Machine learning for full lifecycle management of lithium-ion batteries

Developing advanced battery materials, monitoring and predicting the health status of batteries, and effectively managing retired batteries are crucial for accelerating the closure of the whole industrial chain of power lithium-ion batteries for electric vehicles. Machine learning technology plays a vital role in the research, production, service, and retirement of lithium-ion batteries due to its robust learning and predictive capabilities. While there have been detailed and valuable reviews on this topic, a comprehensive summary of machine learning progress from the perspective of full lifecycle management of lithium-ion batteries is still lacking. This review divides the full lifecycle of lithium-ion batteries into three stages: pre-prediction, mid-prediction, and late prediction phases, and summarizes recent advances in different machine learning methods categorized as materials screening, life prediction, and cascade utilization. It also emphasizes the implementation and evaluation of hybrid machine learning models. Finally, potential research opportunities and future directions are presented, mainly from two aspects of battery databases and the application of large-scale time-series models. This review provides valuable guidance and reference for researchers and practitioners to broaden the scope of machine learning for its application in lithium-ion batteries. •Full lifecycle management of LIBs based on machine learning methods was expounded.•The pros and cons of hybrid machine learning models were particularly discussed.•Perspectives of database and large-scale time-series models are provided.