Polymer electrolyte membrane fuel cell fault diagnosis and sensor abnormality identification using sensor selection method

In this study, a sensor selection technique is proposed to identify various system faults including polymer electrolyte membrane (PEM) fuel cell stack faults and faults of ancillary systems, thus improving system reliability and durability. With proposed technique, the information in sensors could be investigated without fuel cell numerical model, and sensors more sensitive to the fuel cell system performance change could be identified for reliable fault diagnosis. Moreover, the reliability of sensors can be evaluated during the system operation with proposed technique. The performance of selected sensors with proposed technique in identifying fuel cell system faults is investigated using test data from a PEM fuel cell system, where the data-driven fault diagnostic framework is applied. Results demonstrate that with the selected sensors, different levels of fuel cell stack faults can be distinguished with good quality, and the sensor faults can also be identified during the fuel cell operation. Therefore, the proposed sensor selection technique can be beneficial in practical PEM fuel cell systems for the identification of various system faults, from which mitigation strategies could be taken to improve the system reliability and durability, while the maintenance cost could be reduced by avoiding unnecessary system stop and maintenance actions.