Characteristic Frequency Screening of Electrochemical Impedance Spectroscopy in Proton Exchange Membrane Fuel Cells Based on Genetic Algorithm

In the diagnosis of membrane flooding and drying faults in a Proton Exchange Membrane Fuel Cell (PEMFC) through Electrochemical Impedance Spectroscopy (EIS), this paper proposes a Genetic Algorithm (GA)-based feature selection method for selecting the required frequency points of failure, to reduce the measurement time taken by EIS while ensuring high diagnostic accuracy. This feature selection method searches the feature space through GA and proposes an encoding method tailored to this problem. During the searching process, three algorithms, i.e., Backpropagation Neural Network (BPNN), K-Nearest Neighbor (KNN), and eXtreme Gradient Boosting (XGBoost), are used to extract various features and select higher diagnostic rates of feature frequencies. Comparisons are made between the feature frequencies selected by the proposed method and those selected by conventional methods based on empirical experience, and it is found that the feature frequencies selected by the proposed method have better diagnostic performance.