Exploring Performance Degradation of Proton Exchange Membrane Fuel Cells Based on Diffusion Transformer Model

Proton exchange membrane fuel cells (PEMFCs) stand at the forefront of energy conversion technology, efficiently converting the chemical energy of hydrogen and oxygen directly into electricity. Research on predicting the remaining useful life of PEMFCs has long been a focus, as it plays a crucial role in preventing failures and mitigating safety risks. This paper introduces a robust diffusion transformer (DiT) model, which is a novel approach leveraging generative artificial intelligence (GAI) technology to innovate the existing methods for predicting the performance degradation of PEMFCs. This model employs random Gaussian noise to generate stable performance degradation data of PEMFCs under specified conditions. The predictive accuracy is then assessed by benchmarking against a bi-directional long short-term memory recurrent neural network (Bi-LSTM) using two distinct experimental datasets. The evaluation shows that the DiT model achieves higher predictive accuracy than the reference model. Specifically, the mean absolute prediction error is reduced by 72.7% under steady-state conditions and 59.3% under dynamic conditions. Correspondingly, the remaining useful life error (RE) is diminished by 80% and 88%, respectively. These findings indicate that the DiT model has significant potential in PEMFCs performance degradation research.