Machine learning for guiding high-temperature PEM fuel cells with greater power density

Area of Science:

• Materials Science
• Electrochemistry
• Computational Science

Background:

• High-temperature polymer electrolyte membrane fuel cells (HT-PEMFCs) offer advantages like low-cost hydrogen utilization and simplified thermal management.
• Improving the power density of HT-PEMFCs is crucial for their widespread adoption in energy conversion technologies.

Purpose of the Study:

• To demonstrate the utility of Machine Learning (ML) tools in optimizing HT-PEMFC performance.
• To identify strategies for surpassing 1 W cm⁻² power density in HT-PEMFCs.

Main Methods:

• Development of a 0-D, semi-empirical model for HT-PEMFC polarization behavior.
• Application of support vector regression with a radial basis function kernel to existing datasets.
• Utilizing dimension reduction and density-based clustering on model-generated synthetic data.

Main Results:

• ML tools efficiently explored vast parameter spaces to guide performance improvements.
• The study revealed specific pathways to achieve power densities exceeding 1 W cm⁻².
• Successful application of ML to analyze and predict HT-PEMFC performance.

Conclusions:

• Machine learning provides a powerful framework for accelerating the optimization of HT-PEMFCs.
• The identified pathways offer practical guidance for designing next-generation HT-PEMFCs with enhanced power output.
• This work highlights the potential of data-driven approaches in advancing fuel cell technology.