On-Line Neural Network Controller for High Efficiency Operation of IPM Synchronous Motors

Abstract

In this paper an on-line neural network-based scheme is proposed for high efficiency operation of an Interior Permanent Magnet Synchronous Motor (lPMSM). The dynamics of the motorlload are modeled on-line and controlled using two neural network (NN)-based schemes. The first controller (NNT) is designed to control the motor torque. The output of NNT is the q-axis stator current equivalent to the torque command. The second controller (NNI) is designed to satisfy two goals; first it adjusts the d-axis stator current using a pre-determined control law for minimizing the iron and cupper losses. Secondly, it adjusts the IPMSM speed to follow a reference speed. Simulation results demonstrate the simultaneous IPMSM torque and speed control while maintaining minimum losses. The response of the controllers to step changes in torque and speed are also presented, proving the feasibility of the two NN-based controllers.