UKACC International Conference on Control 2012

Cardiff, UK, 3-5 September 2012

RESEARCH ON SPEED-SENSORLESS
INDUCTION MOTOR CONTROL SYSTEM
BASED ON AMESIM-SIMULINK SIMULATION
JIA Xiaoyan

XIE Xiaohua

State Key Laboratory of Automobile Simulation and Control

Department of Control Science and Engineering,Jilin
University
Changchun, PR China
E-mail: [email protected]

State Key Laboratory of Automobile Simulation and Control

Department of Control Science and Engineering,Jilin
University
Changchun, PR China
E-mail: [email protected]

WANG Xue

CHEN Hong

State Key Laboratory of Automobile Simulation and Control

Department of Control Science and Engineering,Jilin
University
Changchun, PR China
E-mail: [email protected]

China First Automobile Works Group

Research and Development Center
Changchun, PR China
E-mail: [email protected]

AbstractThis paper will establish the simulation model and the

inverter model of induction motor based on the multi function
simulation software AMESim. Considering the influence on the
motor from the temperature and other complex factors, and
Using Matlab / Simulink software to establish a simulation model
of direct torque control system, then we go on combined
simulation through the combination of the two aspects. The
simulation result shows that the controller make the speed faster
and the robustness stronger of the motor control system, but at
the same time it exposes the shortcoming that torque pulsing will
enlarge at the low speed. Pointing at the shortcoming, this article
starts with the principle of extended kalman filtering algorithm,
uses of measuring current, voltage on the pattern of stator,
deduces from a new motor speed, flux state observer model
without speed sensor, and estimates the rotor speed and stator
flux. The simulation result shows that this method can solve the
problem of flux and torque pulsing effectively.

Direct torque control is a new type of high performanced

AC adjustable-speed transmission control technique. It
abandonees the decoupling control theory in vector control. By
using of the stator flux orientation and instantaneous space
vector theory, and detecting the stator voltages and currents,
this control method observes flux and torque of motor in the
stator coordinate, then compares the observation value with the
given one. By hysteresis controller regulating the error value
gets corresponding control signal. To control motor, through
synthesizing flux and torque signal to select the corresponding
voltage space vector. The advantage of direct torque control is
that the torque has faster dynamic response, and the rotor
parameters change has certain robustness. The defect of this
control method is that torque and flux pulsing is strong
especially in the low speed.
The motor speed and flux observer given in this paper
check the feasibility of the algorithm, by computer simulation
and the comparison between the actual rotor speed and stator
flux. The simulation results show that the design of new state
observer can reduce the motor flux and torque pulsing
effectively, and improve the motor torque performance. And it
make high performance motor control possible.

Keywords- Extended Kalman filter; direct torque control; the

simulation software AMESim; torque ripple

I.

INTRODUCTION (HEADING 1)

The full name of AMESim is LMS Imagine.Lab AMESim,

that is complex system modeling and simulation platform of
the interdisciplinary field. It provides a complete platform of
system project design, this makes it possible that the user can
build complex model of multidisciplinary field system on a
platform. By using of various components provided by system,
the user can using the physical model to build the simulation
platform, and this improved the difficulties that general
modeling software to do it by using of complicated
mathematical model. AMESim model is more close to the
reality.

978-1-4673-1560-9/12/$31.00 2012 IEEE

II.

THE INDUCTION MOTOR STATE ESTIMATION BASED ON

THE EXTENDED KALMAN FILTER

The object of this paper is the AC induction motor, which is

a typical nonlinear system. Under normal circumstances, the
Kalman filter is used for state estimation of linear systems, if it
is used in the state estimation of nonlinear systems, you must
consider to use the extended Kalman filter (Extended Kalman

375

Filter, EKF). The largest difference between the general

Kalman filter and EKF is attempting to make the nonlinear
systems linearization [3-7]. After the Jacobian matrix
linearization, there is little difference with the general Kalman
filter in form.

1
L
s

0
B=
0

0
0

A. The Extended Kalman Filter Design Based On AC

Induction Motor
In two-phase stationary coordinate, in order to use EKF to
predict the rotor speed, we need to add to a new state variable
[8] [9]. The block diagram of the direct torque control system
based on the extended Kalman filter is shown in figure 1. That
is where we use voltage and current value which is easy to
measure to estimate the value of the stator flux and achieve the
speed identification based on the extended Kalman filter
algorithm. So as to improve the performance of induction
motor direct torque control system [10] [11].

