An Adaptive Model-Free Predictive Current Control Method for PMSM Drives

Model-free predictive current control (MFPCC) of permanent magnet synchronous machines (PMSMs) can be achieved by utilizing measured input–output data to eliminate the influence of machine parameter mismatches and enhance the system robustness. Leveraging the relationship between current differences and applied voltage vectors, MFPCC can update the current differences based on two or three successive voltage vectors to estimate the current difference for the other voltage vectors. However, this approach may lead to stagnation when only two vectors are applied for an extended interval. Unlike existing slope-based or look-up table (LUT)-based MFPCC methods, this article proposes an adaptive MFPCC (AMFPCC) method with a modified current difference updating technique. By generating a reference voltage vector based on the tracking error, AMFPCC can ensure continuous and accurate updates of current differences, even when successive voltage vectors are similar. This reference vector adjusts the current differences in each sampling interval, preventing stagnation and improving prediction accuracy. The effectiveness of the proposed AMFPCC is confirmed by numerical simulations, experimental tests, and robustness evaluations compared to other MFPCC schemes and conventional control methods. AMFPCC achieves a 32% reduction in current total harmonics distortion (THD) compared to field-oriented control (FOC) and significantly reduces torque ripple and improves dynamic response.