Modeling of Proton Exchange Membrane Fuel Cell Using Support Vector Regression Machine

Abstract

In this paper, a nonlinear offline model of proton exchange membrane fuel cell (PEMFC) is built by using a support vector regression machine (SVRM) based on particle swarm optimization (PSO) algorithm. During the process of modeling, the PSO aims to optimize the parameters of SVRM. Compared with the artificial neural network (ANN) approach, the prediction results show that the SVRM approach is superior to the conventional ANN in predicting the stack voltage with different hydrogen pressure. The mean absolute percentage error (MAPE) of 36 test samples is 0.73%, such that prediction result was provided by leave-one-out cross validation (LOOCV) test of SVRM. So it is feasible to establish the prediction model of PEMFC system by using SVRM identification based on the PSO.