Model Predictive Control of 5L-ANPC Converter-Fed PMSM Drives with Two-Stage Optimization

Classical model predictive control (MPC) of a five-level active neutral-point-clamped (5L-ANPC) converter-fed PMSM drive faces two limitations: heavy computation burden and poor steady-state performance. This paper proposed an MPC scheme with two-stage optimization to reduce the computational burden and improve steady-state performance. To simplify the computational complexity, the 5L-ANPC converter is decoupled into two parts: the low-frequency cell (LFC) and the high-frequency cell (HFC). In the first stage, the switching states of the LFC are selected based on the sign of the desired output voltage; in the second stage, the optimal duty cycles for the high-frequency cell (HFC) are calculated by the multiple vector MPC. The pulse train for each switching state is generated by phase-shifted pulse-width modulation (PS-PWM) based on the voltage-second balance principle. The phase capacitor and DClink capacitor voltage balance are achieved by the flexibility of the inherent redundancy in the 5L-ANPC converter. An efficient optimization method is also formulated to reduce the classical enumeration algorithm to only 6 times, which significantly simplifies the computational burden of MPC. Due to the interleaved switching manner within each phase, the steady-state performance compared with well the linear controller with PSPWM. Experimental evaluations have been presented to validate the effectiveness of the proposed MPC.