Comparative analysis and evaluation of PEMFC machine learning surrogates by bridging CFD and experimental data

Cam Tu Ngo, Ba Hieu Nguyen, WanTae Lee, HyunChul Kim,
Comparative analysis and evaluation of PEMFC machine learning surrogates by bridging CFD and experimental data,
Journal of Industrial and Engineering Chemistry,
2025,
,
ISSN 1226-086X,
https://doi.org/10.1016/j.jiec.2025.10.027.
(https://www.sciencedirect.com/science/article/pii/S1226086X25006963)
Abstract: Proton exchange membrane fuel cells (PEMFC) require powerful analytical methods to manage their complicated and nonlinear operating dynamics. This study presented a thorough methodology using verified 3D Computational Fluid Dynamics (CFD) simulations combined with machine learning to precisely estimate 25 cm2 PEMFC performance. CFD data was used to train five machine learning models: random forest, gradient boosting, support vector machines, artificial neural networks, and extreme gradient boosting (XGBoost). Hyperparameter tuning and k-fold cross-validation were employed to ensure model stability and avoid overfitting. Beyond conventional performance measures such as R2 and RMSE, model behavior was assessed using error distribution evaluations, learning curves, scatter plots, and error distribution assessments. Results definitively showed that the XGBoost model exhibited exceptional predictive accuracy across all parameters, with R2 values surpassing 0.989 for power density, 0.94 for oxygen delivery uniformity, and 0.996 for system efficiency. Experimental validation further verified the surrogate model accuracy. The proposed approach enables fast and consistent performance by substituting a high-fidelity surrogate for resource-intensive CFD. This work provides a solid foundation for future research on PEMFC system optimization and control and emphasizes the importance of detailed model diagnostics in the development of physically consistent, reliable machine learning applications for renewable energy technologies.
Keywords: Computational Fluid Dynamics; Machine Learning; Surrogate Model; Proton Exchange Membrane Fuel Cell; Operating Conditions; Performance Prediction