Model Predictive Current Control Using Single Layer Neural Network for PMSM Drives

Abstract

Model predictive control (MPC) is regarded as a significant modern control for the current control of permanent magnet synchronous motor (PMSM). However, the computation burden of MPC imposes its advantage to be implemented in sophisticated converter topologies and multistep prediction horizons. Multilayer neural network with MPC (MLNN-MPC) is increasingly used in different converters to overcome the drawback of high computational time. However, it has a higher computational time compared to a single-layer neural network (SLNN). In addition, many parameters need to be optimized such as initial weights, number of iterations, and neurons. In this paper, a SLNN with MPC is proposed to predict the current of PMSM. The proposed SLNN-MPC is trained using the Levenberg Marquardt algorithm. Meanwhile, it shows better performance than MLNN-MPC with lower computational time by optimizing only one parameter. Furthermore, the simulation results are shown to verify the effectiveness of the proposed method.