You are on page 1of 6

# Adaptive Control of the Three-Phase Squirrel Cage Induction Motor with Load Torque Estimator

Marian Gaiceanu, Emil Rosu, Cristi Dache, Romeo Paduraru, Traian Munteanu Dunarea de Jos University of Galati
Marian.Gaiceanu@ieee.org conditions of the parameters, the classical tests must be done . In order to avoid the above mentioned disadvantage, a modern technique based on MRAC , has been adapted for a vector controlled three-phase induction machine. The reference model response is compared with the speed of the three-phase induction motor, and the tracking error is used in order to adjust the control parameters [6-9]. The paper is structured as follows: the mathematical model of the three-phase induction machine in rotor field oriented reference frame is presented in Section 2, the MRAC in direct form is widely described in Sections 3, the load torque estimator is shown in Section 4, and the corresponding simulation results are provided in Section 5. II. MATHEMATICAL MODEL OF THE VECTOR CONTROLLED ELECTRIC DRIVE SYSTEM

I.

INTRODUCTION

Because of the parametric and structural uncertainty characterization of the three-phase squirrel cage induction machine (IM) an adaptive adjustment of control parameters is necessary. Model reference adaptive control (MRAC) of three-phase induction machine in rotor field coordinated is proposed in this paper. The most important perturbation of a drive is the load torque. The direct measurement of the load torque is possible by using adequate but expensive equipment. The second possibility consists of the integration of the mathematical model. Since the mathematical models of the AC drives are more complicated, and have the nonlinearities, the integration needs a strong hardware as a digital signal processing. The proposed method avoids these difficulties using estimator of the 2nd order, which can be implemented easier, because they do not need so complex calculus. Basically, for motor parameters determination the conventional methods are used, i.e. no load and locked rotor tests . Despite of the inflexible conventional methods, the on-line parameter identification ones are more advantageous. The recursive least square identification algorithm is the most used method , . However, in order to obtain the initial

The well-known mathematical model of the AC drive using three phase squirrel cage induction machine in rotor field coordinates supplied from the current inverter (Fig. 1), is of the form : di mR R dt + i mR = i sd 2 M Te = 3 p 1 + i mR i sq R (1) d m J = Te T L Fv m dt i dq = + sq m R i mR dt

in which : isd the flux component current; isq the torque component current; imR the rotor magnetizing current; m the instantaneous angular velocity of the motor; Te the electromagnetic torque; TL the load torque; Fv the viscous friction force; J the combined inertia of the motor and load; M the mutual inductance between the stator and the rotor d, q equivalent windings; R the rotor time constant; R the rotor leakage factor; p the number of the pole pairs. There are two operating domains for the control of the angular speed: the constant and the variable flux. In the first

460

## isA isB isC

is 3/2 is e-jq

iq s ids

kq -T L km
+Te

q +m

Fig. 1. The mathematical model of the three-phase vector controlled induction machine.

in which: the degree of the monic Hurwitz polynomials Nm(s) and Dm(s) are pm and nm, respectively; in order to obtain a zero tracking error, both relative * degree are the same n m = n* p. The model reference adaptive control synthesis with unity relative degree is presented as follows. A. The gradient adjustment law The filtered signals vector is
T v (t ) = v T u (t ) v y p (t )

case, the constant flux, isd, the component flux current, has a constant value corresponding to the rating flux, the latter case is for constant power operation at flux weakening. The derived state space mathematical model of the induction motor driven in rotor field coordinates frame, with the assumption of maintaining rotor magnetizing current at constant value  in a standard form is described as follows:
o Fv m (t ) = o J q (t ) 1 0 m (t ) q (t ) 0 Km + J i sq (t ) K q (2)

y p (t ) r (t )

2n p

(7)

1 + J Tl (t ) 0

in which r(t) is the reference signal, yp (t) is the output signal of the process, the dynamic of the filters connected at n 1 the control v u p , and at the output of the plant n p 1 vy are described below : o v u = v u + hu (8) o v y = v y + hy p
The pair (, h) can be chosen in controllable canonical ( n 1)( n p 1) n form: the matrix p and the vector h p ,

(3)

## v u ( s ) = ( sI ) 1 hU ( s ) v y ( s ) = ( sI ) 1 hY p ( s ) such that : det( sI ) = N m ( s )

(9)

Obviously, the condition isd=imR=ct can be accomplished by providing an adequate control of the current inverter. III. ADAPTIVE CONTROL SYSTEM DESIGN The control parameters vector is obtained by adding two terms (Fig 2): gradient adjustment control component and variable structure adaptive control = g (t ) + v (t ), u (t ) = T (t ) v(t ) (4)
where g 2 n is the parameters vector with gradient adjustment control, and v 2 n is parameters vector of variable structure. The transfer function of the dynamical system (2) is
H p (s) = K p N p (s) D p (s) (5)

(10)

