Multiple-Voltage-Vector Model-Free Predictive Deadbeat Control With Updated Reference Voltage Vector for PMSM Drive

Abstract

Model predictive control (MPC) has been extensively investigated for its impressive dynamic response and efficacy in controlling non-linear systems. However, MPC faces challenges in terms of robustness, primarily due to its dependency on system parameters. In contrast, model-free predictive control (MFPC) offers an alternative that relies on the inputs and outputs of the system without applying system parameters in each control period. To avoid the problem of the stator current gradient update stagnation existing in the conventional MFPC strategy, this article proposes an enhanced MFPC strategy based on a novel reference voltage vector (VV) updating mechanism. In the proposed strategy, the reference VVs for the present control period are determined by evaluating the difference between the reference current and the sampling current, which is related to the previously calculated VVs of the contiguous control period. By substituting the current gradient with the updated reference VVs, the proposed strategy effectively reduces the interferences arising from inaccurate current gradients, enhancing the steady-state performance of the system. The proposed MFPC is verified using a laboratory 0.5 kw permanent magnet synchronous motor drive setup.