1
Ls
0

0
0

1 0 0 0 0
C=

0 1 0 0 0

Where Rs is the stator winding resistance, Rr is the rotor

winding resistance, Ls is the stator cyclic inductance, Lr is the
rotor cyclic inductances, Lm is mutual cyclic inductance,

= 1

L2m

, n p is the number of pairs. r is the rotor

Ls r
speed. It can be seen from the above state equation expression,
in addition to the speed, the rest of the amount are all the stator
side variables or motor constants, but also it shows the system
is a nonlinear system. In other words we can use the extended
Kalman filter algorithm. The actual system contains the system
and measurement noise inevitably, so we add the noise matrix
to the fifth-order model of the AC induction motor. It
constitutes the following stochastic nonlinear control system:
x (t ) = f ( x(t )) + W (t )
y (t ) = h( x(t )) + V (t )

(2)

Among the formula:

f ( x(t )) = A x (t ) + B u (t ) , h( x(t )) = C x(t ) , W (t ) is the
system noise matrix, V (t ) is the process noise matrix. So the
space state equation of an AC induction motor conforms the
form of general nonlinear systems state equation in 2.1 section.
So this paper may be using extended Kalman filter algorithm to
research the AC induction motor. Firstly we should linear and
discrete the state space model (1) and (2) of the AC induction
motor. Specific process is as follows: we can expand the model
using of the Taylor series mentioned in the previous section,
and then ignore the higher order terms and retain only first
order, get the approximation linear model. Assuming that the
state equation of the discrete system is as follows:

Figure 1 the system block diagram of EKF-DTC

Under the conditions that the sampling period is very short,

we can think that the change is zero. Rewriting of the induction
motor state-space model:

x = A x(t ) + B u (t )
y (t ) = C x(t )

(1)

Among the formula:

State variable x = [i s i s r r r ]T , measurement

x = Ak x(k ) + Bk u (k ) + W (k )

variable y = [i s i s ]T , input variable u = [u s u s ]T ,

L Rr + L Rs

Ls L

A=
Rr Lm

Lr

2
m

2
r
2
r

0
2
m

2
r
2
r

L Rr + L Rs
Ls L

Lm Rr

n p Lmr

Ls L2r
n p Lmr

Ls L2r

Ls L2r

R
r
Lr

Rr Lm
Lr

n p r

Lm Rr
Ls L2r
n p r

Rr
Lr
0

y = Ck x ( k ) + V ( k )

The dispersion coefficient is :

T

Ak = e AT I + AT , Bk = e A Bd BT , Ck = C .
0

We substitute A, B, C into the above equation to get:

376

(3)

 L2m Rr + L2r Rs
Ts

Ls L2r

Ak =
Rr Lm

Ts
Lr

1
L
s

0
Bk =
0

0
0

Lm Rr

0
L2m Rr + L2r Rs
Ts
Ls L2r
0

Ts

Ls L2r
n p Lmr

Ts
Ls L2r

Rr
Ts
Lr

Rr Lm
Ts
Lr

n pr Ts

1
Ls
0

0
0

n p Lmr

Ls L2r
Lm Rr

Ls L2r

Ts
Ts

n p r Ts

Rr
Ts
Lr
0

1 0 0 0 0
Ck =

0 1 0 0 0

Figure 2. Simulink Simulation model based on EKF-DTC.

Known by the five formulas of the extended Kalman filter,

we must also know the nonlinear state matrix f , output matrix
h and vector partial differential matrix F H . The F is
called the Jacobian matrix, and the H is called the transfer
matrix.

Induction motor simulation parameters:

rated power PN =1.1kw, stator winding resistances Rs =5.793,
rotor winding resistances Rr =3.421, stator cyclic
inductances Ls =0.368H, rotor cyclic inductances
Lr =0.368H,mutual cyclic inductance Lm =0.363H,moment of
inertia J =0.0267Kg.m2,the number of pairs P =2.

We can get f = Ak x + Bk u from the formula

f ( x(t )) = A x(t ) + B u (t ) , and get h( x(t )) = C x(t )
from h = Ck x . So
F(x(t)) =

Based on the EKF-DTC method and the traditional DTC

stator flux vector, electromagnetic torque and current
simulation waveform, we compare the simulation results.

f (x(t))
x(t )=x(t)
x(t)

the given speed is 30 r / s (the rated speed is 150 r/s), and

the load torque is 3 N. m.

