Pemfc Fault Classification Based on Noisy and Noiseless Eis Impedance Spectra

Detection and classification of Polymer Electrolyte Membrane Fuel Cells faults is effectively based on the analysis of frequency domain data. The latter might be affected by different levels of measurement noise, which might depend on the voltage and current signals amplitudes, on the quality of the measurements instrumentation and on the software approaches used to filter the data. This paper is aimed at showing that different approaches, which are data driven and model based ones, perform differently when they operate on noisy or noiseless impedance spectra. The diagnostic techniques operates on frequency domain experimental data acquired through the Electrochemical Impedance Spectroscopy. The available measurements come from a whole stack as well as from single cells and show differ signal to noise ratios. The method based on the fuel cell physics appears to be more sensitive than the data driven one when noisy measurements are used. The results presented in the manuscript also demonstrate the feasibility of both approaches in on-line and on-site applications, thus when they run on a low-cost embedded platform