Model-Free Predictive Current Control for PMSM Drives Using an Improved Dynamic Linearization Data Model

Model predictive control has attracted much attention in electric drives, but its parameter sensitivity on explicit models poses inherent challenges to the further application. This paper proposes an improved model-free predictive current control (MFPCC) based on a full-form dynamic linearization (FFDL) data model. Through the pseudo gradient (PG) and estimation algorithm, the FFDL modeling method only uses the input-output data of the system. However, the accuracy of the classic FFDL data model relies heavily on the PG initial value, which is related to the plant. In order to reduce the initial value dependence, this paper further improves the FFDL technique with a diagonal gain matrix, while the stability analysis and parameter design guidance are given. With only simple parameter design, the proposed method can establish a prediction model based on the input-output data to deal with uncertain systems. Finally, the performance of the proposed method is verified on a PMSM experimental platform with a three-level neutral-point-clamped inverter.