Abstract
In this paper, the sensor selection approach is investigated with the aim of using fewer sensors to provide reliable fuel cell diagnostic and prognostic results. The sensitivity of sensors is firstly calculated with a developed fuel cell model. With sensor sensitivities to different fuel cell failure modes, the available sensors can be ranked. A sensor selection algorithm is used in the analysis, which considers both sensor sensitivity to fuel cell performance and resistance to noise. The performance of the selected sensors in polymer electrolyte membrane (PEM) fuel cell prognostics is also evaluated with an adaptive neuro-fuzzy inference system (ANFIS), and results show that the fuel cell voltage can be predicted with good quality using the selected sensors. Furthermore, a fuel cell test is performed to investigate the effectiveness of selected sensors in fuel cell fault diagnosis. From the results, different fuel cell states can be distinguished with good quality using the selected sensors.
Highlights
Due to characteristics such as zero-emissions and high efficiency, fuel cell technology has been engineered for a range of applications, including automotive, stationary power stations, and consumer devices.the fuel cell reliability and durability are still two main barriers for its wider application, which leads to a series of studies on fuel cell prognostics and health management (PHM)
This paper investigated the application of the sensor selection approach in fuel cell PHM
The performance of selected sensors in polymer electrolyte membrane (PEM) fuel cell fault diagnosis was investigated using test data from a PEM fuel cell system, and the results are compared with those using all using test data from a PEM fuel cell system, and the results are compared with those using all available sensors
Summary
Due to characteristics such as zero-emissions and high efficiency, fuel cell technology has been engineered for a range of applications, including automotive, stationary power stations, and consumer devices. Compared to fuel cell diagnostics, fewer studies have been devoted to fuel cell prognostics, and among these studies, training data from a fuel cell system is required to generate the input–output relationship of the fuel cell model for the prediction of future performance [14,15,16,17,18]. It can be concluded from the literature review that with commonly used fuel cell diagnostic and prognostic approaches, the performance relies largely on the quality of the fuel cell sensor measurements.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
