ScholarWorks@성균관대학교

초록

Hydrogen production from biogas is strongly affected by upstream feedstock variability and market fluctuations, creating challenges for efficient and sustainable production planning. This study proposes a digital scheduling framework for hydrogen production via combined steam and CO2 reforming (CSCR), integrating feed forecasting, surrogate modeling, and rolling horizon optimization to enable adaptive operation and support future operational planning under variable conditions. First, detailed process simulations of the integrated CSCR system were conducted to generate technical, economic, and environmental performance data over a wide range of operating conditions. A deep neural network–based surrogate model was then developed to represent the nonlinear behavior of the process with low computational cost. Forecasts of feed biogas CH4 concentration were generated using a gated recurrent unit (GRU) model over a seven-day horizon. Rolling horizon (RH) optimization was employed in a multi-step scheduling scheme to determine optimal daily operating parameters within predefined bounds, aiming to meet hydrogen demand while adapting to predicted feed variability and market fluctuations. The results showed that the GRU model provided accurate forecasts (R2 = 0.9411), enabling advanced estimation of feed biogas composition. Informed by these forecasts, the RH approach allowed flexible adjustment of the CSCR operating parameters, leading to increased hydrogen production, reductions in net CO2-equivalent emissions of up to 1470 kgCO2/day, and improvements in cumulative profit of up to 47% compared with fixed scheduling over the seven-day horizon. Overall, the proposed scheduling approach effectively coordinates production with feedstock and operational fluctuations, enhancing adaptability and supporting sustainable hydrogen production planning.

키워드