Fault Diagnosis of Proton Exchange Membrane Fuel Cell Based on Nonlinear Impedance Spectrum

Electrochemical impedance spectroscopy (EIS) contributes to developing the fault diagnosis tools for fuel cells, which is of great significance in improving service life. The conventional impedance measurement techniques are limited to linear responses, failing to capture high-order harmonic responses. However, nonlinear electrochemical impedance analysis incorporates additional nonlinear information, enabling the resolution of such responses. This study proposes a novel multi-stage fault diagnosis method based on the nonlinear electrochemical impedance spectrum (NEIS). First, the impact of alternating current excitation amplitude on NEIS is analyzed. Then, a series of experiments are conducted to obtain NEIS data under various fault conditions, encompassing recoverable faults like flooding, drying, starvation, and their mixed faults, spanning different degrees of fault severity. Based on these experiments, both EIS and NEIS datasets are established, and principal component analysis is utilized to extract the main features, thereby reducing the dimensionality of the original data. Finally, a fault diagnosis model is constructed with the support vector machine (SVM) and random forest algorithms, with model hyperparameters optimized by a hybrid genetic particle swarm optimization (HGAPSO) algorithm. The results show that the diagnostic accuracy of NEIS is higher than that of traditional EIS, with the HGAPSO-SVM model achieving a 100% accurate diagnosis under the NEIS dateset and self-defined fault labels.