Dynamic modeling of SOFC based on support vector regression machine and improved particle swarm optimization

For predicting the electrochemical and heat transfer dynamics synchronously, a dynamic identification model of the solid oxide fuel cell (SOFC) is reported. In this study, support vector regression machine (SVRM) is proposed to model the nonlinear dynamic characteristics of the SOFC. In addition, a kind of improved particle swarm optimization (IPSO) is preferably chosen for the parameter optimization of the SVRM model. The applicability of the proposed SVRM with IPSO (IPSO-SVRM) model in modeling the voltage and the temperature transient responses to the hydrogen input flow rate change of the SOFC is illustrated by the simulation. Furthermore, the comparisons between the IPSO-SVRM model and the SVRM model are provided which show a substantially better performance for the IPSO-SVRM model. The results also show that IPSO algorithm outperforms the crossover validation method in terms of parameters choice of the SVRM model.