Water Fault Diagnosis for PEMFC based on an Improved Equivalent Circuit Model

The water faults significantly impact the durability and performance of the proton exchange membrane fuel cell (PEMFC). As a non-invasive detection method, the equivalent circuit model (ECM) based water fault diagnostic method offers great interpretability for internal states of the PEMFC. However, the conventional ECM of the PEMFC has poor fitting accuracy in the high-frequency region. To address this, an improved ECM is proposed by introducing constant phase element (CPE) and short Warburg impedance elements, which can reflect the effects of plate unevenness and dielectric heterogeneity. Compared to the second-order RC and R(RC)(RC) models, the proposed ECM increases the fitting accuracy by 96.5% and 97.9%, respectively. The features derived from the ECM are used to develop the water fault classification model based on the particle swarm optimization clustering. To validate the proposed method, a PEMFC water fault dataset are generated from experiments, and all 35 sets of data are accurately classified. Compared to the K-means clustering, the classification precision of the proposed method analysis increased by 23.86%.