Data-Driven Model Predictive Current Control of PMSM Incorporating an Adaptive Machine Learning Model

Model predictive control (MPC) has become a cornerstone in controlling permanent magnet synchronous motor (PMSM) drives powered through voltage-source inverters. However, MPC models rely on an accurate mathematical model of the PMSM and are generally built based upon accurate parameters of the motor. This in turn poses significant challenges in real-time control of PMSMs because of dynamic variations of motor parameters. To tackle this shortcoming of conventional MPC approaches, this paper is devoted to the design of a novel and efficient model-free data-driven control strategy incorporating an evolving fuzzy system (EFS). In particular, the proposed method initially makes use of historical data collected from the MPC solution to the control problem. The structure and parameter learning of EFSs is then realized through the collected data. The constructed data-driven EFS model is then employed in place of the MPC module to deal with the control problem, enabling the real-time solution. The resulting predictive EFS model, equipped with a rule base, achieves superior steady-state and dynamic performance compared with the MPC solution without relying on the accurate mathematical modeling of the PMSM. The attained results validate the robustness and effectiveness of the proposed data-driven control strategy in various scenarios.