Nonlinear identification of a DIR-SOFC stack using wavelet networks

Application of wavelet networks for identification of a direct internal reforming solid oxide fuel cell (DIR-SOFC) stack is reported in this paper. The SOFC is a complex system particularly when it is directly fueled with hydrocarbons (natural gas, coal gas, etc.). Most of the traditional models of the SOFC, based on the reforming, electrochemical and thermal modeling, are too complicated. To facilitate controller design and analysis of systems, the wavelet network dynamic model of the DIR-SOFC is constructed, avoiding the consideration of the complex processes in the fuel cells. The input and output data are used for initializing and training the wavelet network by a recursive approach. The Gram–Schmidt algorithm, the Cross-Validation method and immune selection principles are applied to optimization of the network. The simulation is performed and comparisons of characteristics under different operating conditions are given. The results show high static and dynamic accuracy of the identified model. Further, the obtained wavelet network model can be used for developing the model-based controllers of DIR-SOFC.