Radial Basis Function Neural Network Optimization of Weighting Factors for Model Predictive Control of LSLSM

Long stator linear synchronous motor (LSLSM) has been considered in the traction of high-speed maglev train, while its control with multi-mode modulation become quite complicated. The model predictive control (MPC) is very attractive for motor drives with multiple objectives. However, it takes a lot of time to adjust the weight factors, and the weight factors significantly affects the inverter switching frequency, neutral-point (NP) voltage balance and the total harmonic distortion (THD). To optimize the weight factors of cost function in MPC, a weight factor optimization method based on radial basis function neural network (RBFNN) is proposed in this paper. The RBFNN has a simple structure and can predict the optimal weight factor quickly and accurately. In addition, the proposed method has strong robustness for the resistance change of the feeder cables and the stator. Finally, the superiority of the proposed method in weight factor optimization is evaluated by simulation.