You are using an outdated browser. Please upgrade your browser to improve your experience.
Javascript is disabled in your web browser. For full functionality of this site it is necessary to enable JavaScript.
This website is using cookies.
We use them to give you the best experience. If you continue using our website, we'll assume that you are happy to receive all cookies on this website.
CPOTE2020 logo
6th International Conference on
Contemporary Problems of Thermal Engineering
Online | 21-24 September 2020

Abstract CPOTE2020-1128-A

Book of abstracts draft
slider slider slider slider slider slider

Modeling electrochemical reaction in solid oxide fuel cell's anode with an artificial neural network-supported numerical simulation

Szymon BUCHANIEC, AGH University of Science and Technology, Poland
Marek GNATOWSKI, AGH University of Science and Technology, Poland
Grzegorz BRUS, AGH University of Science and Technology, Poland

The presented study focuses on a numerical simulation of the charge 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 electrochemical reaction symmetry coefficient. A modified formulation of the problem includes data-driven correction of reaction symmetry coefficient in the electrochemical reaction model. A dedicated computational scheme was developed in which a deep-artificial neural network updates symmetry coefficients depend on operational conditions and available data sets. Neural network was learned on 12 experimental data points of polarization curve of an anode obtained from literature. Learning set contained data for the anode operating in two different temperatures - 800 and 1000 degree Celsius. Test set contained six data points for anode operating in 900 degree Celsius. Coefficients' form proposed by the ANN differ from typically used in the literature. Instead of being independent from temperature and withdrawn current, ANN proposed coefficients that change nonlinearly with these variables. Change of coefficients with temperature and withdrawn current is in accordance with electrochemical laws. The results of predictions are juxtaposed with the experimental data from the literature, giving an excellent agreement and indicating improvement of the mathematical model of SOFC. Mean square error on learning data set was 5.312e-5 and on test data points equal to 4.697e-5. It was shown that such a combined approach could lead to successful prediction of anode overpotential and could be a useful tool in the optimization-design process.

Keywords: Solid oxide fuel cell (SOFC), Artificial neural networks, Evolutionary algorithm, Mathematical modelling, Anode
Acknowledgment: This work was supported by the Foundation for Polish Science under FIRST TEAM program No. First TEAM/2016-1/3 co-financed by the European Union under the European Regional Development Fund