Optimizing PEMFC parameter identification using improved pufferfish algorithm and CNN

In this research, a novel approach has been proposed for enhancing the accuracy of proton exchange membrane fuel cell (PEMFC) models based on convolutional neural networks (CNNs) and an improved optimization method (called the improved pufferfish optimization algorithm). This method has been suggested to improve the predictive performance of the model by optimizing the estimation of PEMFC parameters. The results with almost 2.5-unit training error for the PEMFC parameter using sample data during training provide a promising output for the system. These results confirm the effectiveness of the proposed method with acceptable errors and good performance on the identification of parameters for PEMFCs over diverse circumstances. The yield voltage–current curves predicted by the optimized CNN model display a high degree of correlation with the empirical voltage–current curves, clearly proving the accuracy of the achieved CNN model and its ability to learn the underlying relationship, thus helping to improve the PEMFC parameter estimation.