Abstract: Lithium plating on the negative electrode is one of the critical issues restricting the safety and lifespan of lithium-ion batteries. To enhance the safety and extend the lifespan of lithium-ion batteries, a lithium plating diagnosis method is proposed which is based on multidimensional feature mining and cluster analysis. Low-temperature lithium plating experiments are designed, and experimental data of batteries are collected. A high-precision equivalent circuit model of the battery is established, and a lithium plating feature extraction method based on model parameter identification and capacity increment analysis, as well as a feature space dimension reduction method based on principal component analysis, are proposed. Based on this, an adaptive grading diagnosis method for lithium-ion battery lithium plating faults is proposed using a density-based clustering algorithm optimized by particle swarm optimization, and the accuracy of the proposed method is verified based on the difference in capacity before and after lithium plating and scanning physical detection methods. The diagnostic results show that the lithium plating diagnosis results based on multi-dimensional features are optimal. Compared with single-dimensional lithium plating diagnosis methods based on battery model features, the missed diagnosis rate decreases by 8.00%, and compared with single-dimensional lithium plating diagnosis methods based on capacity increment curve features, the missed diagnosis rate decreases by 8.00% and the misdiagnosis rate decreases by 3.63%. At the same time, scanning electron microscope and inductively coupled plasma inspection results are consistent with diagnostic results, and can accurately diagnose mild and severe lithium plating, realizing graded diagnosis of lithium plating in lithium-ion batteries.

Key words: lithium-ion batteries, lithium plating detection, cluster analysis, model parameter identification, principal component analysis