摘要
锂离子电池健康状态(SOH)描述了电池当前老化程度,其估算难点在于缺乏明确统一的定义、无法直接测量以及难以确定数量合适、相关性高的估算输入量。为了克服上述问题,该文从容量的角度定义SOH,并将锂离子电池恒流-恒压充电过程中的电压、电流、温度曲线作为输入,提出采用一维深度卷积神经网络(CNN)实现锂离子电池容量估算以获取SOH。在NASA锂离子电池随机使用数据集和牛津电池老化数据集上进行的实验结果表明,该方法能够实现准确的SOH估算,且具备网络参数少、占用内存小的优势。另外,通过实验讨论了网络输入、模型结构、数据增强对所提出的SOH估算方法的影响。
State of health(SOH)of Lithium-ion battery describes the current aging degree of the battery.The difficulty of its estimation lies in the lack of a clear definition,the inability to directly measure,and the difficulty in determining the appropriate number and high correlation of the estimation input.In order to overcome the above problems,this article defines SOH from the perspective of capacity,and takes the voltage,current,and temperature curves of the lithium-ion battery constant current-constant voltage charging process as input,and proposes to use a one-dimensional deep convolutional neural network(CNN)to achieve lithium-ion battery capacity estimation to obtain SOH.Experimental results on NASA's lithium-ion battery random use data set and Oxford battery aging data set show that this method can achieve accurate SOH estimation,and has the advantages of fewer network parameters and less memory.In addition,the influences of network input,model structures and data augmentation on the proposed SOH estimation method are discussed through experiments.
作者
李超然
肖飞
樊亚翔
杨国润
唐欣
Li Chaoran;Xiao Fei;Fan Yaxiang;Yang Guorun;Tang Xin(National Key Laboratory of Science and Technology on Vessel Integrated Power System Naval University of Engineering,Wuhan 430033 China)
出处
《电工技术学报》
EI
CSCD
北大核心
2020年第19期4106-4119,共14页
Transactions of China Electrotechnical Society
基金
国家自然科学基金(51907200)
国防科技创新特区资助项目。
关键词
锂离子电池
健康状态
卷积神经网络
深度学习
Lithium-ion battery
state of health(SOH)
convolution neural network
deep learning
