Predictive current limiter for LQ based control of AC drives

Abstract

Linear quadratic control (LQ) is a popular technique for design of control strategy. Its main advantage is the ability to optimize a quadratic cost function on a very long horizon (potentially infinite). In its basic form, it requires linear system which is not the case in many AC drive control problem. However, various approximations, such as gain scheduling or state dependent Riccatti Equation (SDRE) approach, were proposed and applied to this task. They have certain advantages, however, they all still suffer from the essential problem of LQ approaches, namely its inability to respect hard constraints on the system state variable. In this paper, we interpret the LQ controller as an approximate solution of the constrained problem and use the limited lookahead approach of approximate dynamic programming to introduce the constraint on the amplitude of the drive currents. This yields a constrained quadratic optimization problem, for which we design an iterative solution. Performance of the algorithm is tested in simulation and experimentally on a laboratory prototype of a 10.7 kW PMSM drive.