An Advanced Policy Gradient Based Vector Control of PMSM for EV Application

Traditional proportional-integral (PI) based control has proven to be an influential control technique due to its advantage of easy implementation and robust performance for EV application. However, these controls are sensitive to plant parameters since the motor parameters change over time resulting in reduced performance and efficiency. This paper proposes a new vector control in permanent magnet synchronous motor (PMSM) with deep reinforcement learning (DRL) based controller in EV application. An advanced policy gradient algorithm is used to develop the DRL-based vector controller to overcome non-linearity in machine parameters and mitigate the decoupling inaccuracy issues in conventional control schemes. This contribution further evaluates the adaptive and dynamic performance against the conventional PI-based control and the research outlook is presented in the subsequent sections of this paper.