A Novel Degradation Prediction Method of PEMFC Based on CEEMDAN and Informer-LSTM Parallel Prediction

Abstract

Proton exchange membrane fuel cells (PEMFC), as an important part of clean energy technology, are widely used in transport, portable power sources and stationary power systems. PEMFC experience aging during use, resulting in degradation of their performance and shorter lifespan. In this paper, a hybrid model of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), informer, and long short-term memory (LSTM) is proposed to predict the aging trend. The data are decomposed into multiple Intrinsic Mode Function (IMF) through CEEMDAN, which are reconstructed according to sample entropy (SE) to provide stable data for the model. A new prediction approach is proposed to predict informer and LSTM in parallel while extracting multifaceted features. Different datasets, different training stopping points (TSP), and multiple models are used to validate the accuracy and stability of the model. The root mean square error (RMSE) and mean absolute error (MAE) can reach 0.00137 and 0.00060 for the steady state dataset, and the prediction is better for the quasi–dynamic dataset with RMSE and MAE reaching 0.00126 and 0.00065.