AI based Current Controller for Permanent Magnet Synchronous Motor

Due to the widespread usage of a Permanent Magnet Synchronous Motor in the automotive and industrial applications, it calls for efficient control techniques that cater every specific requirement. A novel Reinforcement Learning approach has been introduced to control the inverter current in a Permanent Magnet Synchronous Machine. The reference current to the reinforcement learning block has been provided by a speed controller loop. The system has been simulated in MATLAB - Simulink using the reinforcement learning toolbox. Error between the reference and the measured currents are passed as the observation vector and the eighth possible switching states in a three phase inverter has been considered as the action vector. The reward function has been modelled such that the agent aims at reducing the error between the reference and the measured current. The proposed technique is observed to provide better results in terms of power quality, ripple and DC link voltage utilization.