文档摘要：锂离子电池老化实验是研究电池老化性能的基本手段,但针对大量电池的老化实验一般很耗时.为了节约时间和测试成本,获得更多电池数据,该文将变分深度嵌入(VaDE)模型与带有梯度惩罚的生成对抗网络(WGANGP)相结合,组成VaDE-WGANGP架构,进而基于该生成模型设计了一种电池老化特性建模与数据生成的方法.该文以一套开放的电池全寿命周期测试数据集为依据展开研究,首先,将电池放电过程中的电压、电流和放电容量这三个外特性作为模型的输入,通过VaDE的编码器将原始数据映射到隐空间,再通过优化获得符合特定规则的分布;然后,通过一定方式对该分布空间进行采样,并将采样所得的隐变量输入解码器中进行数据生成;后续数据测试表明,VaDE-WGANGP在电池外特性数据生成上具有较好的性能,可以实现对电池老化过程中基础外特性的模拟,在数据量不足时也可以为某些数据驱动算法提供有效的扩展数据资源.
Abstract：Lithium-ionbatteryagingtestsareprimarilyusedtostudybatteryagingperformance.However,agingexperimentsforalargenumberofbatteriestaketimeandeffort.Therefore,combinedwithWassersteinGANandgradientpenalty(WGANGP),theVariationaldeepembedding(VaDE)modelisusedtoformaVaDE-WGANGParchitecture.Then,amethodofbatteryagingperformancemodellinganddatagenerationisdesigned.Thispaperaimstogeneratesimulateddataonexternalcharacteristics(suchasbatteryvoltageandcharge/dischargecurrentduringworking).TheVaDE-WGANGPgenerationmodeliseffectiveduringdifferentbatterySOHintervals,whichindicatesthatsimulatedbatterydatacanbeobtainedacrossthewholebatterylifespan(SOH∈[100％,80％])andcoversbatteryagingperformance.Furthermore,batteryexternalcharacteristicsdifferfordifferentbatterycellsevenunderthesameSOHstate.Accordingly,thediversitiesinbatteryexternalcharacteristicsamongdifferentbatterycellsunderthesameSOHstatesareconsidered.Anopen-sourcebatterytestdataset,includingdatafromToyotaResearchInstitute(TRI)inpartnershipwithMITandStanfordandworkingdataonLFP/graphitecells,isused.Firstly,thethreeexternalcharacteristics(voltage,current,anddischargecapacityduringthedischargeprocess)areselectedastheinputofthemodel.ThelatentspacebytheVaDEencodercanbemapped,anditsdistributionwithspecificrulescanbeobtainedthroughoptimization.Then,thisdistributionspaceissampled,andthelatentvariablesobtainedbysamplingareusedasinputintotheVaDEdecoderfordatageneration.SOHestimationtestsshowthattheVaDE-WGANGPhasgoodgenerationperformanceforbatteryexternalcharacteristics,andtheprimarybatteryworkingperformanceissimulatedduringtheagingprocess.Besides,effectiveandextendedbatterydataresourcescanbeprovidedbythismeasurefordata-drivenalgorithms,especiallyincasetheprimitivedataamountisinsufficient.Thefollowingconclusionscanbedrawn.(1)Regardingbatterydataclusteringandbatteryextrinsiccharacteristicssimulationgeneration,theintegratedmodelofVaDE-WGANGPhasbetterperformancethanVAEandVaDE.ThestatisticalcharacteristicsofbatteryextrinsiccharacteristicsindifferentSOHintervalsareaccurate.(2)VaDE-WGANGPcangeneratehigh-qualitysimulationdataofbatteryexternalcharacteristics.TheSOHestimationresultsprovethatthedatageneratedbyVaDE-WGANGPisofhighquality.Thispaperprovidesanovelideaforanalyzingtheexternalcharacteristicsofbatteriesduringaging.Giventheshortageoforiginaldataindata-driventechnology,thisschemecanalsoprovideaneffectivemethodfordataextensionandishelpfulforbatteryscreening.
作者：李弈 张金龙 漆汉宏 魏艳君 张迪Author：LiYi ZhangJinlong QiHanhong WeiYanjun ZhangDi
作者单位：燕山大学电气工程学院秦皇岛066004
刊名：电工技术学报 ISTICEIPKU
Journal：TransactionsofChinaElectrotechnicalSociety
年，卷(期)：2024, 39(13)
分类号：TM911
关键词：锂离子电池 老化特性 生成模型 变分深度嵌入 带有梯度惩罚的生成对抗网络
Keywords：Lithium-ionbattery agingperformance generationmodel variationaldeepembedding WassersteinGANwithgradientpenalty
机标分类号：TP311.52TM912.9U463.33
在线出版日期：2024年7月22日
基金项目：秦皇岛市科学技术研究与发展计划项目基于变分深度嵌入-带有梯度惩罚的生成对抗网络的锂离子电池老化特性建模[
期刊论文] 电工技术学报--2024, 39(13)李弈 张金龙 漆汉宏 魏艳君 张迪锂离子电池老化实验是研究电池老化性能的基本手段,但针对大量电池的老化实验一般很耗时.为了节约时间和测试成本,获得更多电池数据,该文将变分深度嵌入(VaDE)模型与带有梯度惩罚的生成对抗网络(WGANGP)相结合,组成VaDE...参考文献和引证文献
参考文献
引证文献
本文读者也读过
相似文献
相关博文
关键词：锂离子电池,老化特性,生成模型,变分深度嵌入,带有梯度惩罚的生成对抗网络,
- 文件大小:
- 16.65 MB
- 下载次数:
- 60
-
高速下载
|
|
1
