Sie sind auf Seite 1von 2

Speed regulation of an induction motor using methods from nonholonomic control

K. Y. Pettersen and O. Egeland Department of Engineering Cybernetics Norwegian University of Science and Technology N-7034 Trondheim, Norway
where a and b denote the components of a vector with respect to a xed stator reference frame. The rotor ux is , the stator current is i, the rotor speed is !, the stator voltage input is u, and TL is the load torque. The model parameters are the moment of inertia J of the rotor and of any tool attached to it, the number of pole pairs np , the stator and rotor resistance Rs and Rr , the autoinductances Ls and Lr , the mutual inductance M , and = 1 ? (M 2 =Ls Lr ). Note that for the development of the control law in Section 3 we assume the ideal conditions of zero load torque and full state measurements. In 1] it was established that nonholonomic systems cannot be asymptotically stabilized by a continuous purestate feedback law. One approach to evade this negative result has been the use of continuous time-varying feedback laws. 6] was the rst to use this approach, showing how such a feedback law can asymptotically stabilize a nonholonomic vehicle. In this paper we will show how this approach can be used to derive an exponentially stabilizing feedback law for the induction motor. Let !ref denote the reference speed of the induction motor. Let z = ! ? !ref , and let x = (z; a ; b ; ia ; ib ). De ne d id (z; a ; b ; t) = ?k1 ( a ? a (z; b ; t)) (2) a d (z; ; t) = sin(t=") (3) b a d (z; ; t) = ?k ( ? d (z; ; t)) (4) ib b 2 b b b 2 d (5) b (z; b ; t) = ? z sin(t=")
4 1 (6) (z; b) = (z 2 + b ) 4 We propose the following continuous periodic timevarying feedback law ua (x; t) = ?k3 (ia ? id (z; a ; b ; t)) (7) a d (z; ; t)) ub (x; t) = ?k4 (ib ? ib (8) b k1 ; k2 > 0, sgn(k3 ) = sgn(k4 ) = sgn( Ls ).

The paper shows how methods developed for control of nonholonomic vehicles can be used to derive a feedback control law for speed and ux regulation of an induction motor. The controller stabilizes the system, giving exponential convergence to the reference speed. Simulation results are presented. The paper shows how methods developed for the control of nonholonomic and underactuated vehicles can be used to derive a feedback control law for speed and ux regulation of an induction motor. Using this approach, the following is achieved: As opposed to eld oriented control of the induction motor where the ux amplitude is regulated to a nonzero reference value to yield linear rotor speed dynamics, the ux amplitude here instead varies as a continuous function of the speed error. This is an interesting property from an energy point of view. The proposed feedback control law provides a stable closedloop system, and the speed is regulated to the reference speed with exponential convergence. Moreover, the proposed control law is continuous and has no singularities for ux amplitude or any other variable equal to zero. The proposed controller does not cancel any dynamics. The stability result does not depend on exact knowledge of the model parameters, and consequently the exponential stability is robust to model parameter uncertainty. This is a useful property, for instance the value of the rotor resistance can vary and often the exact value is not known. It seems like a promising approach to combine the knowledge gained from the control of induction motors and control of nonholonomic systems. In 2] a eldoriented control approach is used to control a nonholonomic double integrator. The following model is used np M TL != _ ( a ib ? b ia ) ? JL J

Abstract

1 Introduction

3 The control law

2 The induction motor model

_a =

r Rr r ? Lr a ? np ! b + Rr Mia L _ b = ? Rr b + np ! a + Rr Mib (1) Lr Lr 2 2 M Rr + Lr Rs np M 1 MRr _ ! b?( ua )ia + ia = a+ Ls L2 Ls Lr Ls L2 Ls r r


=

_ ib

MRr M 2 Rr + L2 Rs np M r )i + 1 u b b 2 b? L L ! a?( Ls Lr L s L2 Ls s r r

For su ciently large absolute values of k1 , k3 and k4 , and for su ciently small values of " > 0, the feedback control law (7{8) locally exponentially stabilizes the equilibrium (!; a ; b ; ia; ib ) = (!ref ; 0; 0; 0; 0) of (1). Proof. The system (1) with the coordinate transformation (!; a ; b ; ia ; ib ) 7! x = (z; a ; b ; ia ; ib ), and with

