A novel sensor preference method for proton exchange membrane fuel cell flooding fault diagnosis based on multi-algorithm.

During periods of high-power operation, proton exchange membrane fuel cell (PEMFC) produce water, which gradually accumulates at the cathode. It causes flooding faults and significantly impacts on PEMFC safety. The data collected by the sensors is a fundamental prerequisite for the flooding faults diagnosis in PEMFC. Consequently, it is critical to select sensors. However, the conventional methods for selecting sensors consider sensitivity and noise immunity, the number of sensors selected and the flood diagnostic performance are not optimized. To solve the above problems, this paper proposes a sensor preference method that combines the Mantel test and Spearman coefficient (MS method). This method first uses the Mantel test to analyze the correlation between two flooding evaluation indicators and the sensors, it identifies suitable sensors for diagnosing flooding faults. Subsequently, the Spearman coefficient eliminates redundant sensors. Additionally, a one-dimensional convolutional neural network is employed to construct a flooding diagnosis model. The MS method is contrasted with two traditional sensor selection methods. The results show that the MS method selected sensors yield superior model training and enhanced accuracy in diagnosing flooding, thereby ensuring PEMFC safety. [ABSTRACT FROM AUTHOR]