Applying Machine Learning in Rechargeable Batteries from Microscale to Macroscale

Emerging machine learning (ML) methodologies are widely applied in chemistry and materials science studies and are constructing a data-driven research paradigm. This review summarizes the applications of ML in rechargeable batteries from microscale to macroscale. Specifically, ML affords an emerging strategy to explore new functional for density functional theory calculations and new potentials for molecular dynamics simulations, which are expected to significantly enhance the challenging description concerning with interfaces or amorphous structures. Besides, ML possesses great potential to mine and unveil valuable information from both experimental and theoretical datasets. A quantitative "structure-function" correlation can thusly be established, the application of which includes predicting ionic conductivity of solids as well as forecasting battery lifespan. ML also exhibits great advantages in strategy optimization, such as fast-charge protocols. Lastly, an outlook on the future combination of multi-scale simulations, experiments, and ML is provided and the role of human in the data-driven research paradigm is highlighted.