You are on page 1of 5

First International Power and Energy Coference PECon 2006 411

November 28-29, 2006, Putrajaya, Malaysia

Simulation of MRAS-based Speed Sensorless


Estimation of Induction Motor Drives using
MATLAB/SIMULINK
Ahmad Razani Haron, Nik Rumzi Nik Idris, IEEE Member

Abstract - Model Reference Adaptive System (MRAS) based MRAS based techniques have been proven to be one of the
techniques are one of the best methods to estimate the rotor best methods being proposed by the researchers due to its
speed due to its performance and straightforward stability good high performance ability and straight-forward stability
approach. These techniques use two different models (the approach [1].
reference model and the adjustable model) which have made The method was first proposed in [2] followed by [3]
the speed estimation a reliable scheme especially when the
motor parameters are poorly known or having large which consists of a reference model (RM), an adjustable
variations. The scheme uses the error vector from the model (AM) and an adaptation mechanism. RM is
comparison of both models as the feedback for speed independent of the rotor speed whereas AM requires the
estimation. Depending on the type of tuning signal driving the rotor speed information. Through Landau' s idea of
adaptation mechanism, there could be a number of schemes comparing the outputs of RM and of AM, the error between
available such as rotor flux based MRAS, back e.m.f based the two models can be minimized using the adaptation
MRAS, reactive power based MRAS and artificial neural mechanism [4], as depicted in Fig. 1.
network based MRAS. All these schemes have their own trends Various structures of MRAS can be realized depending
and tradeoffs. In this paper, the performance of the rotor flux
based MRAS (RF-MRAS) and back e.m.f based MRAS on the quantities that form the error vector. They are rotor-
(BEMF-MRAS) for estimating the rotor speed was studied. flux based MRAS, back e.m.f based MRAS, reactive power
Both schemes use the stator equation and rotor equation as the based MRAS and artificial intelligence based MRAS [5].
reference model and the adjustable model respectively. The Each of these schemes has their own trends and tradeoffs.
output error from both models is tuned using a PI controller
yielding the estimated rotor speed. The dynamic response of
the RF-MRAS and BEMF-MRAS sensorless speed estimation
is examined in order to evaluate the performance of each
scheme.
Index terms - BEMF-MRAS, MRAS, parameter variations, RF-
MRAS, sensorless speed, tracking capability.

I. NOMENCLATURE

lds lqs, idr , lqr


, d- and q-axis stator and rotor currents in
the stationary reference frame.
Vds, Vqs Stationary reference frame d- and q-axis Fig. 1. Basic configuration of MRAS-based speed sensorless estimation
scheme.
stator voltages.
Ls Lr Lm Stator-, rotor- and magnetizing self This paper studied the performance of the rotor-flux
inductance.
Stator resistance. based MRAS (RF-MRAS) and back e.m.f based MRAS
Rs (BEMF-MRAS) for estimating the rotor speed. The
Tr Rotor time constant.
behavior of the estimators prior to the IM parameter
C4 )(r Actual and estimated rotor speed. variations also has been studied.
ds 2s dr' YVsr
s
d- and q-axis stator and rotor flux
linkage in the stationary reference
frame. II. MRAS ESTIMATORS
Total leakage reactance. The study in this paper compares the performance of two
MRAS techniques, the RF-MRAS and BEMF-MRAS
emd,emq,emd,emq Actual and estimated induced back
e e

estimator. Both schemes use the stator equation and rotor


e.m.f in d- and q-axis. equation as the reference model and adjustable model
respectively. The RF-MRAS estimator is constructed from
II. INTRODUCTION equations (la) - (Id) and the structure for this estimator is
The study of speed sensorless control of the IM has depicted in Fig. 2. The tuning signal for RF-MRAS
undergone through maturing years when new techniques estimator is the rotor-flux error vector which is fed to the
came into introduction to improve the previous techniques. adaptation mechanism to ensure that the system will be
The motivation is to find one method that can cater the stable and the estimated quantity will converge to the ideal
entire problem related to speed sensorless IM. Among them, (actual) value.
1-4244-0273-5/06/$20.00 (©2006 IEEE
412

where dimdd I I
d+
I s

(a) Reference Model: Tr Tr

qi=r (Vqs -Rsiqs )t-tLsJiqsJ (la) dimq I .I i I i


LM
dt dt rrmdT
T~r
mq
+Tr"
Tr

Vdr = k(vd-R js dt- LsJds] (lb)


LM

where = I
Ls Lr
(b) Adjustable Model:
d qr
(Ic)
dt T- VQr
Tr
+
WOrVdr+
Tr~.s
s

dt T Vdr Vqr +
Lr ds
(Id)
dt Tr Tr

Vd,

Fig. 3. Block diagram of BEMF-MRAS scheme.

