Machine learning in energy storage material discovery and performance prediction

Abstract

Energy storage material is one of the critical materials in modern life. However, due to the difficulty of material development, the existing mainstream batteries still use the materials system developed decades ago. Machine learning (ML) is rapidly changing the paradigm of energy storage material discovery and performance prediction due to its ability to solve complex problems efficiently and automatically. Various excellent works are constantly emerging in the field of ML assisted or dominated development of energy storage material, such as exploring of new materials, studying of battery performance, investigating of battery aging mechanism. In this paper, we methodically review recent advances in discovery and performance prediction of energy storage materials relying on ML. After a brief introduction to the general workflow of ML, we provide an overview of the current status and dilemmas of ML databases commonly used in energy storage materials. The typical applications and examples of ML to the finding of novel energy storage materials and the performance forecasting of electrode and electrolyte materials. Furthermore, we explore the dilemmas that will be faced in the development of applied ML-assisted or dominated energy storage materials and propose a corresponding outlook. This review systematically summarizes the current development of ML-assisted energy storage materials research, which is expected to point the way for its further development.