Data-based adaptive learning control for uncertain nonstrict-feedback PMSM: A gradient descent approach

This paper investigates the rotational speed control problem for a sort of permanent magnet synchronous motor (PMSM) systems, in which a novel data-based gradient descent learning scheme is put forward. The presented control algorithm is designed with measurable states and predefined rotational speed, which relaxes the limitation of prior information from system dynamics, stator/rotor parameters, and external load torque, while simultaneously ensuring a more satisfactory speed regulation performance and allowing the drifting of system parameters. In this paper, the model of PMSM in d−q coordinate is transformed to the nonstrict-feedback form. Then, a gradient descent-based adaptive learning mechanism is constructed to improve the approximation accuracy of fuzzy logic systems (FLSs) to the optimal control input, by which the effects of uncertain PMSM dynamics and unknown control gain on speed regulation performance are effectively compensated. The Lyapunov theory shows the semi-global uniform ultimate boundedness (SUUB) of speed regulation errors. Finally, a simulation example and some comparison results are provided to validate the efficacy of the developed scheme in driving an unknown PMSM under different rotation commands.