Abstract: With the continuous increase in the number of new energy vehicles, the safety issues of their traction battery systems have gradually emerged. To address the issues of complex and diverse fault types in power battery systems and the insufficient detection accuracy of traditional diagnostic methods, this paper combines decision tree (DT) and adaptive boosting (AdaBoost) algorithms to propose a battery fault diagnosis and prediction method based on the DT-AdaBoost model. By constructing a multi-dimensional feature dataset of typical battery fault characteristics, including voltage, current, temperature, and cycle number, and after data cleaning, feature selection, and standardization preprocessing, the adaptive weight adjustment mechanism of the AdaBoost algorithm is used to integrate the weak learner DT. The training process of the DT-AdaBoost model is designed, key parameters are optimized, and stability is ensured through cross-validation. This research provides an effective solution for the intelligent diagnosis and prediction of traction battery faults in new energy vehicles, which can reduce maintenance costs and improve battery system safety, holding significant engineering application value.

Key words: new energy vehicles; traction battery; fault diagnosis; prediction and early warning

摘要： 随着新能源汽车保有量的不断增加，新能源汽车动力电池系统的安全性问题逐步显现。 针对动力电池系统故障类型的复杂多样、传统诊断方法检测精度不足等问题，文章结合决策 树（DT）与自适应增强（AdaBoost）算法，提出基于 DT-AdaBoost 模型的电池故障诊断与预 测方法。通过构建电压、电流、温度、循环次数等电池典型故障特征多维度特征数据集，经 数据清洗、特征筛选与标准化预处理后，利用 AdaBoost 算法的自适应权重调整机制集成弱学 习器 DT，设计 DT-AdaBoost 模型训练流程并优化关键参数，通过交叉验证保障稳定性。研究 为新能源汽车动力电池故障的智能化诊断与预测提供了有效解决方案，可降低维护成本、提 升电池系统安全性，具有重要的工程应用价值。

关键词: 新能源汽车；动力电池；故障诊断；预测预警