Optimization of T-Type Three-Level Inverter Algorithm Based on Finite Basis Model Predictive

The conventional Finite Control Set Model Predictive Control (FCS-MPC) algorithm is characterized by challenges in weight coefficient tuning during multi-objective optimization control, which compromises grid-connected current quality. To address this limitation, an improved FCSMPC algorithm is proposed. Switching frequency control is achieved by comparing distances between current reference values and their predicted counterparts without involving weight coefficients, while grid-connected current quality is enhanced through selective traversal of current value functions. The influence of switching vectors on neutral-point voltage was systematically investigated: Zero vectors and large vectors that exert no impact on neutral-point voltage were directly applied. Medium vectors were replaced with virtual vectors to eliminate neutral-point voltage deviations, whereas small vectors were coordinated with their redundant counterparts for neutralpoint voltage regulation. Simulation results demonstrate that the proposed algorithm effectively reduces neutral-point voltage fluctuations, improves current waveform quality, and achieves significant switching frequency reduction.