A GRU-based Automatic Weighting Factors Optimization Strategy for PMSM

The performance of model predictive control (MPC) in permanent magnet synchronous motor (PMSM) still remains a challenging problem due to the large torque ripple in lower speed. The cost function in conventional MPC generally use fixed weighting factors, which is not easy to adapt different operation scenario and keep balance between torque ripple suppression and current tracking accuracy. To address this issue, a gated recurrent units (GRU) network based weighting factors automated-regulation strategy is proposed for the MPC in PMSM. Through offline training, the weighting factors in MPC are tuned in real time and the current tracking accuracy as well as torque ripple in different conditions are ensured, which is superior to the conventional fixed weighting factors-based MPC. Finally, some simulation results verify the effectiveness of the proposed method.