Abstract

Since a data-driven model is adopted to describe the operating state of the plant in the model-free predictive control, it has been widely used in the motor driving realm to eliminate the influences caused by parameter mismatches and enhance the robustness of the system. However, due to the fixed model structure and heavy calculating process, it is difficult to obtain an improved control performance using time-series models in continuous-control-set (CCS) predictive algorithms. To solve these problems, a model-free predictive current control (MF-PCC) using adaptive ultralocalized time-series is proposed in this article, and applied to a permanent magnet synchronous motor driving system as the current controller. The model structure is improved as a variable, and its orders are online adjusted according to the designed adaptive law and the current operating state of the system. The complex discrete-time transfer functions in the model are ultralocalized to simplify the realization in the CCS-type controller. All required coefficients in the model are estimated by the recursive least squares algorithm, and the optimal gain is also found by the particle swarm optimization algorithm. The effectiveness of the proposed method is demonstrated by the experimental results, as well as the advantages of the proposed method, including better model accuracy and current quality with suitable robustness compared with the conventional time-series based MF-PCC.