A Novel Dual Fitness–Distance Balance Chaotic Evolution Algorithm for Parameter Extraction of PEM Fuel Cell Models

ABSTRACT Accurate parameter estimation of proton exchange membrane fuel cells (PEMFCs) is a critical and highly complex optimisation problem due to the nonlinear, multivariable, and multimodal nature of fuel cell polarisation curves. To overcome problems in existing metaheuristic algorithms in the literature, such as the early convergence issue and the inability to properly maintain the balance between exploration and exploitation, this study proposes an effective, successful, and novel optimisation algorithm called dual fitness–distance balance chaotic evolution optimisation (DFDBCEO). The proposed method significantly enhances the standard chaotic evolution optimisation algorithm by integrating a newly developed “Dual FDB” architecture. This unique selection mechanism explicitly evaluates both the global distance to the best‐known solution and the local spatial distance among neighbouring candidates, effectively preserving population diversity and preventing local clustering. To comprehensively validate the robustness and accuracy of the proposed algorithm, rigorous experiments were conducted on BCS‐500 W, Ballard Mark V, NedStack PS6, and Horizon H‐12 stack systems, which are widely used commercial PEMFC stacks in the literature. The performance of the proposed algorithm was evaluated using four different objective functions (sum of squared error, integral of the absolute error, root mean square error, and mean absolute error) across 30 independent runs. The comprehensive statistical results demonstrate that DFDBCEO consistently achieves superior solutions compared to competing algorithms with near‐zero variance and virtually 100% convergence efficiency. Furthermore, non‐parametric statistical evaluations, including the Friedman rank test and the Wilcoxon signed‐rank test, confirmed that DFDBCEO systematically outperforms 15 state‐of‐the‐art metaheuristic algorithms. The proposed DFDBCEO algorithm, which consistently achieved the best average ranking in all 16 experimental cases, has proven to be a highly reliable, robust, and accurate tool for complex PEMFC parameter extraction.