Utilization of Deep Learning Approach to Enhancing PEMFC Efficiency: Analyzing Humidity Variations Toward Sustainability

Abstract

The performance of Proton Exchange Membrane Fuel Cells (PEMFCs) is highly dependent on operating conditions, particularly humidity levels, which significantly affect membrane hydration, ionic conductivity, and overall efficiency. While traditional approaches rely on laboratory experiments to study these effects, this research employs advanced deep learning techniques to model and predict PEMFC performance under varying humidity conditions. In this study, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks, along with attention mechanisms, are used to enhance predictive accuracy and capture complex nonlinear relationships. Numerical simulations conducted in ANSYS Fluent generate a dataset covering five humidity levels (20%, 40%, 60%, 80%, and 100%), which is used to train and validate the deep learning models. The findings indicate that moderate humidity (40%) yields optimal predictions, with the attention-based LSTM model achieving the highest accuracy (R² = 0.98, root mean squared error (RMSE) = 0.01). This study shows the potential of proposed models as efficient predictive tools for PEMFC optimization, providing a surrogate to costly and time-consuming experimental testing. The results also revealed that hydrogen consumption was minimized at 40% humidity, confirming that optimized humidification strategies contribute to both improved efficiency and reduced fuel demand toward sustainability.