Risk early warning of electric vehicle battery system based on machine learning

doi:10.16265/j.cnki.issn1003-3033.2023.02.1289

Abstract

Abstract:

In order to improve the timeliness and accuracy of safety early warning of power battery under real vehicle conditions, the safe early warning of the battery system was refined into two scientific problems: key state prediction and early warning classification based on predicted state. According to the battery state data in the real vehicle operation, the maximum value of single cell voltage and the range of the cell voltage were selected as the key prediction objects. Fisher scoring and Maximum Information Coefficient (MIC) were used to realize key feature selection, and Sample Convolution and Interaction Network model (SCINet) were used for key state prediction. Then, based on the predicted state, a multi-classification RF model was established to classify the safety risks of power batteries. The results show that the proposed model has a strong predictive ability for multiple parametrs of the battery. For example, the root mean square error (RMSE) of the highest cell voltage is 0.027 1 and the highest temperature is 0.054 0 after 1 min of prediction. The prediction accuracy of the safety risk level of the battery system after 1 min is 84%, and the macro-average f₁ score is 74%.

Key words: machine learning, electric vehicle, battery system, risk early warning, sample convolution and interaction network(SCINet), random forest (RF)

HE Shubo, XIANG Wei, SHI Zhongmiao. Risk early warning of electric vehicle battery system based on machine learning[J]. China Safety Science Journal, 2023, 33(2): 159-165.

Figures/Tables 10

Fig.1

Tab.1

Tab.2

Fig.2

Fig.3

Tab.3

Fig.4

Fig.5

Tab.4

Tab.5

References 10

[1]	CHEN Siyan, GAO Zhenghai, SUN Tianjun. Safety challenges and safety measures of Li-ion batteries[J]. Energy Science & Engineering, 2021, 9(9): 1647-1672.
[2]	黄强, 陶风波, 刘洋, 等. 气液灭火剂对磷酸铁锂电池模组灭火能效研究[J]. 中国安全科学学报, 2020, 30(3): 53-59. doi: 10.16265/j.cnki.issn1003-3033.2020.03.009
	HUANG Qiang, TAO Fengbo, LIU Yang, et al. Study on performance of gas-liquid extinguishing agent for lithium iron phosphate battery modules[J]. China Safety Science Journal, 2020, 30(3): 53-59. doi: 10.16265/j.cnki.issn1003-3033.2020.03.009
[3]	刘同宇, 李师, 付卫东, 等. 大容量磷酸铁锂动力电池热失控预警策略研究[J]. 中国安全科学学报, 2021, 31(11): 120-126. doi: 10.16265/j.cnki.issn 1003-3033.2021.11.017
	LIU Tongyu, LI Shi, FU Weidong, et al. Study on early warning strategy of large LFP traction battery's thermal runaway[J]. China Safety Science Journal, 2021, 31(11): 120-126. doi: 10.16265/j.cnki.issn 1003-3033.2021.11.017
[4]	FENG Fei, HU Xiaosong, HU Lin, et al. Propagation mechanisms and diagnosis of parameter inconsistency within Li-ion battery packs[J]. Renewable and Sustainable Energy Reviews, 2019, 112(2):102-113. doi: 10.1016/j.rser.2019.05.042
[5]	LI Da, ZHANG Zhaosheng, LIU Peng, et al. Battery fault diagnosis for electric vehicles based on voltage abnormality by combining the long short-term memory neural network and the equivalent circuit model[J]. IEEE Transactions on Power Electronics, 2020, 36(2): 1303-1315. doi: 10.1109/TPEL.63
[6]	HONG Jichao, WANG Zhenpo, YAO Yongtao. Fault prognosis of battery system based on accurate voltage abnormity prognosis using long short-term memory neural networks[J]. Applied Energy, 2019, 251:DOI: 10.1016/j.apenergy.2019.113381. doi: 10.1016/j.apenergy.2019.113381
[7]	ZHANG Liang, GAO Tian, CAI Guowei, et al. Research on electric vehicle charging safety warning model based on back propagation neural network optimized by improved gray wolf algorithm[J]. Journal of Energy Storage, 2022, 49:DOI: 10.1016/j.est.2022.104092. doi: 10.1016/j.est.2022.104092
[8]	GAN Min, ZHANG Li. Iteratively local fisher score for feature selection[J]. Applied Intelligence, 2021, 51(8): 6167-6181. doi: 10.1007/s10489-020-02141-0
[9]	ZHENG Kangfeng, WANG Xiujuan, WU Bin, et al. Feature subset selection combining maximal information entropy and maximal information coefficient[J]. Applied Intelligence, 2019, 50(2): 487-501. doi: 10.1007/s10489-019-01537-x
[10]	LIU Minhao, ZENG Ailing, CHEN Muxi, et al. Time series is a special sequence: forecasting with sample convolution and interaction[J]. arXiv Preprint arXiv, 2021:DOI: arxiv-2106.09305. doi: arxiv-2106.09305

风险等级	表示	影响
0	无故障	正常
1	1级故障	不影响车辆正常行驶
2	2级故障	影响车辆性能,驾驶员限制行驶
3	3级故障	应立即停车处理或请求救援

编号	名称	编号	名称
A1	充放电时长	A7	最低温度值
A2	总电压	A8	SOC变化值
A3	总电流	A9	单体温度极差
A4	单体电压最高值	A10	单体电压极差
A5	最高温度值	A11	SOC
A6	单体电压最低值

特征	Informer	SSA-LSTM	SCINet
总电压	10.765 0	9.633 6	7.913 1
最高温度值	1.685 9	0.855 9	0.556 7
最低温度值	1.854 9	0.811 1	0.515 6
SOC变化值	13.903 0	12.211 0	9.602 8
温度极差	1.312 1	0.817 1	0.496 6
单体电压极差	0.059 0	0.039 9	0.032 7
单体电压最高值	0.090 1	0.072 7	0.027 1
单体电压最低值	0.084 0	0.066 6	0.054 0

算法	P	R	F₁	R_A
RF	0.84	0.66	0.74	0.79
KNN	0.56	0.67	0.60	0.78
DEC	0.77	0.63	0.69	0.77
GBDT	0.57	0.65	0.61	0.80

时长	P	R	F₁
第4步	0.84	0.66	0.74
第8步	0.80	0.57	0.67
第12步	0.75	0.53	0.62