主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2023, Vol. 42 ›› Issue (3): 422-428.

• • 上一篇    下一篇

基于运行参数的电动汽车火灾原因判别模型

林 烨1，2，3，董红磊1，3，肖凌云1，3，曲现国1，3   

  1. （1.国家市场监督管理总局 缺陷产品管理中心，北京 100101； 2.上海机动车检测认证技术研究中心有限公司，上海 201805； 3.国家市场监管重点实验室（产品缺陷与安全），北京 102200）
  • 出版日期:2023-03-15 发布日期:2023-03-15
  • 作者简介:林 烨（1996- ），女，上海机动车检测认证技术研究中心有限公司检测工程师，主要从事汽车安全与召回工作，上海市嘉定区于田南路68号，201805。
  • 基金资助:
    国家重点研发计划项目（2020YFB1600600）

A fire cause discrimination model for electric vehicles based on operation parameters

Lin Ye1, 2, 3, Dong Honglei1, 3, Xiao Lingyun1, 3, Qu Xianguo1, 3   

  1. (1. Defective Product Administrative Center, SAMR, Beijing 100101, China; 2. Shanghai Motor Vehicle Inspection Certificatior & Tech Innovation Center Co., Ltd., Shanghai 201805, China; 3. Key Laboratory of Product Defect and Safety for State Market Regulation, Beijing 102200, China)
  • Online:2023-03-15 Published:2023-03-15

摘要： 分析2018-2021年的电动汽车火灾事故信息，归纳起火电动汽车典型特点。利用Origin分析电动汽车火灾BMS运行参数数据，发现由电池包进水、动力电池故障（电芯热失控、电池包内线路故障）引发事故的运行参数典型特征（SOC、电压、温度、绝缘电阻等）。基于此利用神经网络方法构建了基于运行参数的电动汽车火灾原因判别模型，12项运行参数为输入信息，3类起火原因为输出信息。模型通过训练达到期望误差水平0.05，检验样本正确率达到了100%，证明了模型可靠性。结果表明，不同原因电动汽车火灾事故，其运行参数具有不同特征及规律，运用原因判别模型能够有效地辅助判别电动汽车起火原因。

关键词: 电动汽车火灾, 火灾原因, BMS参数, 神经网络

Abstract: Abstract: Analyze the information of electric vehicle fire accidents in 2018-2021, and summarize the typical characteristics of electric vehicles on fire. Origin was used to analyze the operating parameters data of electric vehicle fire BMS, and found the typical characteristics of the operating parameters (SOC, voltage, temperature, insulation resistance, etc.) caused by the water ingress of the battery pack and the fault of the power battery (thermal runaway of the battery cell and the fault of the line in the battery pack). Based on this, a judgment model of electric vehicle fire cause based on operating parameters is constructed by using neural network method, with 12 operating parameters as input information and 3 types of fire causes as output information. The model reached the expected error level of 0.05 through training, and the correctness rate of the test sample reached 100%, which proved the reliability of the model. The results show that the operating parameters of electric vehicle fire accidents of different causes have different characteristics and laws, and the use of cause discrimination model can effectively assist in determining the cause of electric vehicle fire.

Key words: Key words: electric car fire, cause of fire, BMS parameters, the neural network