Optimal design of hydrogen storage-based microgrid employing machine learning models

The integration of hydrogen (H₂) into renewable energy-based microgrids enables long-term energy storage, prolongs battery (BT) life, minimizes energy costs, and improves microgrid reliability. However, literature indicates several research gaps in the design of H₂-based microgrids. In existing techno-economic optimization of H₂-based microgrids, fixed electrolyzer (EL), BT, and fuel cell (FC) lifespans predetermined by manufacturers were used, and most of the studies rely on constant EL efficiency. In response to the identified gap, this paper proposes optimal design of a standalone H₂-based microgrid design incorporating a wind turbine (WT), photovoltaic (PV) system, EL, FC, and BT. Using single-cell electrochemical models as a foundation, the study derives EL efficiency curves and calculates BT, EL, and FC lifespans based on operational cycles rather than manufacturer estimates. We evaluate eight different microgrid configurations employing proton exchange membrane and alkaline EL (ALE), as well as vanadium-redox-flow-BT (VRFB) and lithium-ion BT to determine the most cost-effective microgrid for an internally displaced person's camp. Furthermore, following multi-objective optimization of the microgrid, machine learning (ML) algorithms were employed to analyze the techno-economic data of the various microgrid scenarios. Grasshopper optimization algorithm was used to optimize the microgrids to determine the optimal sizes of the microgrid components that minimize the levelized cost of energy (LCOE) at zero loss of power supply probability (LPSP). The H₂-based microgrid incorporating PV, WT, VRFB, FC, and ALE with LCOE of USD 0.508/kWh was found to be more cost-effective compared to diesel generator system. The study shows that extreme gradient boosting ML algorithm yields excellent accuracy in predicting the LCOE and LPSP compared to random forest ML model. Additionally, a comprehensive payback period is performed to provide insights for stakeholders into the financial performances of the microgrid configurations for preliminary screening, risk assessment, and liquidity analysis.