4th International Conference on Electrical Engineering (ICEE 2015)

IGEE, Boumerdes, December 13th -15th, 2015

Induction Motor State Estimation using

Tuned Extended Kalman Filter
Samia Allaoui, Kheireddine Chafaa

Yahia Laamari

Electronics Department, Faculty of Technology

Laboratoire d'Automatique Avance et d'Analyse des
Systmes (LAAAS), University UHL BATNA
[email protected]
[email protected]

Electronics Department, Faculty of Technology,

Annaba University
Electrical Engineering Laboratory,
MSila University
[email protected]

Belkacem Athamena
Department of Management and Management Information Systems
Al Ain University of Science and Technology, UAE
[email protected]

Abstract As a main limitation in the state and parameters

estimation using Extended Kalman Filter (EKF) is that its
optimality is critically dependent on the choice of the right
covariance matrices of state and measurement noise. In order to
overcome this difficulty, a new approach based on the use of the
tuned EKF to estimate simultaneously the speed and rotor flux of
an induction motor drive is proposed. This approach will firstly
optimize the covariance matrices by the Particle Swarm
Optimization (PSO) algorithm and after that, the values of these
covariance matrices are introduced in the estimation loop.
Computer simulation results indicate an accurate estimation and
an acceptable performance in speed-rotor flux estimation after
considerable tuning of the covariance matrices coefficients and
confirm the efficiency of our proposed method.
KeywordsInduction Motors, Estimation, Parameters Tuning,
Stochastic and nonlinear observer, Extended Kalman Filter,
Particle Swarm Optimization.

I. INTRODUCTION
Induction Motors (IMs) are widely present in the main
industrial applications since they are very economical, rugged
and reliable. In mainly control applications of the IMs, the
information about rotor speed is very necessary and essential.
Usually, rotor speed can be directly measured by using an
optical encoder (sensors) which is mounted on the motor shaft.
However, the system cost, volume and weight of the motor are
increased; also, the reliability of drive reduces with hard
environments conditions [1]-[2]-[5].
In recent years, several approaches have been proposed to
eliminate the speed and flux sensors from industrial IM drives
for economical and technical reasons to reduce the cost and
enhance the reliability of the overall system [3]-[7].

Therefore, the need to use an auxiliary dynamic system,

called observer, which is responsible for estimating the state
of the system is necessary. These estimation tools are very
good alternatives to transducers and encoders.
In the last decade, the EKF is the most used observer for
stochastic nonlinear systems due to its ability to taking into
accounts the model uncertainties and inherent non linearities
which is the case of IM drives [3]-[7]. On the other hand, the
key problem of the EKF is that its estimation performance is
greatly influenced by the system parameters and covariance
matrices Q and R of state and measurement noises,
respectively. According to the Kalman filter theory, Q and R
are usually not known and in most cases are used as weighting
factors (tuning parameters).These matrices were first tuned
manually by trial and error methods which are very tedious
procedures [3].

To overcome this problem and to avoid the computational

complexity of trial and error method, genetic algorithms were
used to optimize matrices Q and R automatically [4].
In this paper, we propose a new alternative for the tuning
of Q and R based on a method called PSO, which is an
evolutionary algorithm inspired by social interactions.
Compared to genetic algorithms, PSO method is simpler and
more computationally efficient. The goal of our work is the
simultaneous estimation of the IM state especially rotor speed
and flux by a tuned EKF where the PSO algorithm is
employed.
The following section of this paper describes the basic
mathematical formulation of EKF algorithm. In Section 3 the
discrete IM models are presented. Section 4 gives the principle

2015 IEEE

concepts of PSO. Novel approach using both EKF and PSO

techniques are introduced in section 5. While simulation
results are examined in section 6. Finally, section 7 presents
some concluding remarks

two control variables (stator voltages). In the stationary

reference frame, the states and the stator voltages are:
x t

II. THE EKF ALGORITHM

The EKF is a recursive filter based on the knowledge of the
statistics of both states and noises created by system modeling
and measurements. EKF is an extension of the classical
Kalman filter which considers nonlinear dynamical systems
[7]. In this paper we shall attempt to find the best linear
estimate of the state vector xk of the following discrete-time
nonlinear dynamic system:
x k 1 f (x k , u k ,w k )
(1)
y k h (x k , v k )
where f (.) is the state evolution function representing the
system dynamics, h(.) represents the relationship between the
state vector and the observation y k , uk the system input at
time k and w k and v k are the process and measurement
white Gaussian noise vectors with zero mean and with
associated covariance matrices Q and R , respectively.
To allow application of Kalman filter to the nonlinearity (1),
this later must be linearized by using first order Taylor
approximation
near
a
desired
reference
point ( x k ,w k 0,vk 0) .Thus, the EKF algorithm can be
given by the following recursive equations:
x k 1/ k f (x k / k ,u k , 0)
Pk

