Parameter Identification for A Proton Exchange Membrane Fuel Cell Model 2020-01-0858
The proton exchange membrane fuel cell (PEMFC) system has emerged as the state-of-art power source for the electric vehicle, but the widespread commercial application of fuel cell vehicle is restricted by its short service life. An enabling high accuracy model holds the key for better understanding, simulation, analysis, subsystem control of the fuel cell system to extract full power and prolong the lifespan. In this paper, a quasi-dynamic lumped parameters model for a 3kW stack is introduced, which includes filling-and-emptying cavity sub-models for the relationships between periphery signals and internal states, static water transferring sub-model for the membrane, and empirical electrochemical sub-model for the voltage response. Several dynamic experiments are carried out to identify unknown parameters of the model. According to the periphery measurable signals, the model is parameterized using a hybrid genetic algorithm (GA)/particle swarm optimization (PSO) method, which combined the advantages of conventional GA and PSO to reduce risks of being trapped into local optima. Comparison of the identified results and test voltages shows that the model is capable of predicting the voltage response with the relative errors between simulated and measured results less than 5%. Then, the properties of internal states with respect to the time, such as gas partial pressure within the cavity, the water content of the membrane, and relative humidity, are analyzed. Lastly, a step simulated current is injected into the model under different periphery conditions, and the frequency relationships between current and internal states are analyzed using a fast Fourier transform method. To decrease the spectral leakage error, the time domain data is processed by the Hanning window.