多特征提取的可解释性锂电池健康状态估计方法研究
Research on explainable lithium battery health state estimation method with multi-feature extraction
王奥博 1霍为炜 2贾云旭1
作者信息
- 1. 北京信息科技大学 机电工程学院,北京 100192
- 2. 北京信息科技大学 机电工程学院,北京 100192;新能源汽车北京实验室,北京 100192
- 折叠
摘要
锂离子电池健康状态(SOH)的估算是电池管理系统(BMS)的重要组成部分.准确预测锂离子电池的SOH对于确保其安全稳定运行至关重要.为解决容量衰退状态难获取、"黑箱"模型缺乏透明度的问题,提出了一种基于多特征提取的可解释性锂离子电池健康状态估计方法.在数据处理阶段,通过引入直接测量与二阶处理相结合的方法提取健康特征;通过XGBoost模型估计电池SOH,引入SHAP算法,分别从局部循环和全局水平解释了各健康特征对预测结果的边际贡献;通过对3种电池的SOH预测实验验证了所提方法的有效性.对比实验结果表明:所提出的锂离子电池容量预测模型的平均绝对误差(MAE)和均方根偏差(RMSE)分别小于0.7％和1.0％.
Abstract
The estimation of the state of health (SOH)for lithium-ion batteries is an essential part for battery management system (BMS).It is crucial to have an accurate prediction for the SOH of lithium-ion batteries to ensure the safe and stable operation.In order to address the issues of difficult access to the capacity decline state and the lack of transparency of the"black-box"model,this paper proposes an explainable lithium-ion batteries SOH estimation method based on multi-feature extraction.Firstly,in the data processing stage,the health features are extracted by introducing a combination of direct measurement and second-order processing;then the battery SOH is estimated by the XGBoost model,and the Shapley additive explanations (SHAP)algorithm is introduced to explain the marginal contribution of each health feature to the prediction results from the local loop and global levels,respectively;finally,the effectiveness of the proposed method is verified by SOH prediction experiments on three batteries.The results of the comparative experiments indicate that the mean absolute error (MAE)and root mean square error (RMSE )of the proposed lithium-ion battery capacity prediction model are lower than 0.7％ and 1.0％,respectively.
关键词
锂离子电池/健康状态/可解释性/多特征提取/XGBoostKey words
lithium-ion batteries/state of health/Shapley additive explanations/multi-feature extraction/XGBoost引用本文复制引用
基金项目
国家自然科学基金面上项目(52077007)
出版年
2024
国家科技期刊平台
NSTL
万方数据