Real-Time SIL Validation of a Novel PMSM Control Based on Deep Deterministic Policy Gradient Scheme for Electrified Vehicles

Vector control plays a critical role in a permanent magnet synchronous motor (PMSM) drive to deliver the desired torque in electrified vehicle applications. Motor speed and stator current control depend on various nonlinear motor parameters that influence the performance of PMSM. Moreover, tuning of speed and current controller parameters using conventional control techniques also depends on these PMSM parameters. To enhance the robustness of vector control and tracking methodology against PMSM parameter uncertainties and load disturbances, a novel deep reinforcement learning (DRL) based advanced speed and current control technique is proposed in this article. The proposed method mitigates the effects of disturbance due to parameter variations as well as the load torque. The novel architecture delivers closed-loop reinforcement learning agents trained with the deep deterministic policy gradient learning algorithm in the plant environment where the cost of exploration is expensive. First, an overview and need for the proposed DRL vector control architecture are provided. Subsequently, the design and training methods for the proposed DRL controller are elicited. Thereafter, the proposed control scheme is validated with real-time software-in-the-loop testing under various conditions and compared against adaptive proportional–integral control of the same PMSM in OPAL-RT simulator.