Accurate semi-empirical parameter identification of polymer electrolyte membrane fuel cells via enzyme action optimization

A precise semi-empirical model of polymer electrolyte membrane fuel cells (PEMFCs) is vital for performance prediction and control, making accurate parameter identification a crucial task. Many metaheuristic techniques suffer from slow convergence and frequent entrapment in local minima due to the high nonlinearity and complexity of PEMFC models. This study introduces the Enzyme Action Optimizer (EAO) for robust and accurate estimation of uncertain PEMFC parameters. By refining the exploration–exploitation dynamics, EAO improves both convergence speed and accuracy. The EAO was applied to four standard stacks: 250 W, BCS 500 W, AVISTA SR-12 500 W, and Temasek 1 kW, using the sum of squared errors (SSE) between experimental and simulated voltages as the objective function. Minimum SSE values of 0.6420, 0.0116, 1.0566, and 0.7910, respectively, were obtained, showing close agreement with experimental polarization curves. Comparative analyses with five recent optimizers: dhole optimization algorithm (DOA), equilibrium optimizer (EO), white shark optimizer (WSO), snake optimizer (SO), and puma optimizer (PO), at which statistical analyses and convergence profiles confirmed the EAO’s superiority. Findings also revealed that EAO achieved the lowest mean Friedman rank, standard deviation, and fast convergence within 25–30 iterations across all cases. The algorithm demonstrated superior stability and reduced run-to-run variability. A ±10 % sensitivity analysis further confirmed the dominance of the activation-kinetics coefficients in influencing model accuracy. Overall, EAO provides a fast, consistent, and accurate approach for PEMFC parameter estimation, enabling improved modeling, control, and design optimization.