Multi-objective optimization of comprehensive performance enhancement for proton exchange membrane fuel cell based on machine learning

The comprehensive performance of proton exchange membrane fuel cells depends on operating conditions. This paper innovatively uses the Pearson correlation coefficient to screen the optimization objectives (uniformity index of oxygen, standard deviation of temperature, net power density), and obtains the optimal operating conditions of the proton exchange membrane fuel cell through a multi-objective optimization method. The optimized dataset comes from the simulation results of the three-dimensional numerical model, and the regression model is established through the response surface method. Moreover, the non-dominated sorting genetic algorithm-II is used for processing to obtain the Pareto front solution set, and the optimal operating conditions are obtained from it through the Technique for order preference by similarity to an ideal solution. The analysis of variance result shows that the influence of cathode operating conditions on the comprehensive performance is greater than that of anode, especially the influence of cathode stoichiometry ratio is the most significant. The optimal solution obtained 1.0981 %, 10.5845 %, and 1.0376 % enhancement compared to the optimal values in the simulation results. The differences between the three optimization objectives are only 0.8190 %, 1.0315 %, and 0.8789 % as verified by numerical simulation, thus the machine learning results are reliable and accurate.