Residual-based fault diagnosis for thermal management systems of proton exchange membrane fuel cells

•Hierarchical fault diagnosis using residuals for fuel cell systems is proposed.•Four residuals are used to diagnose thermal management systems.•Initial faults of the pumps, thermocouples, and heat exchanger are diagnosed.•The component faults can be diagnosed based on distinct residual patterns.•This method contributes to the development of fuel cell health management systems. Fuel cells are considered a highly efficient and eco-friendly alternative for energy production. However, for successful commercialization of fuel cells, it is necessary to ensure the reliability of fuel cell systems. This study proposes a real-time fault detection and diagnosis method for the thermal management system, one of the important balance of plants in proton exchange membrane fuel cell systems. A hierarchical method is developed for fault detection and diagnosis using the residuals which are differences between the analytical redundancies and the measured values as indicators. The analytical redundancies are calculated with regression models. The proposed method consists of two steps. After faults are classified into the thermal management system at the subsystem level diagnosis in the first step, specific component faults are diagnosed in the second step. The residual of the heat transfer rate is used in the first step, and the additional residuals related to two coolant flow rates and one coolant temperature are used in the second step. When faults occurred in the pumps, sensors, and heat exchanger, the experimental results showed that the residual-based method accurately diagnosed the causes and locations of problems in fuel cell systems.