Artificial Intelligence and Machine Learning for Oxygen Reduction Reaction: A Review on Catalyst Discovery and Durability Modeling

The oxygen reduction reaction (ORR) at the cathode of proton exchange membrane fuel cells (PEMFCs) remains the principal kinetic bottleneck impeding the widespread adoption of hydrogen‐based energy technologies. The inherently sluggish four‐electron transfer process necessitates high platinum group metal (PGM) loadings, driving up costs and compromising durability, two factors that severely limit large‐scale commercialization. Despite decades of research, the rational design of durable, high‐performance ORR electrocatalysts has proven elusive, constrained by the vast compositional design space and the limitations of both trial‐and‐error experimentation and density functional theory (DFT) calculations. In recent years, artificial intelligence (AI) and machine learning (ML) have reshaped the landscape of catalyst discovery, enabling predictive structure–activity modeling, rapid development of surrogate models, and autonomous exploration of chemical design spaces. This review critically surveys AI and ML applications across the catalyst development pipeline, from dataset curation and supervised or deep learning models for activity prediction to ML‐accelerated DFT screening and Bayesian optimization for autonomous discovery. Special attention is devoted to durability and degradation modeling, which remains underdeveloped compared to activity prediction despite its central importance for practical deployment.