Adaptive optimization of combined steam and CO₂ reforming for hydrogen production from variable biogas feed

Biogas-to-hydrogen (B2H2) conversion offers a sustainable pathway for valorizing organic waste-derived biogas from anaerobic digestion (AD) into clean energy carriers. However, temporal variations in biogas composition strongly influence the endothermic nature of direct reforming reactions, burner fuel quality, and overall process stability. This study systematically investigates the effect of biogas variability on the B2H2 process through a comprehensive parametric analysis, based on a techno-economic-environmental analysis, evaluating three key indicators: unit hydrogen production cost (UPC), net CO₂ emissions (NCE), and hydrogen production efficiency (EFF). Then, a novel multi-task digital twin optimization (MT-DTO) framework is developed for feed-adaptive operation of the B2H2 process based on combined steam and CO₂ reforming (CSCR), enabling efficient biohydrogen production under dynamically changing biogas conditions. The MT-DTO framework integrates a deep learning surrogate model, a multi-task evolutionary optimization algorithm, and a K-nearest neighbors classifier to simultaneously optimize multiple operational objectives under varying biogas compositions. Results indicate that increasing methane content from 55% to 75% reduces UPC from 3.14 to 2.66 USD/kg H₂ and improves EFF from 0.33 to 0.36. The operational temperature of CSCR, critical for endothermic hydrogen production, exerts significant influence on UPC, NCE, and EFF. Validation using real biogas data from a full-scale AD plant in Korea confirms annual operating cost reductions from 56.06 to 53.25 million USD and a 6% decrease in CO₂ emissions, from 47.18 to 44.53 kilotons. Therefore, the proposed MT-DTO framework can provide practical guidance for designing and operating waste-to-hydrogen systems efficiently and sustainably under variable biogas feed conditions.