Three-dimensional Simulation of a Solid Oxide Fuel Cell Based on Current Distribution In-situ Acquired with Segmented Electrodes and Surrogate Model Preparation with Machine Learning

Abstract

The analysis of the three-dimensional current density, concentration, and temperature distribution in solid oxide fuel cell (SOFC) cells and stacks has conventionally required high computational cost. In this study, we measure the current distributions in a planar test cell with segmented electrodes, which are used to construct and verify a three-dimensional finite element (FE) model. A surrogate model is generated with machine learning using current densities for the cell voltages, gas concentrations, and temperatures as training data predicted from the FE model, which is expected to reduce the computational cost.