Model Predictive Control With Inherent CMV Reduction Capability for Multilevel Inverters

The conventional model predictive control (MPC) methods use either a cost function with weighting factors or an offline selection of voltage vectors to minimize the common-mode voltage (CMV) in three-phase multilevel inverters (MLIs). However, such methods significantly affect MLIs harmonic performance. Moreover, these methods are designed based on the three-phase modeling and implementation philosophy. Hence, they are difficult to extend for multiphase systems directly without affecting the computational complexity. In this article, an MPC method capable of minimizing CMV inherently is proposed. The proposed MPC is formulated by following the per-phase implementation philosophy. Moreover, the maximum number of predictions in the proposed MPC is limited to the number of output levels of MLIs only. Consequently, it results in a significant reduction in computational complexity. The proposed method is applied to a four-level MLI, and its performance is demonstrated through experimental studies on a dSPACE-controlled laboratory prototype. Furthermore, a performance comparison of the proposed and the conventional MPC methods is assessed in terms of CMV, total harmonic distortion, and computational complexity.