Enabling low-carbon membrane steam methane reforming: Comparative analysis and multi-objective NSGA-II-integrated Bayesian optimization

This study comparatively investigated and optimized membrane steam methane reforming (M-SMR) over a Ni/ Al2O3 catalyst in a rectangular channel using a Pd-based membrane. Based on Le Chatelier's principle, the latter promotes hydrogen production via SMR and the water-gas shift reaction (WGS) owing to the shift in reaction equilibrium; however, with a tradeoff of an increase in CO2 emission. Herein, the wash-coated (i.e., the WCM- SMRs) and packed-bed (i.e., the PB-M-SMR) catalytic reactors were modeled and compared. A multiobjective Bayesian optimization was applied to determine the optimal membrane distribution that would result in a high hydrogen recovery relative to the membrane length ratio along with low CO2 emission. According to the optimization results, the CO2 selectivities of the WC-M-SMR and the PB-M-SMR were significantly lowered up to 16.38% and 35.7%, respectively, compared to the base configurations associated with full membrane coverage. Additionally, a high separation ratio was attained despite the shorter membrane distance. This methodology can be successfully used for process control to lower CO2 emissions. It also enables key downstream processes such as methanol fuel production and Fischer-Tropsch synthesis by increasing the CO yield-which is low in conventional M-SMR reactors-while maintaining a high hydrogen yield, similar to a fully installed membrane.