The prediction of the polarization curves of a solid oxide fuel cell anode with an artificial neural network supported numerical simulation

The presented study focuses on a numerical simulation of the transport phenomena inside a solid oxide fuel cell anode. The classical mathematical model leads to a notable discrepancy between measured and predicted overpotentials. One of the possible reasons is the assumption of the constant values of electrochemical reaction charge transfer coefficients. A modified formulation of the problem includes data-driven correction of reaction charge transfer coefficients in the electrochemical reaction model. Here we show a dedicated computational scheme in which an artificial neural network updates the charge transfer coefficients depending on the operational conditions and the available datasets. The neural network was trained on twelve experimental data points of an anode's polarization curve obtained from the literature. The training set contained data for the anode operating in two different temperatures - 800 degrees C and 900 degrees C. The test set contained additional six data points for an anode operating at 1000 degrees C. Charge transfer coefficients were proposed by the Artificial Neural Network as a functional relation of the temperature and withdrawn current. The results of the predictions are juxtaposed with the experimental data from the literature. It was shown that an Artificial Neural Network could improve an electrochemical reaction model in Solid Oxide Fuel Cell modeling. (c) 2021 The Author(s). Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).