Optimizing proton exchange membrane fuel cell parameters using adaptive cluster division differential evolution

Proton exchange membrane fuel cells (PEMFCs) are complex systems with many interconnected nonlinear components. To create accurate models of these systems, it is necessary to precisely identify the parameters that govern their behaviour. Metaheuristic algorithms are ideal for this task, as they systematically explore the solution space to find the best-fitting parameter values. This study adopts the Adaptive Cluster Division Differential Evolution (ACD-DE) algorithm and compares it with nine other Differential Evolution (DE) variants, including DE, iLSHADE, CRADE, LSHADE, jSO, HARD-DE, LSHADE-cnEpSin, PCM-DE, and CS-DE, for calibrating the PEMFC model. The process starts by using these algorithms to fine-tune the parameters of a standard PEMFC model. The goal is to make the model's predictions as accurate as possible by reducing the difference between its calculated voltages and the actual voltages measured from six commercial PEMFCs: BCS 500 W, STD 250 W, Nedstack, SR-12, H-12, and HORIZON 500W. This is achieved by minimizing the sum of squared errors between the predicted and actual voltages. After executing 30 independent trials, each consisting of 500 iterations, the algorithms are evaluated based on their lowest and highest SSEs, as well as their average and standard deviation. The results show that ACD-DE slightly surpasses the other nine DE variants in achieving the lowest SSE in all 12 scenarios examined. In addition, the current-voltage (I/V) and power-voltage (P/V) curves produced by the ACD-DE method are very similar to the curves provided in the datasheets for all of the cases studied.