The stable operation of a renewable energy-based green hydrogen supply chain is emerging as a critical challenge for achieving carbon neutrality and expanding the hydrogen economy. However, relying solely on historical meteorological data fails to ade...
The stable operation of a renewable energy-based green hydrogen supply chain is emerging as a critical challenge for achieving carbon neutrality and expanding the hydrogen economy. However, relying solely on historical meteorological data fails to adequately reflect the increasing frequency and intensity of extreme weather events, such as heatwaves and cold waves, induced by climate change. Existing deterministic optimization models often struggle to ensure system
reliability under these unprecedented environmental uncertainties.
To address this gap, this study proposes a novel analytical framework that integrates a generative AI-based weather scenario generator with a Mixed-Integer Linear Programming (MILP) optimization model. First, a Conditional Variational Autoencoder (cVAE) based on 1D-CNN architecture was developed to synthesize physically consistent meteorological time series data. By incorporating FiLM-based modulation and extreme value aware loss weighting, the model successfully generated realistic extreme climate scenarios (heatwaves and cold waves) that are difficult to capture with observational data alone.
These generated scenarios were then integrated into a hydrogen supply chain model consisting of a hybrid renewable energy system(9 MW PV and 6 MW Wind), 2.5 MW PEM water electrolysis, storage, and transportation sectors. The MILP model optimized the system design and operation to minimize the total cost while satisfying demand constraints. The simulation results demonstrated that the system maintained a 100% supply sufficiency rate across all climate
conditions, validating its operational resilience. Comparative analysis revealed that heatwaves slightly reduced hydrogen production
efficiency due to the negative temperature coefficient of PV modules, whereas cold waves compensated for solar losses through increased
wind power generation, highlighting the complementarity of the hybrid energy mix.
Economically, the Levelized Cost of Hydrogen (LCOH) was derived at approximately 19,847 KRW/kg under normal conditions. The analysis quantified the Climate Risk Premium, showing that extreme weather intensity incurs a marginal cost increase of 0.76%. However, the study also confirmed that grid-connected surplus power sales effectively buffer this economic impact, stabilizing the Net LCOH. This research holds significant value as it provides a robust decision-making framework for designing climate resilient hydrogen energy systems by effectively combining AI-driven data augmentation with operations research.