ANN Optimization of Weighting Factors Using Genetic Algorithm for Model Predictive Control of PMSM Drives

Model predictive control (MPC) has become one of the most attractive control techniques due to its outstanding dynamic performance for permanent magnet synchronous motor (PMSM) drives. However, the tuning of weighting factors for the cost function of MPC is a time-consuming procedure and weighting factors can significantly affect torque ripples of PMSM drives. For the purpose of optimizing the weighting factors in cost functions of MPC, this article proposes an artificial neural network (ANN) based method, which applies genetic algorithm as the back propagation algorithm. Since this method is trained offline, it does not increase any computation complexity of MPC. Consequently, any optimization targets that combine the performance metrics of MPC are defined and the optimal weighting factors for the given targets can be fast and precisely found through the proposed method. Besides, this method is robust to the variations of motor parameters like stator resistance and permanent magnet flux. The superiority of the proposed method over conventional ANN is evaluated by comparative simulations, in terms of current total harmonic distortion, tracking performance, and neutral-point potential balance. Finally, the feasibility and robustness of the proposed method are verified on the platform of PMSM drives fed by three-level T-type inverter.