Multi-objective optimization of PEM fuel cell by coupled significant variables recognition, surrogate models and a multi-objective genetic algorithm

This paper aims to present a fast and systematic optimization approach for proton exchange membrane fuel cell (PEMFC) by combining variance analysis, surrogate models and non-dominated sorting genetic algorithm (NSGA-II). First, a three-dimensional steady-state PEMFC computational fluid dynamics (CFD) model is developed as the base model for optimization. Second, six variables that have significant effect on PEMFC performance are selected from numerious common parameters using variance analysis, reducing the number of decision variables from 11 to 6. Then, three data-driven ensemble learning models are trained as surrogate models to accelerate the fitness values evaluation of the optimization algorithm. Finally, three PEMFC performance indexes, including power density, system efficiency and oxygen distribution uniformity on cathode catalyst layer are optimized simultaneously based on NSGA-II. Using the NSGA-II combined with surrogate models, a set of Pareto solutions is obtained in a short time. The results indicate that PEMFCs with optimized parameters perform better than the base model in terms of all three performance indexes, demonstrating the success of this approach in solving time-consuming multi-optimization problems. This study provides a fast and systematic approach for PEMFC multi-objective optimization and can be a guide for engineering applications.

Automated summary

This paper presents a fast and systematic optimization approach for PEMFC by combining variance analysis, surrogate models, and NSGA-II. By optimizing power density, system efficiency, and oxygen distribution uniformity simultaneously, the study demonstrates the success of this method in solving time-consuming multi-optimization problems efficiently.