and D. W. Novotny, “Observer-based di-
is drastically reduced, while the fast response is preserved. rect field orientation: Analysis and comparison of alternative methods,”
IEEE Trans. Ind. Applicat., vol. 30, pp. 945–953, July/Aug. 1994.
The same conclusions are evident for the no-load speed tran- [9] P. L. Jansen and R. D. Lorenz, “A physically insightful approach to the
sients—from 5 to 50 Hz—presented in Fig. 16 for DTC-SVM design and accuracy assessment of flux observers for field oriented I.M.
and in Fig. 17 for DTC. A zoom of torque proves the fast torque drives,” IEEE Trans. Ind. Applicat., vol. 30, pp. 101–110, Jan./Feb. 1994.
[10] H. Kubota, K. Matsuse, and T. Nakano, “DTC-based speed adaptive flux
response of the proposed strategy. observer of induction motor,” IEEE Trans. Ind. Applicat., vol. 29, pp.
Fig. 18 shows the speed reversal from 25 to −25 Hz—speed, 344–348, Mar./Apr. 1993.
flux, and current—for DTC-SVM. Some small flux oscillations
can be observed when the flux changes due to the absence of the
decoupling term in the flux controller. Cristian Lascu received the M.Sc. degree in elec-
The system’s stability is influenced by the precision and the trical engineering from the University Politehnica of
Timisoara, Timisoara, Romania, in 1995.
speed of convergence of the flux and speed estimation. The He became an Assistant Professor in 1995 at the
speed estimator is not a very fast one, and this can be seen from University Politehnica of Timisoara. His research
Fig. 18 where some speed oscillations occur. The DTC-SVM areas are ac drives, power electronics, and static
power converters. In 1997, he was involved in the
controller does not depend on motor parameters and is relatively Danfoss Professor Programme in Power Electronics
robust as was proved by simulation. and Drives at the Institute of Energy Technology,
Aalborg University, Denmark. He is currently a
Visiting Research Scholar at the University of
VII. CONCLUSIONS Nevada, Reno.
Mr. Lascu was the recipient of a Prize Paper Award at the IEEE Industry
This paper has introduced a new direct torque and flux control Applications Society Annual Meeting in 1998.
strategy based on two PI controllers and a voltage space-vector
modulator. The complete sensorless solution was presented.
The main conclusions are as follows. Ion Boldea (M’77–SM’81–F’96) is a Professor of
Electrical Engineering at the University Politehnica
• DTC-SVM strategy realizes almost ripple-free operation of Timisoara, Timisoara, Romania. He has also
for the entire speed range. Consequently, the flux, torque, repeatedly been a Visiting Professor with the
University of Kentucky, Lexington, Oregon State
and speed estimation is improved. University, Corvallis, the University of Glasgow,
• The fast response and robustness merits of the classical U.K., and Aalborg University, Aaalborg, Denmark.
DTC are entirely preserved. He has worked and published extensively on
linear and rotary machines and drives, mainly on
• The switching frequency is constant and controllable. In linear motor Maglevs and linear oscilomotors and
fact, the better results are due to the increasing of the generators, vector control (direct torque and flux
switching frequency. While for DTC a single voltage control of both induction and synchronous motors), reluctance synchronous
machines, and drives and automotive new alternator systems. He has authored
vector is applied during one sampling time, for DTC-SVM and coauthored 11 books in English, the latest, with S. A. Nasar, being Linear
a sequence of six vectors is applied during the same time. Electric Actuators and Generators (Cambridge, U.K.: Cambridge Univ. Press,
This is the merit of SVM strategy. 1997) and Electric Drives (Boca Raton, FL: CRC Press, 1998).
• An improved MRAC speed estimator based on a full-order
rotor flux estimator as reference model was proposed and
tested at high and low speeds. Frede Blaabjerg (S’86–M’88–SM’97) was born in
Erslev, Denmark, in 1963. He received the Msc.EE.
It can be stated that, using the DTC-SVM topology, the overall degree from Aalborg University, Aalborg, Denmark,
system performance is increased. in 1987 and the Ph.D. degree from the Institute of
Energy Technology, Aalborg University, in 1995.
He was with ABB—Scandia, Randers, Denmark,
from 1987 to 1988. He joined Aalborg University in
REFERENCES 1992 as an Assistant Professor and became an Asso-
[1] I. Takahashi and T. Noguchi, “A new quick response and high efficiency ciate Professor in 1996 and a Full Professor in power
strategy of an induction motor,” in Conf. Rec. IEEE-IAS Annu. Meeting, electronics and drives in 1998. His research areas are
1985, pp. 495–502. power electronics, static power converters, ac drives,
[2] M. Depenbrock, “Direct self control for high dynamics performance of switched reluctance drives, modeling, characterization of power semiconductor
inverter feed AC machines,” ETZ Arch.., vol. 7, no. 7, pp. 211–218, 1985. devices, and simulation. He is involved in more than ten research projects with
[3] A. Mir, M. E. Elbuluk, and D. S. Zinger, “Fuzzy implementation of direct industry. Among them is the Danfoss Professor Programme in Power Elec-
self control of induction motors,” IEEE Trans. Ind. Applicat., vol. 30, pp. tronics and Drives.
729–735, May/June 1994. Dr. Blaabjerg is a member of the Industrial Drives, the Industrial Power
[4] D. Casadei, G. Sera, and A. Tani, “Stator flux vector control for high Converter, and the Power Electronics Devices and Components Committees
performance induction motor drives using space vector modulation,” in of the IEEE Industry Applications Society, as well as being the Paper Review
Proc. OPTIM’96, 1996, pp. 1413–1422. Chairman of the Industrial Power Converter Committee. He is a member
[5] P. Thoegersen and J. K. Pedersen, “Stator flux oriented asynchronous of the European Power Electronics and Drives Association and the Danish
vector modulation for AC-drives,” in Proc. IEEE PESC’90, 1990, pp. Technical Research Council and a Member of the Board of the Danish Space
641–648. Research Institute. In 1995, he received the Angelos Award for his contribution
[6] C. Schauder, “Adaptive speed identification for vector control of induc- in modulation technique and control of electric drives and an Annual Teacher
tion motors without rotational transducers,” IEEE Trans. Ind. Applicat., Prize from Aalborg University. In 1998, he received the Outstanding Young
vol. 28, pp. 1054–1061, Sept./Oct. 1992. Power Electronics Engineer Award from the IEEE Power Electronics Society
[7] H. Tajima and Y. Hori, “Speed sensorless field-oriented control of the and an IEEE TRANSACTION ON POWER ELECTRONICS Prize Paper Award for
induction machine,” IEEE Trans. Ind. Applicat., vol. 29, pp. 175–180, the best paper published in 1997. He also received two Prize Paper Awards at
Jan./Feb. 1993. the 1998 IEEE Industry Applications Society Annual Meeting.

You might also like