Reinforcement Learning based Method for Online Parameter Identification of the Permanent Magnet Synchronous Machines

This paper presents a new method for identifying in-real time the permanent magnet synchronous machines (PMSMs)’ inductances by utilizing machine learning algorithms. The proposed technique combines deep learning (DL) and reinforcement learning (RL) algorithms with the aim of solving the problems of the existing online methods associated with the inverter’s non-linearity, noise, and the accuracy of the measured rotor position. Extensive experimental tests were implemented to judge the estimation accuracy as well as the effectiveness of the proposed solution under critical operating conditions including full load, variation of speed and field weakening conditions.