4 - Neural Networks—A Review of Applications

Abstract

Complementing the detailed application example of an artificial neural network (ANN) described in the previous chapter, we present here a collection of brief reviews of selected papers which apply artificial neural networks. Various approaches to ANN fuel cell modeling are considered. Some methods apply the multilayer perceptron with unidirectional information flow. Hybrid models involving an artificial neural network with radial basis function (RBF) are mentioned without going into detail regarding the network architecture and transition functions. The following topics are outlined: The role of ANN in development of low-pollution vehicles; neural network (NN) hybrid model of direct internal reforming solid oxide fuel cell (SOFC); NN-based dynamic control of proton exchange membrane (PEM) fuel cells (FCs); NN on-board FC power supply modeling; NN with SOFC control of power supply improvement; NN-based modeling of PEMFC and controller synthesis; NN model of drying and thermal degradation; NN modeling of SOFC; NN optimization of energy systems; NN model for fluidized bed; NN power control of an FC; performance prediction and analysis of the cathode catalyst layer; NN model for a PEMFC; hybrid NN models for fuel cells; ANN capillary transport characteristics of FC diffusion media; ANN of the mechanical behavior of a PEMFC; ANN simulators for SOFC performance; ANN modeling for the study and development of fuel cells. ANN to predict SOFC performance in residential spaces; NN modeling of polymer electrolyte membrane FC; genetic algorithm (GA)-RBF neural networks modeling; NN-based quality control with an SOFC plant. A review of central notions of the NN theory is cited. A brief assessment of results obtained in these investigations is also provided.