Model-Free Predictive Control for PMSM Drives Using Ultra-Local With Poincaré Mapping Causality

Model-free predictive control (MFPC) has become a popular choice for addressing the robustness limitations of model-based predictive control (MBPC), by replacing physical models with data-driven ones. However, to accurately describe the motion characteristics and operating states of the plant, data-driven models need to consider more flexible freedoms and achieve better adaptability. To achieve this improvement, an MFPC using an ultra-local model with Poincaré mapping (ULM-PM) causality is proposed in this article, and applied in a permanent magnet synchronous motor (PMSM) driving system. Based on the extended Poincaré mapping and physical model of PMSM, ULM-PM incorporates additional factors of the coupling current component and electrical angular velocity, into its causality, and this allows it to handle multiinput signals effectively. Furthermore, a fast–slow recursive-least-square algorithm is devised to estimate and update the model online, which does not require excessive processor resources. Theoretical analysis demonstrates that ULM-PM and its MFPC have system stability and reflect more linear components than the conventional MFPC using ULMs. As the experimental validations, the proposed method also shows advantages in terms of objective quality and model adaptability with enhanced robustness, compared to conventional strategies.