1/ k

Pk

1/ k 1

Pk

1/ k

K k H k Pk

(2)

x2

u t

x3

x4

x5

(4)

T
s

v
y

(5)

(6)

Considering the state and the measure noises and. Since

Kalman filter is a discrete algorithm, then the following
discrete model for the IM is needed:
x k

1
y k

f (x (k ), u (k ),w (k ))
h x k ,v (k )

(7)

where f describes the discrete nonlinearity and h the

discrete linearity such as:
x1 (k 1)
f

x2 (k 1)
x3 (k 1)
x4 (k 1)
x5 (k 1)
1 T(

1
Ts

1 T(

1
Ts

Fk Pk / k FkT W k QW kT

K k Pk 1/ k H kT (H k Pk 1/ k H kT V k RV kT )
x k 1/ k 1 x k 1/ k K k ( y k h (x k 1/ k ,0))

x1

Tr
Tr

) x1 (k ) T
) x2 ( k ) T

1
LmTr
1
Lm

x3 ( k ) T

1
Lm

px5 (k ) x4 ( k ) T

px3 (k ) x5 (k ) T

1
LmTr

x4 ( k ) T

Lm
1
x1 ( k ) (1 T ) x3 ( k ) Tpx5 ( k ) x4 (k ) w3
Tr
Tr

Lm
1
x2 (k ) Tpx5 ( k ) x3 (k ) (1 T ) x4 (k ) w4
Tr
Tr

pLm
( x3 ( k ) x2 ( k ) x4 (k ) x1 (k )
JLr

(1 T

ff
J

) x5 (k ) T

Cr
J

1
v
Ls

w1

1
v
Ls

w2

w5

(8)

1/ k

where x k 1/ k is the a priori state prediction vector, x k 1 / k 1 is

the a posteriori state prediction vector, Pk 1/ k is the a priori
prediction error covariance matrix, Pk 1 / k 1 is the a posteriori
prediction error covariance matrix and K k is the Kalman
gain. F k , W k , H k and Vk are the Jacobean matrices defined
by:
f (x k ,w )
f (x , 0)
Fk
;W k
;
x
w
x x k
w 0
(3)
h (x k ,v )
h (x , 0)
Hk
;V k
x
v
x x k
v 0

III. DISCRETE INDUCTION MOTOR MODEL

The continuous IM model can be described by fifth order
nonlinear differential equations with four electrical variables
(currents and fluxes), a mechanical variable (rotor speed), and

x1 ( k ) v1
x2 ( k ) v2

(9)

where T is the sampling period.

IV. DEVELOPMENT OF PSO ALGORITHM

PSO is a technique for finding the best solution to a
problem inspired by the social behavior of bird flocking and
fish schooling. It was first introduced by Kennedy and
Eberhart in 1995 [6]. In PSO, the population is called a
Swarm, and each individual in the swarm is called a particle.
Starting with an arbitrarily initialized swarm population, each
member in PSO flies through a searching N-dimensional
solution space, and remembers the best encountered position.
Each particle in the swarm of size N is described by its
N
N
(i 1,2,..., N ) . The
position x i
and velocity v i
position of each particle is a potential solution, and the best
position that each particle achieved during the entire
N
optimization process is memorized as p bi
(best local

position). The swarm as a whole memorizes the best position

N
ever achieved by any of its particles as p g
(best global
position). Position optimality is measured by means of one or
more fitness functions defined in relation to the considered
optimization problem. The position and the velocity of each
particle in the k th iteration are updated according to the
following equations:

v i (k 1) w .v i (k ) c1 .r1 (k ). pbi (k ) x i (k )
c 2 .r2 (k ). p g (k ) x i (k )
x i (k

1)

x i (k ) v i (k

(10)
(11)

1)

where r1 (.) and r2 (.) are two random numbers uniformly

distributed in the interval [0,1] to provide a stochastic
weighting of the different components participating in the
particle velocities [6].
c1 and c2 are positive constants defining the cognitive
acceleration and the social acceleration respectively. w is the
inertia weight factor, it is used to control the impact of the
previous history of velocities on the current one [6].

