NEURAL NETWORK APPROACH FOR PERFORMANCE PREDICTION OF A SOLID OXIDE FUEL CELL

In this study, we develop a neural network (NN) based model for performance prediction of a single tubular solid oxide fuel cell (SOFC). This model utilizes the back-propagation (BP) algorithm to predict the cell voltage and electrical power. Levenberg-Marquardt back-propagation algorithm based neural network has been used to develop the prediction model. The required data for learning process of the neural network model have been obtained from simulation results of a validated nonlinear physical SOFC model under wide operation range of the input variables. After neural network model development, comparisons of the simulation results of neural network model and the nonlinear physical model at the trained and untrained conditions have been carried out. As a result, it has been seen that the neural network model developed in this study is able to make good predictions in performance behavior of the considered fuel cell due to the training strategy and the topology of the neural network adopted