Development of Online Fault Diagnosis Method for PEM Fuel Cell Based on Impedance at Optimal Frequency

Online fault diagnosis efficiency and accuracy are important issues to improve the performance of the proton exchange membrane fuel cell. Electrochemical impedance spectroscopy is an effective tool to detect the state-of-health of fuel cell for laboratory research. However, its measurement time is too long in online use. This paper shows an online fault diagnosis method just based on a single impedance. First, electrochemical impedance spectroscopy was used to deal with monitoring of flooding and drying of a commercial fuel cell stack comprised of six cells and obtained impedance spectra in different state-of-health. Next, the specific frequency impedance is extracted from the impedance spectrum, considering the sampling time and the sensitivity to different water content. The optimal frequency is selected by the combination weighting method rather than subjective decision. Finally, based on the real and imaginary parts of the specified impedance, an online fault diagnosis method is developed and the state-of-health of the fuel cell stack is diagnosed by fuzzy c-means clustering. Fuzzy c-means clustering is an unsupervised algorithm which can be used to organize impedance data into different groups based on similarities among the data points. Different groups correspond to different state-of-health of fuel cell stack. Through experimental verification, the method has demonstrated high classification performance with diagnosis accuracy higher than 99%.