Simulation and Validation of Permanent Magnet Synchronous Motor Drives Using Reinforcement Learning

There has been a drastic improvement in the mechanical design of Permanent Magnet Synchronous Machines (PMSM), improving the power density, torque ratio to inertia, and efficiency. However, the performance of the PMSM depends on the control algorithm used. Field Oriented Control (FOC) is the most commonly used control algorithm in Electric vehicles (EV) and Industrial drives. FOC algorithm provides a suitable voltage that has to be applied to machine terminals to achieve the desired speed. FOC algorithm implements two closed feedback loops, the external speed control, and the inner current control loop. Traditionally, Proportional Integral (PI) controllers are used to model the feedback loops. PI Controllers are proven to have robust performance, and the PI algorithm is easy to implement. However, the motor parameters are non- linear, resulting in reduced performance and efficiency as PI controllers are sensitive to plant parameters and decoupling inaccuracies. The proposed scheme FOC replaces the inner PI current control feedback loop with the RL controllers feedback loop, which helps to mitigate the decoupling inaccuracies and overcome the non-linearity in machine parameters. The simulation study is implemented using Matlab and Simulink. This paper aims to show that Reinforcement learning (RL) controllers have better reference speed tracking with no overshoot than the proportional-integral (PI) controller.