A novel in-tube reformer for solid oxide fuel cell for performance improvement and efficient thermal management: A numerical study based on artificial neural network and genetic algorithm

The pursuit of higher power density and compact structure presents a critical challenge to the thermal management of solid oxide fuel cell. In this study, a novel in-tube reformer is proposed and a Multi-physics simulation-Artificial neural network-Multi-objective genetic algorism based optimization framework is developed to improve the output performance and reduce the internal temperature difference in solid oxide fuel cell. First, a validated multi-physics model is developed for parametric simulation and generating dataset. Afterwards, a surrogate model is obtained by training an artificial neural network to predict the output performance and internal temperature field of solid oxide fuel cell. Finally, multi-objective genetic algorithm optimizations based on the surrogate model are performed to maximize the output performance and minimize the internal temperature difference under different operation strategies. It is found that compared to the conventional configuration (without in-tube reformer), the use of in-tube reformer can effectively promote the electrochemical reactions, increase the fuel utilization (up to 34.2%) and current density (up to 14.5%) while significantly reducing the maximum temperature difference (up to 85.5%) in the cell, resulting in a uniform current density and temperature distribution along the cell. The proposed novel in-tube reformer and optimization framework are demonstrated to be highly powerful and can be easily applied to other fuel cell/electrolyzer systems to effectively improve system performance and realize efficient thermal management under actual demands.