Optimal Parameter Estimation of PEMFC Model Using an Improved Atomic Orbital Search Algorithm

Proton exchange membrane fuel cells (PEMFCs) have drawn much attention lately for their parameter extraction. It is important to carefully determine the optimal values of the uncertain parameters in the PEMFC model to guarantee accuracy and dependability. However, because of their nonlinearity and multi-variability, PEMFC modeling and optimization present a significant difficulty. Therefore, an improved atomic orbital search algorithm based on Lévy flight and chaotic maps, called CLAOS, was proposed in this study for the PEMFC parameter estimation problem, where the sum of square error was minimized. In order to test the effect of the Levy flight and chaotic maps on the performance of the Atomic Orbital Search (AOS) algorithm, ten different AOS variations were created and applied to solve the CEC2020 and CEC2022 benchmark problem suites. Their results were analyzed using Friedman and Wilcoxon tests, and the best variant was called the CLAOS algorithm. To validate the effectiveness of the proposed CLAOS, extensive simulations and performance evaluations were conducted on the PEMFCs model, where a 250W PEMFC stack was considered. Two search ranges for the unknown parameters and two operational conditions were considered. The performance of the proposed algorithm was compared with the six meta-heuristic search algorithms. Accordingly, the proposed algorithm achieved 4.722489 and 0.152027 for Case-1 and Case-2, respectively, which were the best objective function values among its rivals. Moreover, the results of the CLAOS algorithm were compared with the results reported in the literature for both cases. Accordingly, the CLAOS achieved the minimum error value compared to its rivals for both cases. To evaluate the performance of the algorithms statistically, the Friedman and Wilcoxon tests were applied to the results of the algorithms. The Friedman test results show that the proposed CLAOS algorithm ranked first with 1.1667 and 1.0000 score values for Case-1 and Case-2, respectively. All simulation and analysis results demonstrated that the proposed algorithm outperformed its rivals in solving the PEMFC parameter estimation problem.