A two-stage GIS-MCDM optimization framework for urban demand-adaptive siting and sizing of hybrid renewable hydrogen facilities

초록

Urban decarbonization requires hydrogen systems capable of converting surplus solar and wind electricity into a storable energy carrier for flexible local supply within energy networks. However, the performance of green hydrogen infrastructure is highly location-dependent, yet many existing siting approaches rely on static resource assessments or annualized meteorological data, which limits their ability to represent temporal variability under real urban conditions. This study proposes a two-stage geographic information system-multi-criteria decision making (GIS-MCDM) framework for urban demand-driven optimal siting (DDOS) and sizing of hybrid renewable hydrogen facilities. First, a generative artificial intelligence-based spatial autoencoder (GAI-SAE) generates location-specific annual meteorological time series under uncertainty, and an analytic hierarchy process (AHP) framework with constraint mapping identifies feasible zones for photovoltaic-wind-hydrogen deployment. The first-stage uses the DDOS scheme to determine demand-responsive facility locations, and the second-stage applies a TRNSYS-Python based many-objective optimization for site-specific hybrid renewable energy system (HRES) sizing with a dynamic techno-economic-environmental assessment. The proposed GAI-SAE achieved an R2 of 84.9% for global horizontal irradiance (GHI) forecasting on unseen data at a 30 min horizon. Suitability analysis identified 164 km2 of highly suitable land for HRES development. The DDOS stage selected 20 sites with full demand coverage and demand-weighted service distance of 3.41 km, which indicates efficient urban energy distribution. At the sizing stage, most locations achieved self-sufficiency above 95% during high-load periods. Overall, the DDOS framework provides a practical approach to sustainable regional energy planning and lowcarbon urban transitions.

키워드