Performance Analysis of Anode-Supported Solid Oxide Fuel Cells: A Machine Learning Approach.

Area of Science:

• Electrochemical energy conversion
• Materials science
• Computational modeling

Background:

• Solid oxide fuel cells (SOFCs) are efficient energy devices requiring extensive validation.
• Predicting SOFC performance is challenging due to complex component interactions and experimental costs.
• Mathematical models exist but have limitations regarding assumptions and computational demands.

Purpose of the Study:

• To develop a machine learning (ML) approach for predicting anode-supported SOFC performance.
• To overcome the limitations of traditional mathematical modeling in SOFC analysis.
• To accurately predict the current-voltage dependency based on cell parameters.

Main Methods:

• Collected a dataset on Ni-YSZ anode-supported SOFC performance.
• Implemented convolutional neural networks and multilayer perceptron models.
• Evaluated model accuracy using R-squared, mean squared error, and mean absolute error.

Main Results:

• The developed neural network model achieved an R-squared score > 0.998.
• The model demonstrated high accuracy with a mean squared error of 9.6 × 10⁻⁵ and mean absolute error of 6 × 10⁻³ (V).
• The neural network outperformed conventional models like Gaussian processes in predicting current-voltage dependency.

Conclusions:

• Machine learning provides a highly accurate and efficient method for predicting SOFC performance.
• The developed approach surpasses previous models in predicting the impact of cell parameters on current-voltage characteristics.
• This ML methodology is generalizable to other material datasets for energy applications.