Artificial Intelligence and Machine Learning

Application-oriented Design of Machine Learning Paradigms for Battery Science Ying Wang of Fudan University summarizes the current state and possible future directions of machine-learning in battery science, listing key recent advances in:

High energy density batteries

Solid-state electrolytes

Fast-charging electrolytes

Calendar lifetime prediction

Recyclability To further accelerate battery R&D using machine-learning, there are several research frontiers to continue pushing: (a) The integration of state-of-the-art AI models for molecular generation and simulation acceleration, including high-throughput calculation, optimization of properties, and new materials design at all length scales. (b) Developing automation tools by integrating hardware, software, database management, and data analysis and modeling to accelerate discovery. (c) Coupled development of new analytical theory and physics-informed ML investigations to develop and test new hypotheses. (d) ‘Human-in-the-loopʼ discovery, using ML guided by hard-to-encode human intelligence and experimental evidence. Machine-learning has already transformed battery research, impacting academic and industrial work in nearly all stages of battery development and application. Future progress lies not just in applying the current suite of methods, but also pushing the boundary with new technologies alongside the development of new theory, using automated data collection and modeling to generate and test hypotheses. Source: Ying Wang, npj Computational Materials, 2025, 1189

Extracted from the PDF; charts and images may read imperfectly.