Machine learning based predictive model for methanol steam reforming with technical, environmental, and economic perspectives

Abstract

To overcome limitations of conventional H2 production approaches such as steam methane reforming (SMR) in a membrane reactor (MR) such as large CO2 emission and deactivation of catalyst and membrane, a promising alternative H2 production system of methanol steam reforming (MSR) in serial reactors and membrane filters is reported here, affording its high product yield and a compact design. In this study, technical, environmental, and economic feasibility according to 12 techno-economic parameters and detailed effects of each parameter for this H2 production system are comprehensively investigated with a machine learning (ML) based predictive model in the following steps: (1) process simulation using Aspen Plus® with detailed thermodynamic phenomena and environmental performance; (2) numerical model using MATLAB® based on technical and environmental performance from the process simulation results; (3) ML-based predictive model having outputs of H2 production rate, CO2 emission, and unit H2 production cost feasibility trained by 12,000 data sets from a numerical model. It is well noted from this study that # of reactors and operating temperature for technical performance, # of reactors and S/C ratio for environmental performance, and operating temperature, # of reactors, reactant, and labor for economic performance are reported as most influential factors.