DL-Predictive Control of GFM Inverters for Inrush Current Mitigation and Overload Protection

Inrush current and overload conditions in grid-forming inverters can shorten inverter lifespan and trigger microgrid blackouts due to associated thermal and electrical stresses. This paper proposes a deep learning (DL)-based predictive control strategy to mitigate these effects by minimizing energy dissipation during inrush and steady-state overload conditions. A neural network model, enhanced with feedback from prediction errors, forecasts the inverter's current-voltage dynamics. The control framework optimizes a cost function that balances voltage regulation and overload protection, with tuning parameters aligned to system constraints and IEEE 1547 standards. By maintaining the point of common coupling (PCC) voltage within acceptable limits, the proposed method improves inverter reliability. Simulation results show superior performance compared to conventional approaches, including feedforward control, under both transient and steady-state scenarios.