摘要
电动汽车的电池健康状态(state of health,SOH)监测系统对电池的回收和再利用具有重要意义。为了提高电动汽车电池的回收利用率、确保电池性能的最大化和安全性,提出了一种改进的电池容量计算区间及相关特征提取方法。通过密度聚类、Z-score(标准分数)和高斯滤波方法,对电池SOH数据去噪和规范化处理。采用门控循环单元(gated recurrent unit,GRU)模型进行训练和测试。实验结果表明,SOH预测的均方根误差在3.5%以内,平均值误差在2.5%以内。该模型由于神经元个数较少,预测速度较快,对边缘部署有一定的参考价值。
The battery state of health(SOH)monitoring system of electric vehicles is important for the recovery and reuse of batteries.In order to improve the recycling rate of electric vehicle batteries and ensure the maximization of battery performance and safety,an improved battery capacity calculation interval and related feature extraction method are proposed.The battery SOH data are denoised and normalized by density clustering,Z-score(standard score)and Gaussian filtering methods.The gated recurrent unit(GRU)model is used for training and testing.Experimental results show that the root mean square error of SOH prediction is within 3.5%and the mean error is within 2.5%.This model is a good reference for edge deployment due to the smaller number of neurons and faster prediction speed.
作者
陈子标
章忌
方卫东
CHEN Zibiao;ZHANG Ji;FANG Weidong(School of Electronic,Electrical Engineering and Physics,Fujian University of Technology,Fuzhou 350118,China)
出处
《福建理工大学学报》
2025年第1期86-94,共9页
JOURNAL OF FUJILAN UNIVERSITY OF TECHNOLOGY
基金
福州市科技计划项目(2022-Q-009)
福建理工大学科研项目(GY-Z20066)。
关键词
锂离子电池
电动汽车
门循环单元
边缘计算
电池健康度
lithium-ion batteries
electric vehicles
gated recurrent unit(GRU)
edge computing
battery health
