Long-term prognostic framework for PEMFC under dynamic conditions based on EMD-Transformer

In transportation applications, the commercialization process of proton exchange membrane fuel cells is challenged by their limited durability. Predictive and Health Management (PHM) techniques are able to extend the system lifespan by predicting the system health status and taking decisions to avoid failures. A deep learning algorithm combined with modal decomposition is developed to predict the performance degradation of proton exchange membrane fuel cells. Specifically, the degradation trajectory is divided into a main degradation trend part and a periodic degradation-recovery part by a empirical modal decomposition. Experimental results show that the framework achieves an average long-term prediction error below 0.69%, outperforming traditional methods such as ARIMA and LSTM in both performance and stability while reducing error accumulation and overestimation risks. It is proved that the proposed prediction method can help to realize high-precision health state prediction.