Data-driven flooding fault diagnosis method for proton-exchange membrane fuel cells using deep learning technologies

Abstract

Effective and accurate diagnostic methods are necessary to ensure the stable and efficient operation of proton exchange membrane fuel cells (PEMFC). In practice, the voltage drop is commonly used as the detection indicator of the flooding fault. However, the output power changes also affect the voltage of the PEMFC. To better diagnose the flooding fault under load-varying conditions, a data-driven method for PEMFC using deep learning technologies is proposed, which can automatically extract fault features for the raw data to diagnose the flooding fault. First, the indicators for the fault diagnosis model are selected to meet the actual situation according to the water transport mechanism and auxiliary systems of the general fuel cell stack. And the collected data are transformed into a 2-D graph to visually represent the characteristics of the time-series data. Then, the convolutional neural network is adopted to develop the fault diagnosis model. In addition, the batch normalization method is used to alleviate feature distribution differences and enhance the model generalization. Finally, the trained model is applied to detect the flooding fault. A real PEMFC experiment dataset is adopted to verify the diagnostic performance of the method. Experiment results show that the proposed model can effectively identify the flooding fault of the fuel cell accurately under load-varying conditions, and achieves over 99% accuracy.