Compared with traditional power source vehicles, there is still a gap in the durability of fuel cell vehicles. On fuel cell vehicles, gas starvation caused by start-stop and frequent load changes is one of the significant factors that cause the fuel cell lifetime decay. Gas starvation refers to the state in which the reactant gas of the fuel cell works under substoichiometric numbers. However, due to the complexity of the fuel cell structure and reaction mechanism, the existing researches have not reached a unified and accurate diagnosis method. It is difficult to establish a clear relationship between the gas starvation state of fuel cell and its external characteristic parameters. Therefore, this paper innovatively uses the adaptive network-based fuzzy inference system (ANFIS) method in gas starvation diagnosis. In this study, a proton exchange membrane fuel cell (PEMFC) is modelled and analyzed by ANFIS. Meanwhile, CFD simulation is used as an auxiliary tool to obtain 102 samples’ data. Experimental data are obtained to verify that the steady-state response of the PEMFC is consistent with the simulation data. This research finally achieves a 92% accuracy of gas starvation prediction, and the ANFIS model can predict the size of starvation area in PEMFC cells, providing a feasible technical route for the prediction of starvation in fuel cell systems and vehicles.
Keywords Proton exchange membrane fuel cell, Gas starvation diagnosis, Artificial neural network, Adaptive network based on fuzzy interface system, Computational fluid dynamics