Rs (Ls ) 1 (Tr )
r (Ls ) is +r (Ls )
r
Rr Lr

r
Rs (Ls ) 1(Tr ) r (Ls ) Rr Lr
is r (Ls )

=
Rs
0
0
0
0

0
0
0
0
Rs

2
2
2
2
(n J)

(np J)r
(np J)is (np J)is
0
r
p

2
1.5

C
x(t )

x ( t ) = x ( t )

1 0 0 0 0
=

0 1 0 0 0

Startor Flux(Wb Wb

H=

After calculating the above functions and the corresponding

matrix, we can apply the five steps of the Extended Kalman
Filter. Substituting into the following recursive formula:

x(k + 1 / k ) = x(k ) + T f ( x(k ), u (k ))

(4)

Pk +1/ k = Pk + (F(x(k)) Pk + Pk FT (x(k))) T + Q

(5)

K (k + 1) = Pk +1/ k H T [ HPk +1/ k H T + R]1

Pk +1 = Pk +1/ k K (k + 1) H Pk +1/ k

(6)

1
0.5
0
-0.5
-1
-1.5

(7)
(8)
x(k +1) =x(k +1/ k) + K(k +1)(y(k +1) H x(k +1/ k))
Among the formulas, K is the Kalman gain, P is the state
error covariance matrix, Q and R are noise covariance matrix
whose initial value is artificially given. This article gets each
parameter by trial and error method under the premise of
ensuring the steady-state tracking and filtering divergence.

-2
-2

-1

Startor Flux(W b

(a1) stator flux waveform of traditional DTC

B. System Simulation Model Based On The EKF-DTC

System
When the simulation model is set up, the extended Kalman
filter is using of the S function. Figure 2 is the Simulink.
Simulation diagram with the EKF-DTC.

377

30

1.5
20

10

0.5

Current A

Stator Flux Wb

0
-0.5

-10

-1
-20

-1.5
-2
-2

-1

-30
0

0.5

1.5

2.5

Time s

Stator Flux W b

(c1) Current waveform of traditional DTC

(a2) stator flux waveform of EKF-DTC

1.5

60
50

0.5

30

current A

Torque N.m

40

20
10

-0.5

0
-10

-1

-20
0

0.5

1.5

2.5

-1.5
0

Time s

0.2

0.3

0.4

0.5

time(s

(b1) Electromagnetic torque waveform of traditional DTC

(c2) Current waveform of EKF-DTC

40

31

30

30.8
30.6

20

30.4

10

Speed(rad/s)

torque(N.m)

0.1

0
-10

30.2
30
29.8
29.6

-20

29.4

-30

29.2

-40
0

0.5

1.5

2.5

29
0

time(s)

0.5

1.5

Time(S)

(b2) Electromagnetic torque waveform of EKF-DTC

(d1) Given speed waveform

378

2.5

35
30

speed(rad/s)

25
20
15
10
5
0
-5
0

0.5

1.5

2.5

time(s)

Figure 4 AMESim simulation model of Co-simulation

(d2) speed waveform of EKF-DTC

Figure 3 Comparative experiment waveforms

III.

SIMULINK AND AMESIM SIMULATION REALIZATION

AND VALIDATION OF THE ALGORITH

A. The Structures Of The Joint Simulation Model

We use the motor model in AMESim to instead of the
induction motor model of control system in front. At the same
time its inverter drive section is also created in the software, to
verify the effect of this control algorithm designed. In
AMESim, the mathematical model of the induction motor takes
the relationship between the temperature and the parameters
into account, the system makes it closer to the actual motor
system environment, and makes the powerful validation of the
effectiveness of the proposed algorithm. In AMESim model,
the relationship between the temperature and the motor
parameters is as follows:

Figure 5 Simulink simulation model of Co-simulation

B. The Simulation Experiment Of Algorithm Verification

Simulation: the motor parameters still choose the same
parameters as the previous ones.
The experimental setup: the reference temperature is 25
degrees Celsius. The given speed is 30rad / s. Observing the
stator flux and speed simulation waveform.

Rs = Rs0 (1 + alphaRs (Temp T0 ))

tauR = Tr0 (1 + alphaTr (Temp T0 ))
Co-simulation electronic control system and Simulink
single software simulation model are different. In the cosimulation electronic control system, we introduce the voltage,
current sensor signal and speed signal in AMESim into the
Simulink model. We can calculate the inverse change switch
signal by the computing in the Simulink model and then act on
the inverter device of AMESim model. As shown in Figure 4,
the S-function in the simulation model is corresponding to the
AMESim co-simulation interface in figure 5, to complete the
data reception and transmission.

2
1.5

Stator flux Wb

1
0.5
0
-0.5
-1
-1.5
-2
-2

-1

Stator flux Wb

Figure 6. Stator flux waveform of Co-simulation

379

change the original open-loop stator flux model into a closedloop control, and eliminate the error in the ring. At the same
time EKF achieve speed identification and the speed sensor
control, and also maintain the advantages of the DTC fast
dynamic response.

