该专利提出了一种利用支持向量机（SVM）算法对退役电池进行快速分选的方法。首先根据退役电池的外部参数构建二值特征向量，然后进行一次分类，区分梯次使用和直接回收的电池。之后对电池进行高倍率充电，获得增量容量（IC）曲线，提取关键特征作为二次分选指标。将这些结果与一次分类相结合，输入到多分类模型中，实现精准分选，提高处理不同健康状态电池的效率和准确性。

Recycling of massive spent lithium-ion batteries (LIBs) is urgently required with the development of electric vehicles and energy storage industries. However, due to their complex composition and uncertain state, spent LIBs pose significant fire hazards during the recycling process. In this work, liquid nitrogen (LN) and dry ice (DI) were utilized as refrigerants to investigate the inerting mechanism and thermal stability of spent LIBs. Post-mortem and thermal analyses indicated that when spent LIBs are subjected to low temperatures (below −60 °C), the solidification of the electrolyte and the separation of internal components cause an increase in internal resistance, leading to a drop in terminal voltage where it cannot deliver energy. Nail penetration tests demonstrated that cryogenic freezing effectively suppresses thermal runaway, reducing peak internal battery temperatures from 921.2 °C to below 150 °C, with a temperature rise rate suppressed to under 3 °C/s. Additionally, DI exhibited a more sustained cooling effect than LN and is proposed as a safer and more cost-effective alternative for enhancing safety in LIBs recycling.

With the increasing number of lithium-ion batteries (LIBs) reaching the end of their life, the potential for reusing these retired LIBs has become a critical area of research. However, inconsistencies among retired batteries and high sorting costs remain major obstacles. Current methods rely on experimental testing or algorithm-based data analysis to identify aging indicators, yet they often fail to balance precision with efficiency. To address these challenges, this study introduces an optimized high-rate incremental capacity (IC) curve acquisition method for the rapid and accurate extraction of sorting features. By systematically evaluating the trade-offs between model accuracy and sorting time, two feature combination strategies are proposed to improve the sorting process. The proposed method employs a two-step classification strategy for high-efficiency classification, achieving an accuracy of 98.1 % and 95.1 % for binary and multi-class classification, respectively. Compared to conventional methods, it increases sorting efficiency sixfold and significantly improves battery consistency, with capacity and internal resistance enhanced by 64.85 % and 82.71 %, respectively. This high-precision, machine-learning-based sorting approach addresses critical barriers in LIBs reuse, enabling reduced sorting costs and improved battery performance. It represents a significant step toward a circular economy for LIBs and contributes to the broader development of sustainable energy storage technologies.