III. ADAPTATION MECHANISM


The most important part of the estimation process is the
designs of the adaptation mechanism which at the end will
determine whether the system is stable or not. Therefore, the
estimated quantity will converge to the actual value with
suitable dynamics characteristics.
As described by Landau [4], the practical synthesis
technique for MRAS structures which is based on the
concept of hyper-stability can be used to design the
adaptation mechanism. These rules will ensure that the state
Fig. 2. Block diagram of RF-MRAS scheme.
error equations of the MRAS structures are globally

For the BEMF-MRAS estimator, it is constructed using asymptotically stable. However, this method does not
establish the dynamics of the convergence process which in
induced back e.m.f as given by equations (2a) - (2d). The
turn, analysis of the system equations, linearized about a
structure is realized in Fig. 3. The distinct feature of BEMF-
selected operating point is required.
MRAS from RF-MRAS estimator is the elimination of pure
In this paper, the detail of the adaptation mechanism
integration both in the RM and AM.
design is not elaborated since it has been clearly discussed
in [3] and [6]. Therefore, the design of the adaptation
(a) Reference Model: mechanism for RF-MRAS and BEMF-MRAS estimators
has yielded the estimated rotor speed as given by equations
dis (3) and (4) respectively.
emd = Vds s ids s (2a)
dt )

I) K +
K

(3)
emq =qs - jiqs +s dtj (2b) (r p )I dndi

Or = (Kp + s)em (S em ) (4)


(b) Adjustable Model:
IV. SIMULATION RESULTS
emd = L (_r mc
- i + iI s ) (2c) The simulations have been carried out using
Tr
MATLAB/Simulink platform and the estimators were
implemented in the direct torque control IM drive. The
irinmd qs ) (2d) parameters of the IM are given in Table 1.
mq L
L
qT
+
r
413
TABLE 1
INDUCTION MOTOR'S PARAMETERS FOR SIMULATION
Parameter Value
Stator Resistance (Q) 5.5
Rotor Resistance (Q) 4.51
Stator Self Inductance (H) 0.3065
Rotor Self Inductance (H) 0.3065
Mutual Inductance (H) 0.2919
Moment of Inertia (kgm2) 0.03
Load Torque (Nm) 1
No. of Poles 4
Rated Speed (rpm) 1410

(c)
The performance of the estimators has been studied in
terms of its ability to converge to the actual speed and the
response of the estimators towards parameters changes.
Thus, to fit that purpose, the values of parameter in the IM
are kept unchanged while the values of the estimators are
varied from its rated value.
Fig. 4 shows the speed response of the RF-MRAS speed
estimator when the IM is operated at different reference
speeds. It shows that the estimated speed tracks the actual
speed reasonably well at high speed and even at low speed.
But, as the speed decreased, the steady state values of the
estimated speed tend to deviate from actual speed. This X agD S aI 11 a2 D:M .3 a3 S a 45 D
deviation is caused mostly by the error in the stator
resistance Rs [2]. (d)
As a result, the speed estimation below reference speed
Fig. 4. RF-MRAS estimator's tracking performance at reference speed
of 20rad/s is not applicable and generalization of the results (a) 100rad/s, (b) 70rad/s and (c) 50rad/s (d) 30rad/s.
applied only for this range of speeds.
In addition to its well tracking performance, the effect
Weeid Response
of parameter variations at high speed is quite negligible but
as the frequency decrease, for stator resistance, the term
Rsi/p becomes relatively large and therefore the values of
stator resistance will seriously affect the estimated speed.
This situation is depicted in Fig. 5.
The effect of rotor resistance and moment of inertia
variations also being studied but the changes is not
significant since no direct relationships establish between
these parameters with rotor speed as given in equations (la)
(2d).
In the case of BEMF-MRAS estimator, it shows an
(a) improvement in terms of speed tracking performance and
effect of parameter variations as compared to the RF-
0 m Om FIS D -32 3 am 0 a4D Us MRAS. The good tracking capability as depicted in Fig. 6 is
due to the elimination of the pure integration process in the
reference model as shown in the equation (2a) - (2d).
This estimator also has less parameter dependent
compared to its predecessor. In Fig. 6, prior to four different
reference speeds, it shows that the estimator has better speed
tracking response at high and even at low speed.
BEMF-MRAS estimator shows an outstanding
performance over incorrect setting of parameter values
(especially stator resistance value, Rs) even as large as 100%
deviations from the rated value.
The BEMF-MRAS estimator's response at different R,
(b) values is depicted in Fig. 7. In this figure, it shows that this
scheme is completely robust to parameters variations since
no significant changes in estimated speed can be noted prior
to the variations.
_f\-~ spedro_
414
Effect of Rotor Resistance Variation Speed Response
r r l r r r ~ ~
100 actual - a4ual
"If
-

