A New Model Reference Adaptive Control of PMSM Using Neural Network Generalized Inverse

摘要

A new strategy of model reference adaptive control (MRAC) system based on neural network generalized inverse (NNGI) algorithm, termed as MRAC-NNGI system, is proposed for the current and speed regulations of permanent magnet synchronous motor (PMSM) drives. Due to the fact that PMSM is a multivariable nonlinear system with strong couplings, this paper gives an analysis of generalized reversibility combined with NN. The developed scheme of NNGI is transformed into a pseudo-linear system from connecting the motor plant and achieved the purposes of decoupling and linearization with Levenberg-Marquardt algorithm off-line. Therefore, an adjustable gain of closed-loop adaptive controller is developed by introducing MRAC into this kind of pseudo-linear system. The self-adaptive law is given for the gain regulation of linear system. Comparison of simulation results from others widely used algorithms confirm that it incorporates the merits of model-free learning, high-precision tracking and strong anti-interference capability.