A Novel Reconstructed Voltage Predictive Control for the Multilevel T-NPC Converter: Design, Simulation, and Experimentation

This article presents a novel reconstructed voltage model predictive control (RVMPC) approach for a three-level T-type neutral point clamped inverter. Unlike the conventional model predictive controllers, which require exhaustive searches across the entire finite state space, the proposed method uses the load model to estimate the reference voltage and exploit a level-shifted pulsewidth modulation (LSPWM) technique to control the switching actions. The integration of RVMPC with LSPWM not only ensures a consistent switching frequency throughout operation but also effectively balances the midpoint dc-link voltage. In addition, the proposed modulation technique significantly reduces the computational time and complexity compared to the existing finite control set model predictive control (FCS-MPC) methods. By eliminating the need to search the entire state-space, the approach does not require optimization of weight factors or cost function control, thereby simplifying the overall design. The proposed RVMPC method reduces computational time from 17 μs (as seen in the conventional FCS-MPC) to 6 μs, achieving a 64.7% reduction. Furthermore, it improves dc-link voltage utilization and enhances the inverter performance, with a 7% reduction in line-to-line total harmonic distortion (THD) and over 72% improvement in phase current THD. The effectiveness and performance of the proposed modulation technique are validated through computer simulations and experimental verification on a small-scale laboratory prototype.