Model Predictive Control of Three-Level NPC Inverter-Fed PMSM Drives Based on a Novel Vector-Selection Scheme

Existing model predictive control (MPC) methods mostly adopt multi-vector mode to achieve better steady-state control performance. But this increases system complexity, especially for three-level inverters. In addition, various vector combinations need to be evaluated in the cost function, and cumbersome tuning of weighting factors is also involved when the common-mode voltage (CMV) and neutral point potential (NPP) imbalance issues are considered. This paper proposes a novel multi-vector-based MPC scheme to deal with these challenges. The key is to map the reference voltage vector to sub-hexagons, and the candidate region is narrowed down. Then, the dwell time of the determined voltage vectors is obtained from the cost function, which minimizes the error between the predicted reference voltage vector and the synthesis vector. In addition, the basic vectors with higher CMV amplitudes are reconstructed, and the NPP imbalance is addressed due to the employment of a hysteresis controller. Experimental results verify that the proposed method has superior performance to other multi-vector MPC algorithms.