Adjacent-Vector-Based Model Predictive Control for Permanent Magnet Synchronous Motors With Full Model Estimation

The finite-control-set model predictive control (FCS-MPC) shows appealing features in permanent magnet synchronous motor control but suffers from high current ripples. Multiple-vector-based MPC (MVB-MPC) methods reduce the ripples but may lose some of the merits possessed by the FCS-MPC. Furthermore, the performance of both the FCS-MPC and the MVB-MPC is dependent on the accurate knowledge of the system parameters. This article proposes an adjacent-vector-based MPC (AVB-MPC) combining the FCS-MPC and the MVB-MPC. Instead of finding a solution with minimum cost, the AVB-MPC searches from a lookup table of adjacent vectors for the solution with the least switching actions to keep the cost within a tolerable band. Hence, the current ripple is reduced under the same switching frequency, while the merits of fast dynamic responsiveness and inherent overmodulation ability are retained. In addition, a full model estimation method is proposed so that the knowledge of the machine parameters is not required. The rank deficiency problem of multiparameter identification is also solved by the tolerable band. And these benefits are gained without adding much computational effort. Finally, the effectiveness and benefits of the AVB-MPC are validated by comparative experimental results comparing with several other typical predictive controllers.