Transformer-based prediction of the RUL of PEMFC

Abstract

Recently, the proton exchange membrane fuel cell (PEMFC) is of increasing interest to researchers and is considered to have a wide range of applications, because of its low pollution and high energy density. Remaining Useful Life (RUL) prediction is a major problem in driving the widespread use of PEMFC. This paper presents a transformer-based algorithm for RUL. The first step in this algorithm is to extract the periodicity and non-periodicity of the time series using time2vec. Then, the algorithm adds a convolutional network to the transformer to extract the temporal correlation and spatial correlation of the input time series. Moreover, we combine the handcrafted features with automatically learned features to boost the performance of the RUL prediction. The algorithm uses operational data from actual PEMFC vehicles for comparison experiments, and the prediction performance of our proposed algorithm outperforms prediction results of other algorithms.