Data-Driven Fault Diagnosis for Automotive PEMFC Systems Based on the Steady-State Identification

Data-driven diagnosis methods for faults of proton exchange membrane fuel cell (PEMFC) systems can diagnose through the state variable data collected during operation PEMFC system. However, from system stack switching between different operating points easily cause false alarms, such that practical value is reduced. To overcome this problem, a fault method based on steady-state identification proposed in paper. The support vector description (SVDD) and relevance machine (RVM) optimized by artificial bee colony (ABC) are used diagnosis. density-based spatial clustering applications with noise (DBSCAN) linear least squares fitting (LLSF) to identify abnormal datasets estimate change rates variables respectively. automatically points, so accuracy be improved alarms has certain provide reference further study.