Torque Ripple Minimization of PMSM Drives with Multiple Reference Frame and DNN Based High-Fidelity Torque Model

Abstract

Torque ripple minimization is important to realize optimal operating state of permanent magnet synchronous machines (PMSMs). In order to suppress the torque ripples, this paper proposes a novel control technique for PMSM drives with a deep neural networks (DNN) based high-fidelity motor model and multiple reference frame (MRF). The high-fidelity model of PMSM is utilized to produce a reference current corresponding to a reversed torque harmonic to offset the torque harmonic inherent in the motor, and MRF is adopted to achieve follow-up control of high frequency signal. By optimizing the d-q axis current waveforms, the torque ripples in the resultant output torque will be minimized. The efficiency of the proposed scheme is explored through simulations. The simulation results of transient and steady state illustrate that the feasibility of the proposed control algorithm in torque ripple suppression.