Model predictive torque control of PMSM based on data drive

Abstract Aiming at the problem of bad real-time performance caused by the large calculation of the finite control set model predictive torque control(MPTC), a data-driven method for MPTC of the permanent magnet synchronous motor(PMSM) is proposed. The data produced by PMSM-MPTC system are used to train DNN to learn its selective laws and then replace it to select the optimal voltage vector. This paper also analyzes the data-driven out-of-control problems caused by the difference between the steady data and dynamic data of the MPTC and solves the above issue by extending and balancing the data set, which further proves the feasibility of the data-driven MPTC. The simulation experiments show that the torque ripple, stator flux ripple and the system switching frequency of the data-driven MPTC are further reduced, indicating has certain superiority.