V. TUNING THE COVARIANCE MATRICES OF EKF USING PSO

In the literature, these matrices were first tuned manually
by trial and error method which is very laborious tasks [4]. In
order to surmount this problem and to avoid trial and error
method, genetic algorithms were used to tune and optimize
matrices Q and R automatically [4].
In this section, we propose a new alternative for the tuning
and optimization of Q and R based on the PSO algorithm. The
structure of the method is constituted of two steps. As a first
step (see Fig.1) we present a PSO-EKF structure which works
in an offline manner and allows finding the optimal values of
Q and R. As a second step, obtained values Q and R from step
one are introduced into EKF estimator running online to
estimate the rotor speed and fluxes of the IM.
Actual (measured) rotor speed ( ) of IM and estimated
rotor speed ( ) of EKF are set to be inputs to a performance
evaluator through a comparator. The performance evaluator
calculates the fitness function which is a Mean Square Error
(MSE) criterion between
and . Then, obtained MSE will

be applied to the particle Swarm optimizer. Based on MSE

values, PSO optimizer will calculate and optimize the
unknown parameters unknown parameters of covariance
matrices Q and R by updating the particle solutions according
to updating (10) and (11) to provide particles better sets.
The new particle solutions and updated Q and R of the
EKF for the next iteration until a preset number, of iterations
has been reached, and then optimal values of Q and R are
obtained. Finally, optimized values Q and R are injected into
EKF estimator running online to estimate rotor speed and flux.
VI. SIMULATION AND RESULTS
In our simulations, the IM is fed via an ac drive in open
loop with a sinusoidal input voltage to achieve more realistic
tests with sampling period 0.0001sec. The ratings of the three
phase 50 Hz, 1.5 kW, 220/380V, 4 poles, 1420 rpm squirrel
cage induction motor are:
Rs
Lm

4.85 , R r

3.805 , L s

0.274 H , L r

0.258 H , J

.031 kg .m , f f

0.274 H

0.008 N .m / rad / s

In the simulation, error covariance matrix P of the EKF is

set to a unit 55 matrix and Q and R matrices with dimensions
55 and 22, respectively, are assumed to be:
Q diag ([q i s , q i s , q r ,q r , q ])
and
R diag ([ri s , ri s ]) .
Assuming that the two components of currents have
approximately the same magnitude and dynamics, the first two
entries in the diagonal of Q have been chosen to be identical .
For comparison purposes, the performance of the EKF with
diverse compositions of Q and R is evaluated by using the
MSE criterion between the estimated speed and the actual
rotor speed as follows:
MSE

1
K

K
i 1

est
i

(12)

where esti is the estimated speed,

K the number of data samples.

the actual speed and

In Table1 we show typical covariance matrices of EKF with

their corresponding performances (MSEs) obtained by
a trial and error method.
Table.1 Trial-error estimations and EKF Performances

Input

Induction
Machine

Measured rotor
speed

Measured
Output y

Performance
evaluator

EKF

Estimated rotor
Speed

Q,R

PSO

Mean Square
Error

Fig. 1 Diagram showing IM state estimation using PSO-EKF

Case

Values of Q and R

qi

ri

2.3761103

Estimation
quality
Very Poor

33.1022

Poor

MSE

0.1

qi

1e 7, q

0.01, ri

0.1

qi

0.0001, q

0.01, ri

0.1

2.2877

Good

0.1

0.3866

Very good

qi

0.001, q

0.001, ri

1600

Estimated speed
Actual speed

2000

Speed (rpm)

1400
1200

Speed (rpm)

1000

1600

1000

1500
0
1400

800

0.38
-1000

1550

600

0.1

0.2

0.3

0.4

1500

400

0.4

0.42

0.5
Time (sec)

0.6

0.5

2.04

1.5
Time (sec)

2.06 2.08

2.1

2.12

2.14

2.5

2.16

0
-1

Fig. 2 Speed estimation result with trial and error method

-2

TABLE2. EKF performance with PSO-EKF

PSO parameters: c1=1, c2=1.5 and w=0.8

0.05

0.1

0.15
Time (sec)

0.2

0.25

0.3

Estimated beta flux component

Actual alpha flu component

Values of Q and R

Flux (Web)

itera
tions

MSE

10

ri

0.205; q i

0.3424; q

0.0054; q

0.0117

0.3897

20

ri

1.1597; q i

0.2512; q

0.0152; q

0.0279

0.2619

ri

15.2927; q i

0.4156; q

0.2518; q

30

Flux (Web)

-200
0

1350

Worst estimation: case 1 of Table 1

Actual speed
Best estimation: case 4 of Table 1

0.9

Actual alpha flux component

1400

0.46

0.8

Estimated alpha flux component

1450

200

0.44
0.7

0.0823

0
-1
-2
0.5

0.51

0.52

0.53

0.54

0.0848

0.55
0.56
Time (sec)

0.57

0.58

0.59

0.6

Fig.3 Simulation results obtained by PSO-EKF(30 iterations)

In Table2 above we give optimized parameters of EKF

with their corresponding MSEs obtained by our proposed
approach PSO-EKF given in Fig. 1.

