Prediction and optimization of gas distribution quality for high-temperature PEMFC based on data-driven surrogate model

Suitable operating conditions can improve the internal gas distribution of high-temperature proton exchange membrane fuel cell (HT-PEMFC), reduce the occurrence of local gas starvation, and thus prolong the lifespan. Therefore, it is important to investigate the gas distribution quality of HT-PEMFC based on quantitative indicators. In this paper, data-driven surrogate models are established based on the simulation results of a three-dimensional validated numerical model to study the gas distribution quality of HT-PEMFC via two qualitative evaluation indexes, the value of mean and standard deviation of the reactant gas concentration in catalyst layer. In order to obtain the surrogate model more efficiently and accurately, genetic algorithm optimized deep belief network (GA-DBN) and Autogluon (an automated machine learning method) are applied. Based on the surrogate models, the influence of the operating condition parameters on each evaluation index is revealed, which provides theoretical basis for the life-extension design of HT-PEMFC. Then, Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS), a Multi-Criteria Decision Aid (MCDA) method, is used to select the optimal operating conditions according to the two evaluation indexes of gas distribution quality from the extensive computational results of the surrogate models, which provides specific guidance for the life duration improvement of HT-PEMFC.