Model Predictive Control to Promote PV Intake in a Multi-Energy Based Power Grid

With the growing integration of renewable energy into power grids, photovoltaic systems face challenges due to intermittency, resulting in significant PV curtailment under low-load conditions. While hydrogen energy storage coupled with water electrolysis and fuel cells offers a promising solution to balance supply-demand fluctuations. Existing control strategies for multi-energy systems often rely on open-loop frameworks that lack real-time adaptability. As a result, sub-optimal PV utilization and grid unstable may happens. This paper proposes an innovative closed-loop model predictive control (MPC) scheme enhanced by hybrid artificial intelligence to maximize PV intake and minimize grid-user power exchange in a PV-hydrogen-microgrid system. The MPC mechanism uniquely incorporates feedback correction and a hybrid prediction algorithm (EMD-KPCA-LSTM) to address PV uncertainty: empirical mode decomposition (EMD) extracts intrinsic mode functions from environmental and PV data, kernel principal component analysis (KPCA) reduces feature dimensions, and long short-term memory (LSTM) networks enable high-accuracy PV power forecasting. Iterative genetic algorithm (GA) optimization generates optimal control sequences for 24-h operation. Case studies using real-world data from China demonstrate that the proposed scheme reduces grid interaction by 42.6% and increases PV utilization by 13.7% in average compared to baseline methods. Key innovations include (1) a feedback-corrected MPC framework for dynamic adaptation to PV and load fluctuations, (2) the EMD-KPCA-LSTM hybrid model enhancing prediction accuracy, and (3) an SSM constrained optimization strategy specifically targeting PV intake maximization rather than conventional cost minimization. This work provides a scalable solution for renewable-rich grids transitioning to low-carbon multi-energy systems.