For a gradient adjustment law the following parameters are obtained: o gu = g sign(k p ) vu e0 o gy = g sign(k p ) v y e0 (11) o gp = sign(k ) y e g p p 0 o gr = g sign(k p ) r e0

which guarantee the asymptotic cancelation of the tracking error lim e 0 (t ) = 0 , g being a positive constant. t The controller parameter vector

in which: the degree of the monic polynomials Np(s) and Dp(s) are mp, and np respectively; the sign of the gain Kp is known; the relative degree n * p = n p m p of Hp(s) is known. The transfer function of the reference model is given by N ( s) H m (s) = K m m ; (6) Dm (s)

2n p

T T = gu (t ) gy (t ) gp (t ) gr (t ) p n p 1

(12) .

## includes the subvectors gu

and gy p

n p 1

The adaptive control consists of a scalar product of the vectors g and v T u (t ) = g v(t ); (13)

461

Hm(s)

ym

r= (t)

gr vr

+ + + + + + +

pu - i*sq=u(t) Km Kq Hp(s)

+ +

T^L
1 Js+Fv 1

m
q

Hnm(s)+1

yp

e0

gu + + vu

( , h)

( ,h)

vu

gy + + vy vy gp + vp +

Fig.2. MRAC with vector controlled drive system for three-phase squirrel cage induction machine (SQIM)

B. Variable structure adaptive law The variable structure adaptive control is given by T v(t ); u(t ) = v
T T v (t ) = v u (t ) vy p (t ) y p (t ) r ( t )

The vT = vu vy p y p r vector components are the solutions of the following differential equations: (14)
vu = vu vu v e0 vu vy p = vy p vy p v e0 vy p y p = y p y p v e0 y p r = rr v e0r
o o o o

## The adjustment of the parameter vector v (t ) is based on the signum function:

vu = vusign( k p )sign( e0 vu ) vy p = vy p sign( k p )sign( e0 v y )
p

(17)

(15)

## p are positive diagonal in which vu , vy p p matrices, y p , r , v are positive constants.

(n 1)(n 1)

IV.

Variable structure adaptive control with parameter adjustment thanks to the signum function assures a fast and smooth system response, but with asymptotically small oscillations around the equilibrium point. Parameter T , will include an approximation of sign(x) adjustment law, v function by using a k-sigmoid function , :
e ke 0 v u 1 e ke0 v u + 1 e e
ke 0 v y ke 0 v y
p p

vu vu

sign( k p ) sign( k p )

The block diagram of the second degree estimator is presented in Fig. 3, where the input needs the measure of the angular speed m(s) and the calculus of the torque current component isq. The output of the estimator is the estimated load torque TL^(s) . The load torque estimator (Fig.3), after some manipulations, gets the form presented in Fig. 4. Using Laplace transform, the first equation from (2) (neglecting viscous friction) gets the form
- m(s) Km
+T e(s)

vy p vy p y p y p r r
e e

1 +1

ke 0 y p ke 0 y p

(16)

is q(s)

1 +1

+ -

sign( k p )

1/sJ

1/s

TL(s)

e ke0 r 1 e ke 0 r + 1

sign( k p )
Fig. 3. The mathematical model of the three-phase vector controlled induction machine.

462

sJ m ( s ) = Te ( s ) T L ( s )
or in another form

(18) (19)

Te ( s) sJ m ( s) = T L ( s)

This means that the block diagram from (Fig. 4) can be redrawing as in (Fig. 5).

A. Calculus of the load torque parameters The problem consists of the calculation of the parameters c and such that the error between the estimated torque TL^(s) and the actual torque Te(s) to be insignificant. The transfer function of the estimator (Fig.5) is given by:
G(s ) = 1 T L (s ) = TL (s ) s 2 J + s + 1 c
^

1 2

=e

(26)

(20)

## a small step time response

ta = ln( 0.05 1 2 )
2 0

## Considering a step variation for the TL(s), by setting:

m(s) sJ i sq(s) km T e(s) + c c+J 1/s TL(s)

(27)

## Fig. 4. The redrawing load torque estimator.

TL ( s ) =

TL s

(21)
1 + s + 1)

Fig.7 Matlab/Simulink implementation of the three-phase SQIM in rotor field reference frame

