Data-Driven Fault Diagnosis Method for PEMFC Based on NCA-KNN

In this paper, a data-driven fault diagnosis method for proton exchange membrane fuel cell (PEMFC) is proposed based on the neighborhood component analysis (NCA) and the k- nearest neighbor (KNN). Firstly, the electrochemical impedance under normal and fault states is measured by the real system bench and, together with the operating parameters, form the database applied in this paper. Then, in combination with the corresponding frequency range of each polarization process in the mechanism studies, the variations of impedance amplitude under different states are analyzed to prove that the impedance amplitude at each frequency point can reflect the evolution of the internal polarization process. Average single chip voltages, currents and impedance amplitude at each frequency point are selected as feature data for fault diagnosis (classification). Finally, the NCA-KNN algorithm is applied for feature dimension reduction and fault classification and is compared with other methods to demonstrate the advantages of this method in diagnostic accuracy and efficiency with test accuracy of 97.67% and diagnosis time of 0.0009885s.