Long-Horizon Nonlinear Model Predictive Control of Modular Multilevel Converters

Modular Multilevel Converters (MMCs) are a topology that can scale several voltage levels to obtain higher efficiency and lower harmonics than most voltage-source converters. MMCs very attractive for renewable energy applications fast charging stations electric vehicles, where they improve performance reduce costs. However, due the complex architecture large number of submodules, current control modular multilevel converters is challenging task. The standard solution in practice relies on hierarchical decoupling single-input-single-output loops, which limited performance. Linearization-based model predictive was already proposed MMCs, as it optimize transient response better handle constraints. In this paper, we show validity linear MMC models significantly limits prediction horizon length, propose nonlinear MPC (NMPC) solve issue. With NMPC, employ long horizons up 100 compared 10, limit range model. Additionally, an alternative corresponding cost function, enables directly controlling circulating improves capacitor voltages’ behavior. Using state-of-the-art sequential quadratic programming developed NMPC algorithm meet real-time constraints MMCs. A comparison with time-varying linearization-based used ultra-fast vehicles illustrates benefits approach.