2000
1500
S p e e d (rp m )

We show in Fig.2 that a bad choice of R and Q values can

create large estimate error (case 1 in Table 1 and dashed line).
We note that the best value of MSE corresponds to the best
estimate of rotor speed (case 4 in Table 1 and solid line
in Fig. 2).

1000
1500
500

1400

-500
0

Estimated Speed
Actual speed

1300

0.58
0.1

0.2

0.6
0.3

0.62

0.64
0.4

0.66

0.68
0.5
Time (sec)

0.7

0.72
0.6

0.7

0.8

0.9

6.5
6

Rr +50%

5.5

R o to r r e s is t a n c e

To evaluate the effectiveness of our PSO-EKF method,

two scenarios with load torque variation and parameter
changes (due to temperature rise) are inspected: Fig.3 shows
the actual and estimated speed and rotor flux of the IM under a
load torque of 10 N.m applied in the interval from 0.4s to 0.7s
only. By inspecting the results of this first scenario, it can be
noted that we have a quite performance in the estimation of
the speed and in both components of the rotor flux. For the
purpose of parameter changes, the motor resistances used in
EKF model are kept constant at their nominal value. Fig.4
shows the actual and estimated speeds when the rotor
resistance is increased due to a load torque of 10N.m is
applied starting from 0.4s. For this purpose, PSO-EKF shows
a better accuracy of speed even when the rotor resistance is
increased by 25% or 50%.

Rr +25%

4.5

Nominal Rr

3.5
3
0

0.1

0.2

0.3

0.4

0.5
Time (sec)

0.6

0.7

0.8

0.9

Fig.4. Speed estimation result of PSO-EKF under rotor resistance variation

VII. CONCLUSION
This paper adapts a new method combining EKF and PSO
algorithms. PSO technique was used for the optimization of
the performances of EKF. The proposed approach enables the
noise covariance matrices Q and R, which directly affects the
estimation performance, to be suitably selected. Simulation
results demonstrated the efficiency of this approach, where
estimations are more accurate and faster. This estimation
algorithm was also tested under load torque and rotor
resistance variations of where better solution and good
performances were obtained.

APPENDIX
TABLE3 Nomenclature

, : stator index
u : stator voltage

L s (L r )
Lm

i s : stator current
r

: rotor flux

: rotor speed
T s (T r ) : stator (rotor) time

p
J
ff

: stator (rotor) inductance

: mutual inductance
: total leakage coefficient
: pole pairs number
: rotor inertia
: friction coefficient

References
[1]
[2]

[3]

[4]

[5]

[6]

[7]

P. Vas, Sensorless, vector and direct torque control, Oxford University

Press, London, 1998.
S. Aksoy, A. Muhurcu, H. Kizmaz, State and parameter estimation in
induction motor using the Extended Kalman Filtering algorithm., In:
Proc. of the International Symposium, Modern Electric Power Systems
(MEPS), Wroclaw, Poland, pp.1-5,2010.
M. Barut, S. Bogosyan, M.Gokasan, Speed-sensorless estimation for
induction motors using extended kalman filters., IEEE Trans. on
Industrial Electronics, vol. 54, pp. 272-280,2007
KL. Shi, YK. Wong, SL. Ho, Speed estimation of an induction motor
drive using an optimized extended Kalman filter., IEEE Trans. on
Industrial Electronics, vol. 49, pp. 124-133,2002
O.Aydogmus, S.Sunter, Implementation of EKF based sensorless drive
system using controlled PMSM fed by a matrix converter, Electrical
Power and Energy Systems, vol.43,pp.736-743,2012
Kennedy J, Eberhart RC, Particle swarm optimization, In: Proc. of the
IEEE International Conference on Neural Networks, Piscataway, USA,
pp. 19421948,1995.
T.Du, P.Vas, AF. Stronach, MA.Brdys, Application of Kalman filters
and extended Luenberger observers in induction motor drive, In: Proc.
Intelligent motion, pp 369-387,1994.