Applications of machine learning in the design and discovery of electrocatalysts

With the continuous growth of global energy demand and increasingly severe environmental issues, developing efficient and clean energy conversion technologies has become a critical task for sustainable development. Electrocatalysis, which can couple with renewable energy sources like solar and wind to produce clean fuels and high-value chemicals, plays a pivotal role in mitigating climate change and environmental pollution. It is widely regarded as a core technology supporting sustainable development goals. The fundamental challenge lies in discovering and designing high-performance, stable, and low-cost catalytic materials that can efficiently and selectively drive key reactions such as the oxygen evolution reaction, hydrogen evolution reaction, oxygen reduction reaction, and carbon dioxide reduction reaction. Despite significant progress in materials science and electrochemistry, the discovery and optimization of electrocatalysts still rely heavily on time-consuming and repetitive trial-and-error approaches. Both experimental workflows—involving the synthesis and characterization of numerous candidate materials—and theoretical simulations, such as computationally expensive density functional theory (DFT) calculations, are often constrained by empirical judgment and high resource costs. These methods struggle to navigate the vast, multi-dimensional composition-structure-property space of potential catalysts efficiently. In recent years, the rapid development of data science and artificial intelligence (AI) has created new opportunities for accelerating electrocatalysis research. Machine learning (ML), in particular, demonstrates strong potential for intelligently exploring the chemical and material design space. By learning complex, non-linear relationships from existing experimental or computational datasets, ML models can rapidly predict key properties like adsorption energies, overpotentials, and stability, bypassing the need for exhaustive calculations or experiments. This data-driven paradigm is fundamentally transforming traditional research workflows. This review systematically summarizes the AI-driven paradigm shift in electrocatalyst design over the past five years, with a focus on representative material systems, including alloys, metal oxides, and single-atom catalysts. We highlight recent progress in ML-assisted electrocatalysis from three key perspectives: performance prediction, high-throughput screening, and mechanism analysis. This review begins by outlining the general ML workflow, covering data acquisition, feature engineering, model selection, and validation strategies. We then discuss recent advances in ML-enabled high-throughput performance prediction and screening, which have accelerated the identification of promising catalyst candidates. The review also examines how ML aids in the microstructural optimization of known materials, for instance, by guiding doping strategies, surface modification, or defect engineering to enhance activity and durability. Furthermore, we highlight the emerging trend of integrating theoretical modeling with experimental validation through active learning and autonomous robotic platforms, paving the way for a closed-loop, intelligent design framework for electrocatalysts. This iterative cycle of ML prediction, automated synthesis, and high-throughput testing significantly shortens the innovation timeline. Finally, we address key challenges and future research directions, including the scarcity of high-quality, standardized data, the interpretability of complex ML models, and the need for closer integration between computational and experimental communities. Future advancements are expected to involve multi-task learning, generative models for inverse design, and more sophisticated digital platforms that unify data, models, and experiments. By addressing these issues, AI-powered electrocatalyst discovery promises to not only innovate research paradigms but also accelerate the transition towards an efficient and sustainable energy system. This review aims to provide new insights and methodologies for researchers working at the intersection of electrochemistry, materials science, and artificial intelligence.