Genetic algorithms in model structure identification for fuel cell polarization curve

Abstract Evolutionary optimizers, such as genetic algorithms, have earlier been successfully applied to find the parameter values for the fuel cell polarization curve models. The structure of these, typically semi-empirical, models have evolved during the decades. In this study, the model structures were reviewed and a new model structure was generated. Genetic algorithms were used to determine the optimized model structure with linear model parameters. Four different fuel cells, one with varying operating conditions, were studied. The results show that the model can outperform the semi-empirical model utilized in number of studies without increasing the model complexity