汽车工程 ›› 2023, Vol. 45 ›› Issue (5): 814-824.doi: 10.19562/j.chinasae.qcgc.2023.05.011

所属专题: 新能源汽车技术-动力电池&燃料电池2023年

• • 上一篇    下一篇

基于数据驱动的电动汽车电池安全风险预测

胡杰(),余海,杨博闻,程雅钰   

  1. 武汉理工大学,现代汽车零部件技术湖北省重点实验室,武汉  430070
    2.武汉理工大学,汽车零部件技术湖北省协同创新中心,武汉  430070
    3.新能源与智能网联车湖北工程技术研究中心,武汉  430070
  • 收稿日期:2022-09-15 修回日期:2022-11-09 出版日期:2023-05-25 发布日期:2023-05-26
  • 通讯作者: 胡杰 E-mail:auto_hj@163.com
  • 基金资助:
    湖北省科技重大专项(2021AAA001)

Battery Safety Risk Prediction for Data-Driven Electric Vehicles

Jie Hu(),Hai Yu,Bowen Yang,Yayu Cheng   

  1. Wuhan University of Technology,Hubei Key Laboratory of Modern Auto Parts Technology,Wuhan  430070
    2.Wuhan University of Technology,Auto Parts Technology Hubei Collaborative Innovation Center,Wuhan  430070
    3.Hubei Technology Research Center of New Energy and Intelligent Connected Vehicle Engineering,Wuhan  430070
  • Received:2022-09-15 Revised:2022-11-09 Online:2023-05-25 Published:2023-05-26
  • Contact: Jie Hu E-mail:auto_hj@163.com

摘要:

为了对电池安全风险进行准确预测,本文提出基于一种车-天气-驾驶员的多指标电池安全风险预测方法。首先提取车内外多维度信息即运用数据挖掘提取了天气状况、汽车行驶工况和驾驶风格等多指标特征,以模拟实际的电池应用场景;然后通过随机森林和SHAP组合模型的方式对特征进行筛选,从而提高了模型的泛化性和鲁棒性;最后将电池安全风险预测问题解耦为机器学习预测和时间序列预测问题,分别选择XGBoost和随机森林模型进行预测,并在此基础上建立新的Stacking集成模型对电池安全风险进行预测。最终模型的预测效果和数据实验的结果表明,该方案对电动汽车电池安全风险能做出较为准确的预测,可以为安全化、智能化的电池管理系统提供辅助决策信息。

关键词: 电池安全, 多指标特征, Stacking集成, 数据实验

Abstract:

In order to accurately predict the battery safety risk, a multi-index battery safety risk prediction method based on vehicle-weather-driver is proposed in this paper. Firstly, multi-dimensional information inside and outside the vehicle is extracted, i.e. multi-index characteristics such as weather condition, driving conditions and driving style are extracted by data mining to simulate the actual battery application scenario. Then, features are filtered by random forest and SHAP combination model, which improves the generalization and robustness of the model. Finally, the battery safety risk prediction problem is decoupled into machine learning prediction and time series prediction problems, and XGBoost and random forest models are selected to predict respectively. On this basis, a new Stacking integrated model is established to predict the battery safety risk. According to the predictive effect of the final model and the results of data experiment, the scheme can make a more accurate prediction of the battery safety risk of EV and provide decision-making information for safe and intelligent battery management system.

Key words: battery safety, multi-index feature, Stacking integration, data experiment