Intelligent lithium plating detection and prediction method for Li-ion batteries based on random forest model

Abstract

Lithium plating in lithium-ion batteries (LIBs) is one of the main causes of safety accidents in electric vehicles (EVs). The study of intelligent machine learning-based lithium plating detection and warning algorithms for LIBs is of great importance. Therefore, this paper proposes an intelligent lithium plating detection and early warning method for LIBs based on the random forest model. This method can accurately detect lithium plating during the charging process of LIBs, and play an early warning role according to the detection results. First, pulse charging experiments of LIBs, including normal and lithium plating charging tests, were completed and validated using in situ characterization methods. Second, the normalized internal resistance from the pulse charging test is used to detect lithium plating in LIBs. Third, a lithium plating feature extraction method is proposed to address the lack of useful lithium plating information for LIBs during the charging process. Finally, the Random Forest machine learning technique is used to classify and predict the lithium plating of LIBs. The model validation results show that the detection accuracy of lithium plating is greater than 97.2%. This is of significance for the study of intelligent lithium plating detection algorithms for LIBs.