Design optimization and parameter estimation of a PEMFC using nature-inspired algorithms

Numerical simulation of proton-exchange membrane fuel cells (PEMFCs) requires an adequate model and precise parameters for reproducing their operational performance quantified by the polarization curve. Bioinspired algorithms are well-suited for optimization. The simulator is stressed by inputting thousands of randomly generated parameters, and hence, a robust numerical model is required. Once the proper model and parameters reproduce the experimental data, they can be used for design improvement. This article proposes a reformulation of a macrohomogeneous mathematical model to provide higher numerical stability to the solutions. We introduce optimization problems for parameter estimation and design optimization by applying three bioinspired algorithms to maximize its performance and minimize the platinum mass loading mPt\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m_\mathrm{{Pt}}$$\end{document}. The results are validated by comparing the experimental polarization curves with those simulated from the estimated parameters. We compare a base design’s performance with the optimized design for maximum performance. We also compare a base design with the optimized design for minimum mPt\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m_\mathrm{{Pt}}$$\end{document}. The results show that the particle swarm optimization requires the lowest computational cost and performs the best in most cases, fitting the experimental data with errors lesser than 10-17\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10^{-17}$$\end{document}. The minimization of mPt\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m_\mathrm{{Pt}}$$\end{document} reduces the amount by 42% compared to the base case.