Review of machine learning application in porous transport layers for the design of fuel cells and electrolyzers.

Gas diffusion layers (GDLs) of fuel cells and porous transport layers (PTLs) of electrolyzers are central to mass transportation, heat management, and longevity in electrochemical energy systems. Their microstructures, being porous or sintered, present challenges to traditional simulation methods due to multiscale and nonlinear transport phenomena. Machine learning (ML) has emerged as a potential substitute in the last few years for modeling, optimization, and reconstruction of such porous media. This review consolidates 69 peer-reviewed articles that applied ML techniques such as convolutional neural networks (CNNs), artificial neural networks (ANNs), physics-informed neural networks (PINNs), and generative models to GDLs and PTLs. Bibliometric analysis indicates a dramatic explosion in publication after 2020, with a preponderance of interest in GDLs in PEM fuel cells and little interest in PTLs in water electrolyzers. Across the literature, CNNs and ANNs dominate while generative and physics-aware techniques are underrepresented. Key challenges are shortage of datasets, generalizability of material, and rare explainability or quantification of uncertainty. We highlight the need for benchmark datasets, physics–ML hybrid models, transfer learning between domains, and digital twin platforms for real-time deployment. Our review provides a blueprint towards merging ML with physical laws to accelerate porous material design in fuel cells and electrolyzers.