Smart reforming for hydrogen production via machine learning

Abstract

The integration of diesel autothermal reforming and solid oxide fuel cells holds a key position in hydrogen-electricity co-generation, where optimization of the reforming process is essential for generating high hydrogen content for fuel cells while maintaining thermal neutrality. Aspen Plus, despite its common usage, suffers from the inability to reflect the interaction effects between parameters, which machine learning can enhance it. In this study, twelve machine learning models have been developed using data from thermodynamic calculations and then screened with Extreme Gradient Boosting for analysis, prediction, and optimization of hydrogen production. Machine learning provides detailed explanations of the impacts of individual and multiple factors, demonstrating powerful optimization capabilities through predictive evaluation and fit with experimental data. This method highlights the potential of integrating machine learning with Aspen technology for optimizing chemical processes.