Identifying the unknown parameters of PEM fuel cells based on a human-inspired optimization algorithm

Abstract

With the increasing reliance on hydrogen as an energy carrier, significant advancements are being made in the evolving energy sector. Proton exchange membrane fuel cells (PEMFCs) play a central role in this transformation, offering an efficient and sustainable alternative to fossil fuels. Accurate modeling of PEMFCs is essential for performance analysis and optimization. This study proposes a novel approach to estimate seven unspecified PEMFC parameters using the Adolescent Identity Search Algorithm (AISA), a human-inspired optimization technique. The AISA algorithm is applied to minimize the sum of squared errors (SSE) between experimental and predicted voltage values for three commercial PEMFC models: the Horizon 500W Stack, the BCS500W Stack, and the NedStack PS6 Stack. The proposed approach achieves a minimum SSE of $7.6376\times {10}^{-3}$, $1.2869\times {10}^{-2}$, and $2.2881$ for the three respective models. Further validation with four additional PEMFC datasets (Ballard Mark, H12-3, SR-12, and STD-4 stacks) confirms AISA's exceptional performance, achieving minimum SSE values of 9.3575E-01, 6.1870E-02, 1.0532E+00, and 2.0453E-01, respectively, with remarkable stability. Statistical validation through the Wilcoxon signed-rank test shows AISA outperforms comparison algorithms in 14 out of 18 pairwise comparisons, with no instances of being outperformed. Friedman test rankings position AISA as the top-performing algorithm across all case studies, with mean ranks of 1.17, 2.13, and 2.20, respectively. Comparative analysis with state-of-the-art metaheuristic algorithms—including the Gradient-Based Optimizer (GBO), Peafowl Optimization Algorithm (POA), and Honey Badger Algorithm (HBA)—confirms AISA's superior accuracy, stability, and computational efficiency, achieving runtime values as low as 0.5766 s. Furthermore, AISA exhibits superior convergence behavior, reaching optimal solutions with fewer function evaluations. These results highlight AISA's potential as an effective and computationally efficient tool for PEMFC parameter identification, fuel cell modeling, and performance optimization. Future research will focus on expanding the analysis to additional PEMFC models and exploring hybrid optimization strategies to further enhance accuracy and robustness.