35

30

speed(rad/s)

25

Analyzing Figure 6 and Figure 7

The stator flux trajectory is almost circular, but the
trajectory is coarse, and it shows that the pulse is large.
However, the shape and thickness of the track can still
determine that it is limited within an acceptable range. The
given speed is 30rad / s. It can be seen from the experimental
results that the actual speed can track a given speed. Above all
we can conclude that the design algorithm of this paper is
effective.

20

15

10

0
0

REFERENCES

10

time(s)

[1]

M. Depenbrock, Direct self-control of inverter-fed machine. IEEE

Trans. on PE, vol. 3, no. 4, pp. 420-429, 1988J. Clerk Maxwell, A
Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford:
Clarendon, 1892, pp.6873.
[2] I. Takahashi, T. Noguchi, A new quick-response and high-efficiency
control strategy of an induction motor. IEEE Trans. on IA, vol. 22, no.
5, pp. 820-827, 1986.
[3] ANDERSON, B.O., and MOORE, J.B.: Optimal filtering (Pren-ticeHall, 1979).
[4] JAZINSKI, A.H.: Stochastic processes and filtering theory
(Academic Press, 1970).
[5] KIM, Y., SUL, S., and PARK, M.: Speed sensor less vector con- troll of
induction motor using extended Kalman filter, IEEE Trans., 1994,
IA-30, (5), pp. 1225-1233.
[6] KUBOTA, H, and MATSUSE, K.: Speed sensorless field-ori-ented
control of induction motor with rotor resistanc adapta- tion, IEEE
Trans., 1994, IA-30, (5), pp. 1219-1224
[7] HENNEBERGER, G., BRUNSBACH, B.J., and KLEPSCH, T.: Fieldoriented control of synchronous and asynchronous drives without
mechanical sensors using a Kalman filter. Proceedings of EPE, Firenze,
Italy, 1991, Vol. 3, pp. 664671
[8] Zaky, M.S, Khater, M.M, Shokralla, S.S, Yasin, H.A Wide-SpeedRange Estimation With Online Parameter Identification Schemes of
Sensorless Induction Motor Drives[J]. IEEE Transactions Industrial
Electronics, 2009, 56(5): 1699-1707.
[9] Brandstetter, P. Kuchar, M. Vinklarek, D.Estimation Techniques for
Sensorless Speed Control of Induction Motor Drive[J]. IEEE
International Symposium Industrial Electronics, 2006, 1(7) :154-159.
[10] Ziane, H., Retif, J.M., Rekioua, T., Fixed-switching-frequency DTC
control for PM synchronous machine with minimum torque ripples[J].
Canadian Journal of Electrical and Computer Engineering, 2008, (33) 3:
183189.
[11] Dong Zhe, You Zheng. Anovel extended Kalman filter for a class of
nonlinear systems[J]. Progress in Natural Science, 2006, 16(9) :912-918.
[12] Dong Zhe, You Zheng. Anovel extended Kalman filter for a class of
nonlinear systems[J]. Progress in Natural Science, 2006, 16(9) :912-918.

Figure 7. The waveform of the observables under Low-speed

IV.

CONCLUSION

Analyzing figure 3:
1. It is visible from the comparison between Figure (a1)
and Figure (a2). When the motor is running at low speed 30rad
/ s, the stator flux trajectory is a hexagon in the Traditional
DTC method. The control effect is far from ideal. But the stator
flux trajectory is approximately circular in EKF-DTC method,
and the pulse also significantly reduces.
2. It is visible from the comparison between Figure (b1)
and Figure (b2). The electromagnetic torque ripple reduces to a
large extent based on the EKF-DTC method.
3. It is visible from the comparison between Figure (c1)
and Figure (c2). The current waveform pulse is obvious in
traditional DTC method. The current sine wave has been
significantly improved based on EKF-DTC method.
4. It is visible from the comparison between Figure (d1)
and Figure (d2). The value of a given speed is 30rad / s. The
EKF identification value is also about 30rad / s. That is to say
that we can achieve the speed recognition based on the EKFDTC and the no speed sensor technology.
It can clearly be seen that the method reduce the stator flux,
electromagnetic torque, the current waveform pulsation to a
large extent and improve the steady-state performance of a
direct torque control system based on EKF-DTC. Analyzing
the reason is that the stator flux observer model is the UI model
in the traditional DTC, the accuracy declines seriously at a low
speed. However the EKF can remove the error caused by the
cumulative integral. The introduction of EFK is equivalent to

380