Reduced-Order Parameter Identification of a Polymer Electrolyte Membrane Fuel Cell Using a DE-Enhanced BO Variant for PEMFC Fitting

Abstract

Polymer electrolyte membrane fuel cells (PEMFCs) are a promising technology for sustainable energy systems because of their high efficiency, low emissions, and quiet operation. Accurate parameter identification is essential for reliable modeling and performance optimization of PEMFC systems under varying operating conditions. This study proposes the mutated bonobo optimizer (Mu-BO), a hybrid optimization method that incorporates a Differential Evolution (DE) mutation operator into the bonobo optimizer (BO). Mu-BO is used to identify the parameters of semi-empirical PEMFC models while improving search robustness and reducing sensitivity to parameter settings. The method is evaluated using datasets from three PEMFC systems: a 250 W cell, an SR-12 500 W stack, and a NedStack PS6 6 kW stack. Its performance is compared with contemporary state-of-the-art optimization methods and the original BO algorithm. The results show that Mu-BO achieves higher accuracy, faster convergence, and more consistent performance across all datasets. A sensitivity analysis investigates the effect of parameter variations on the sum of squared errors (SSE), providing further insight into model performance. In addition, a reduced-order PEMFC model is developed to improve computational efficiency, reducing computation time by 12.17%, 15.32%, and 29.01% for the 250 W, SR-12, and NedStack PS6 systems, respectively, without compromising accuracy.