## the estimated torque gets the form

T L (s ) = TL
^

V. SIMULATION RESULTS (22) The MRAC of SQIM is shown in Fig. 2, in which Hm (s) is transfer function of the reference model and Hnm(s) is the transfer function of the unmodelled dynamics. The Matlab/Simulink-based compound law with both components gradient adjustment and variable structure adaptive control is presented in Fig.6. The mathematical model of a 7.5 [kW], 1480 [rpm] induction motor under a load torque of 5.5 [Nm] with a constant magnetizing current imR=isd=1.8 [A] is shown in the Figure 7. By using adaptive control (14), u(t ) = isq (t ) , the angular velocity m and the rotor field magnetizing angle (q) are delivered. The Forward Vector Transformation (FVT) block (Fig.7) transform the two phase (rotor field reference frame) components into three phase input currents (stator reference frame isA, isB, isC ) by using Park and Clarke transformations. In order to perform these transformations, the instantaneous angle of the rotor field, q, is required. The simulation results of the adaptive electric drive system for a no load starting, followed at t = 0.5[s] of an applied load torque to three-phase induction motor shaft, are shown in Fig. 8-14. The load torque estimator provides the information

s( s 2

J
c

T L (s ) = TL
^

1
2 2 s(T0 s

+ 2T0 s + 1)

(23)

=

0
2
1
2 T0

(24)
= c

## and the pulsation factor

2 0 =

(25)

The parameters c and are chosen such as that the response TL^(s) to have an acceptable overshoot
TL(s) + c 1/s TL(s)

c+sJ

463

## iqs current control 60 3

longitudinal current

40 iqs [A] isd [A] 0 1 1.5 2 2.5 time[s] speed reference-actual speed 0.5

20

0.5

2.5

## 0.4 0.2 0 -0.2 -0.4

nref, n [rpm]

0.5

1 1.5 time[s]

2.5

Fig.8-11 The main simulation results: the torque current component (isq), the flux component (isd), the speed reference (nref) and the actual one (n), the tracking error (e0).
A phase current 40 20 iA [A] 0 -20 -40 iB [A] 40 20 0 -20 -40 B phase current

0.5

2.5

0.5

2.5

iC [A]

0.5

1 1.5 time[s]

2.5

## Fig. 12-15. The phase currents under the load torque

about the load machine. Moreover, the load torque bandwidth is at low frequency. The load torque value is necessary in order to maintain the control under the admissible load torque; at maximum load torque the saturation nonlinearity appears and the control law deduction is more complex. Another reason to know the load torque variation is to use it

in order to synthesize another advanced control strategies, like optimal control. VI. CONCLUSION In order to evaluate the proposed MRAC three-phase IM, the simulation results based Matlab/Simulink software are provided. The model reference adaptive control, in direct

464

## 5 estimated load torque 4 TL_^, TL [Nm]

function) assures smooth response around the equilibrium point, robustness of the system, and a zero tracking error. The simulation results (Fig.16-17) confirm a good estimation of the load torque, without additional equipment. The utility of the load torque estimator is important for the torque control or for the advanced control synthesis. REFERENCES
 IEEE Standard Test Procedure for Polyphase Induction Motors and Generators (2004). IEEE Std 112-2004 (Revision of IEEE Std 1121996).  Cirrincione M., M. Pucci, G. Cirrincione, G.A. Capolino, A new experimental application of least-squares techniques for the estimation of the induction motor parameters, Industry Applications, IEEE Transactions on, vol.39, n.5, Sept.-Oct. 2003, pp. 1247-1256.  Netto A.J., P.R. Barros, C.B. Jacobina and A.M.N. Lima, Estimating the parameters of an induction motor in open-loop and closed-loop operation, Industry Applications Conference, 39th IAS Annual Meeting. Conference Record of the 2004 IEEE, vol.1, 2004, pp.26-31  Filipescu, A., A compound Law for Adaptive and variable Structure Control with Sigmoid Function, for the Process with Unitary Relative Degree Model", The 9'th Symposium on Modelling and Identification systems, SIMSIS8'94, Galati, Romania, 1994, pp.14-18  Petros Ioannou, and Baris Fidan , Adaptive Control Tutorial, The Society for Industrial and Applied Mathematics, 2006.  Cheng-Hung Tsai, Ming-Feng Yeh, Application of CMAC neural network to the control of induction motor drives, Elsevier, Applied Soft Computing 9, 2009, pp.11871196  Azzolin R. Z., and H. A. Grndling,.A MRAC Parameter Identification Algorithm for Three-Phase Induction Motors, Electric Machines and Drives Conference, 2009. IEMDC '09. IEEE International, pp.273-278  Zheng Li, Model Reference Adaptive Controller Design Based on Fuzzy Inference System, Journal of Information & Computational Science 8: 9, 2011, 16831693, http://www.joics.com  Hongjie Hu and Yuqiao Miao, A Model Reference Adaptive Control Scheme Based on Neural Network Systems Modeling and Simulation 2007, Part 2, Part 3, pp. 73-77  Murphy, J. M.D, and Turnbull, F. G, Power Electronic Control of A.C. Motors, Pergamon Press, 1988  Gaiceanu, M., and Rosu, E., Neuro-optimal controller for vector controlled induction motor, Proceedings of the 9th International Conference, EPE-PEMC 2000.

0.5

1 time[s]

1.5

2.5

Fig. 16. Comparision between the real load torque and the estimated one.

6

5.5

TL , TL [Nm]
5

4.5