关键词: 永磁同步电机, 模型预测转矩控制, 空间虚拟矢量调制, 中心差分卡尔曼滤波

Abstract: Aiming at the issues of weight factor tuning,significant torque ripple,and strong parameter dependence in traditional model predictive torque control, an improved multi-vector model predictive torque control strategy based on parameter identification is proposed. The reference voltage vector is derived using deadbeat control, simplifying the cost function to an expression containing only the voltage vector error in the rotating reference frame, thereby eliminating the challenge of tuning the flux and torque weighting factors. Based on the phase angle and amplitude of the reference voltage vector,sectors are dynamically divided, and candidate voltage vectors are pre-selected. This compresses the number of
candidate vectors to 2 ~ 4, significantly reduces computational complexity and suppresses torque ripple compared to the traditional approach employing eight candidate voltage vectors. A central difference Kalman filter is utilized to estimate and compensate for motor parameter variations in real-time,thereby enhancing the accuracy of the predictive model. Simulation results demonstrate that,under steady-state operating conditions,the total harmonic distortion of the current is reduced from 6. 43% to 2. 45%. This indicates that the proposed strategy effectively diminishes torque ripple and improves steady-state performance. Furthermore,under conditions of parameter mismatch,the strategy achieves reductions of 26% in torque ripple and 34% in flux linkage error,respectively,exhibiting strong parameter self-adaptiveness.

Key words: permanent magnet synchronous motor, model predictive torque control, spatial virtual vector modulation, central difference Kalman filter