80 fi 4 'b
- estimated
~~~~~~~~~-
speed error_
100
=- eastimrated
1
60 -
80
40 -

f
-7'
28
20 I .a 60
lt7- 0
EU

cn -20 - .40
co
-40

-60 -
20

-80 -

0
-100

0 0.2 0.4 0.6 O.S 1 1.2 1.4 1.6 1.8 2 c0 0.05 0.1 0.1 5 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time (s) time (s)
(a) (a)

Effect of Stator Resistance Variation Sopeed Response


80
ioc0 _ pffAfS ~~~~~-
a4tual
4 i ~~~~~~~~~-
eslimated
oc0 .7/ \2 ~~~~~~~~~~-
speeci error_- 70

6C0 60

4C0
50
'.;' 2C0
.8 .8 40
2 2
-
I0

. 30
co -2C0 co
-4C0 20

-6C0 0

-8c0
0
-1 oc0
-lo
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0.05 0.1 0.1 5 0.2 0.25 0.3 0.35 0.4 0.45 0 .5
time (s) time (s)
(b) (b)

Effect of Stator Resistance Variation Sopeed Response


ioc0
f/2 -~~~~~~~~~~~~~esftimated - eslimated
8c0

6C0
_ ,r i~~~~~~~~~~~
4C0

'j;' 2C0
.8
-
2
I0 KJ zuv1X\ f8fd
40

co -2C0 cofA
-4C0

-6C0

-8C0

-1 oc0
-1 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.05 0.1 0.1 5 0.2 0.25 0.3 0.35 0.4 0.45 O.E.5
time (s) time (s)
(c) (C)

Effect of Stator Resistance Variation Speed Response


35
ioc - a4ual
estimated - eslimated I. f
8c

6C 25 ,, _
4C

'j;' 2C .n.
.8
2 fz

-
I
10 . ,errIorI
spe fd
10 _
co -2C
0 0. 0. 0. 0. 1 1. 1. 1. 1. 2
-4C
,Y
-6C -5 _

-8C I1

-1 oc
0 0.05 0.1 0.1 5 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time (s) time (s)
(d) (d)

Fig. 5. Effect of incorrect setting of RS values to the RF-MRAS estimator's Fig. 6. BEMF-MRAS estimator's tracking performance at reference speed
speed response. (a) Rs (b) Rsnew = 1.1 Rs (C) Rsnew = 1.5 Rs (d) Rsnew = 2 RS. (a) 100rad/s, (b) 70rad/s and (c) 50rad/s (d) 30rad/s.
8~ ~actul
~ /=~ ~esrtimad_
415
Speed Response
V. CONCLUSION
100 < X ~~~~~~~~- actual_

80
Performance of RF-MRAS and BEMF-MRAS
60
estimators based on the tracking capability and parameter
40
sensitivity was presented. The result shows that the BEMF-
20 MRAS estimator is more superior to the RF-MRAS
0 estimator at that particular defined range of reference
-20 speeds. This is prior to the elimination of pure integrators
-40 used in the RF-MRAS scheme. However, the BEMF-MRAS
-60
estimator is more difficult to design due to the non-linear
-80
effect of the adaptation gain constants.
-100
c0 0.2 0.4 0.6 08 1.2 1.4 1.6 1.8 2
Therefore, as a whole, considering all the key criteria of
timne (s) comparison, it can be concluded that the BEMF-MRAS
(a) scheme embrace the requirement as a versatile estimator. It
demonstrate good tracking capability and superb in
Effect of Stator Resistance Variation
insensitivity to parameter variations.
100 - actual _

80 VI. REFERENCES
60 [1] M. Ta-Cao, Y. Hori and T. Uchida, "MRAS-based speed sensorless
40
control for induction motor drives using instantaneous reactive
20
power", IEEE-IES Conference Record, pp. 1717-1422. 2001.
[2] S. Tamai, H. Sugimoto, M. Yano, "Speed-sensorless vector control of
0
induction motor with model reference adaptive system", Conf. Record
-20
of the 1985 IEEE-IAS Annual Meeting, pp. 613-620, 1985.
-40 [3] C. Shauder, "Adaptive speed identification for vector control of
-60 induction motor without rotational transducers", IEEE Trans. Ind.
-80 Application, Vol. 28, No. 5, pp. 1054-1061, Sept./Oct. 1992.
-100
[4] Y.P. Landau, "Adaptive Control: The model reference approach",
co 0.2 0.4 0.6 0.8 1.2 1.4 1.6 1.8 2 Marcel Dekker, New York, 1979.
timne (s) [5] M.N. Marwali, A. Kehyani, "A comparative study of rotor flux based
(b) MRAS and back e.m.f based MRAS speed estimators for speed
sensorless vector control of induction machine", IEEE-IAS Annual
Meeting, New Orleans, Louisiana, pp. 160- 166, 1997.
Effect of Stator Resistance Variation
100
/ ~~~~~~~~~-
estimated
80 _~~~~~~~~~~~~
-error

60

40

20

-20

-40

-60 0 0. 0. .6 12 14

-80
8~~~~~~~~~tm (s
I -100
;~~~~~c

100

80

60

40 o 0.2 0.4 0.6 0.8 1.2 1.4 1.6 1.8 2

20

-20

-40

-60

-80

-100

time (s)
(d)

Fig. 7. Effect of incorrect setting of Rs values to the BEMF-MRAS


estimator's speed response. (a) Rs (b) Rs,ew = 1.1 Rs (c) Rs,ew = 1.5 Rs
(d) Rs,ew = 2 Rs.

You might also like