Model predictive control of three-level grid-connected inverters for weak power grids

Abstract

In weak power grids， grid-connected inverters often suffer from poor robustness and significant harmonics. Moreover， model predictive control （MPC） heavily relies on system parameters， limiting its applicability in weak grids. To address these issues， a data-driven model-free predictive control （MFPC） strategy base on virtual voltage vectors for neutral-point clamped （NPC） three-level grid-connected inverters is proposed. Firstly， the MFPC based on virtual voltage vectors constructs numerous virtual voltage vectors through different switching sequences. Then， a cost function is utilized to pre-select and reduce the computational burden of the control algorithm. Finally， appropriate switching sequences are chosen for optimization to suppress neutral-point voltage fluctuations. Simulation results demonstrate that the proposed strategy effectively mitigates neutral-point voltage fluctuations and grid current harmonics， while exhibiting robustness to parameter variations.

Keywords

• weak grid
• npc three-level inverter
• data-driven model predictive control
• robustness