Parameter Identification for One-Dimension Fuel Cell Model Using GA-PSO Algorithm 2019-01-5041

When studying on how to identify the proton exchange membrane fuel cell model parameters accurately and quickly, the model frequently used is a lumped parameter model. Compared to this kind of model, one-dimensional dynamic proton exchange membrane fuel cell model can correlate the physical parameters with output characteristics of fuel cell to predict the effects of design parameters, materials and environmental conditions, thus reducing the need for experimentation. However, there is little literature about parameter identification for one-dimensional dynamic models currently. In this paper, a one-dimension dynamic proton exchange membrane fuel cell model with many assumptions for reducing the complexity of calculation is realized in Matlab-Simulink environment. The model consists of five interacting subsystems. The GA-PSO hybrid optimization algorithm is used to identify the parameters of fuel cell model to emulate the output characteristics of different proton exchange membrane fuel cells. This hybrid algorithm is an improved Particle Swarm Optimization Algorithm relying on Genetic Algorithm's strong global search ability, with the aim of maintaining the population diversity and avoiding premature convergence. The result shows that the Relative Errors between experimental data and the corresponding simulated data are less than 1% by setting the model parameters using the results from the new hybrid optimization method. In addition, the dynamic simulation is also carried out. Taking the dynamically changing current as an input, the voltage and power response of this model are compared with conclusions of other researchers. The rationality of the dynamic simulation results further verifies the reliability of the model.