Proposition 1

control (7{8), can be written x = l(x; t) + h(x; t) _ (9) where 2 3 np M JLr ( a ib ? b ia ) 6 7 r ? Rr a ? np !ref b + Rrr M ia 6 7 L L 6 7 Rr Rr M i ? Lr b + np !ref a2+ Lr 2 b 6 7 l(x; t)=6 M Rr +Lr Rs np M 1 7 6 MRr2 a + L L !ref b ? ( L L2 )ia + L ua 7 4 Ls Lr 5 s r s r s M 2 Rr +L2 Rs np M MRr 1 r )ib + Ls ub 2 b ? Ls Lr !ref a ? ( 2 Ls Lr Ls Lr Figure 1: The rotor speed ! rad/s]. and h(x; t) is de ned by the remaining terms. As the system satis es the conditions of 4, Prop. 2], the equilibrium x = 0 is locally exponen- ux regulation of an induction motor. The resulting contially stable with respect to the dilation (x; t) = trol law yields stability and exponential convergence to ( 2 z; a ; b ; ia ; ib ; t) if it is a locally asymptotically the speed and ux reference. The ux amplitude takes values depending on the speed error, as opposed to takstable equilibrium of the system ing a nonzero prescribed value. The control law has no x = l(x; t) _ (10) singularities, it does not cancel any dynamics, and the exWe reduce the system (10), by de ning id =ia , id =ib ponential stability provided by the control law is robust a b d and a = a as new control variables. The control func- to model parameter uncertainty. However, the results are tions are chosen to be the functions given in (2{4). The local. Also, the robustness to the load torque disturbance resulting system is is not included in the analysis. For methods developed for nonholonomic systems, little work has been done on the np M 2 (?k2 b sin(t=") ? 2k2 z sin (t=")) (11) z= _ analysis of robustness to disturbances. Future work is in JLr progress to achieve global exponential stabilization, and _ b = ? Rr (1 + k2 M ) b + (np !ref ? 2Rr Mk2 z ) sin(t=") to analyse the robustness of the controller to load torque Lr Lr (12) disturbance. Also, inclusion of a state observer should be analyzed in future work. The \averaged system" 3] of (11{12) is np M z (13) References z = ?k2 _ 1] R.W. Brockett. Asymptotic stability and feedback staJ Lr bilization. In R. W. Brockett, R. S. Millman, and Rr _ b = ? (1 + k2 M ) b (14) H. J. Sussmann, editors, Di erential Geometric ConLr trol Theory, pages 181{191. Birkhauser, 1983. The origin of this system is asymptotically stable, and thus by 3, Th. 4.1] there exists an "0 > 0 such that for any 2] G. Escobar, R. Ortega, and M. Reyhanoglu. Regula" 2 (0; "0 ) the origin of (11{12) is locally asymptotically tion and tracking of the nonholonomic double integrastable. tor: A eld-oriented control approach. To appear in We then use a backstepping approach, and extend Automatica, 1997. the system (11{12) by including the equation for _ a and 3] R. T. M'Closkey and R. M. Murray. Nonholonomic d replacing a (z; b ; t) by a , and similarly extending the systems and exponential convergence: some analysis _ system with the equation for ib and nally with the equatools. In Proc. 32nd IEEE Conf. on Decision and _ tion for ia . The extended system satis es the conditions of Control, pages 943{948, San Antonio, Texas, Decem5, Prop. 1], and thus for su ciently large values of k1 , k3 ber 1993. and k4 the origin of (10) is asymptotically stable. Then, by 4, Prop. 2] (z; a ; b ; ia ; ib ) = (0; 0; 0; 0; 0) is a locally 4] P. Morin and C. Samson. Time-varying exponential stabilization of the attitude of a rigid spacecraft with exponentially stable equilibrium of (9), with respect to two controls. In Proc. 34th IEEE Conf. on Decision the dilation (x; t). and Control, pages 3988{3993, New Orleans, LA, Dec. 4 Simulations 1995. The simulations were performed under the ideal assumptions of zero load torque and full state measure- 5] K. Y. Pettersen and O. Egeland. Robust control of an ments. The speed reference was 100 rad/s, and the initial underactuated surface vessel with thruster dynamics. values of the system were all zero. The parameter " in In Proc. 1997 Am. Control Conf., Albuquerque, New the control law was chosen to be 0.005. The speed conMexico, June 1997. verged to the reference value. The other variables varied as functions of the speed error, and nally converged to 6] C. Samson. Velocity and torque feedback control of zero as we have no load torque. This is consistent with a nonholonomic cart. In C. Canudas de Wit, edithe analysis of Section 3. tor, Advanced Robot Control, Proc. of the Int. Workshop on Nonlinear and Adaptive Control: Issues in 5 Conclusions Robotics, Grenoble, France, Nov. 1990, pages 125{ The paper has shown that methods developed for con151. Springer-Verlag, 1991. trol of nonholonomic vehicles can be applied to speed and
angular speed [rad/s] 100 90 80 70 60 50 40 30 20 10 0 0 0.5 1 1.5 2 2.5 time [s] 3 3.5 4 4.5 5

Das könnte Ihnen auch gefallen