文档名：基于片段充电数据和DEKFWNNWLSTM的锂电池健康状态实时估计
摘要：实时准确地评估电动汽车锂电池健康状态(SOH)对电动汽车的稳定行驶至关重要.因此,该文提出一种基于锂电池日常片段充电数据和双扩展卡尔曼滤波-小波神经网络-小波长短时记忆神经网络(DEKF-WNN-WLSTM)的电池全充时间估计模型,进而提高了片段充电数据评估电池健康状态的准确度.首先,设计双扩展卡尔曼滤波预测-校正算法,分别用来估计片段充电数据对应的全充时间和校正扩展卡尔曼滤波的状态初值,以提高估计的准确性.然后,设计了小波神经网络-小波长短时神经网络来学习扩展卡尔曼滤波递推过程的观测值.最后,通过实验仿真,验证了所提算法在锂电池健康状态实时估算中的准确性和有效性.
Abstract：Asacleantechnologytosolvecarbonemissions,electricvehicleshavebeenwidelyusedinmodernvehicles.Duetoitshighenergydensity,lightweight,longlifeandlowselfdischarge,lithium-ionbatterieshavebecomethemainenergystorageequipmentofelectricvehicles.Realtimeandaccurateevaluationofthestateofhealth(SOH)ofthelithiumbatteriesiscriticaltothestabledrivingofelectricvehicles.However,mosttraditionalSOHforecastmethodsareoffline,whichmakesitdifficulttoobtaintheSOHofthebatteriesinrealtime.Recently,somemethodswerepresentedtoforecasttheSOHoflithium-ionbatteries,butmostofthemsufferedfrominconvenientadjustmentofbatterymodelparametersandaccumulationoferrors.Toaddresstheseissues,thispaperproposesabatteryfullchargingtimeestimationmodelanddualextendedKalmanfilters-waveletneuralnetwork-waveletlongshort-termmemoryneuralnetwork(DEKF-WNN-WLSTM).Bytakingthedailysegmentchargingdataoflithiumbatteriesasinput,topredictthefulltimechargingofthebattery,andthengettheSOHinrealtime.Firstly,basedonthestrongrobustnessofwaveletneuralnetwork(WNN)andtheabilityoflongshorttermmemory(LSTM)toextractthetimeseriesfeaturesofthedata,theneuralnetworkofWNN-WLSTMisdesigned.Secondly,twoWNN-WLSTMnetworksaretrainedwithonefullchargingdataandthreefragmentdataoflithiumbatteries,respectively.Thirdly,areal-timeestimationalgorithmnamedDEKFisconstructed,inwhichthefirstEKFisusedtoestimatethefullchargingtimecorrespondingtothesegmentdata,andthesecondEKFisusedtopredicttheerrorbetweentheestimatedandmeasuredbatteryfullchargingtimeunderthecurrentcycle.ThenthetwotrainednetworksareintegratedintoDEKFtoprovidecorrespondingoutputvaluesforthecyclicrecursionofEKF.Finally,areal-timeSOHestimationmodelbasedondailysegmentchargingdataisdesigned.ThesegmentdatafromconstantcurrentchargingtofullchargingatanytimeisusedastheinputofDEKF-WNN-WLSTM,toestimatethecurrentfullchargingtimeoflithiumbatteries,thencalculatetheSOHofthebatteryatthecurrenttime.Inthisreal-timemodel,theWNN-WLSTMalleviatestheinconvenientadjustmentofbatterymodelparametersproblem,addressesthelong-termdependenceproblem.TheDEKFusesthedailysegmentchargingdataastheinput,whichextendsthepracticalapplicationofthemodel.Simulationresultsontheactualbatterycharginganddischargingdatashowthat,themeanrelativeerrorofthepredictionsfortheentire80cyclesis0.0101,theestimatederrorforthefirst50cyclesiscompletelylessthan2%,andlessthan1%atmosttimes.ThecomparisonbetweenDEKF-WNN-WLSTMandextendedKalmanfilterandGaussianprocessregression(EKF-GPR)showsthat,themeanrelativeerrorofEKF-GPRis0.0176,whichishigherthanDEKF-WNN-WLSTM,especiallyinthe170～180cycles,whichindicatesthatthemodelofDEKF-WNN-WLSTMcanalleviatecertainerrorgrowthwiththeincreaseofcycles.Theproposedmethodhasabetterestimationeffectundertheconditionthatnoartificialfullrechargeoperationisperformedtoupdatetheinitialfullchargingtimevalue.Thefollowingconclusionscanbedrawnfromthesimulationanalysis:(1)TheproposedmethodintegratesWNN-WLSTMneuralnetwork,whichaddresstheproblemsoflong-termdependenceandtheinconvenientadjustmentofbatterymodelparameters.(2)ComparedwithEKF-GPR,theDEKF-WNN-WLSTMnotonlyimprovesthepredictionaccuracy,butalsoalleviatestheerroraccumulation.(3)Theproposedmodelonlyneedsthedailysegmentchargingdata.Inthissense,itispracticalintherealworld.
作者：宋显华 姚全正Author：SongXianhua YaoQuanzheng
作者单位：哈尔滨理工大学理学院哈尔滨150080
刊名：电工技术学报
Journal：TransactionsofChinaElectrotechnicalSociety
年，卷(期)：2024, 39(5)
分类号：TM911
关键词：电池健康状态 片段数据 双扩展卡尔曼滤波 小波神经网络 小波长短时记忆神经网络
Keywords：Stateofhealth segmentdata dualextendedKalmanfilter waveletneuralnetwork waveletlongshort-termmemory
机标分类号：TP274~+.2TP391U445.1
在线出版日期：2024年3月19日
基金项目：黑龙江省自然科学基金联合引导项目，山东省自然科学基金联合基金培育项目基于片段充电数据和DEKF-WNN-WLSTM的锂电池健康状态实时估计[
期刊论文] 电工技术学报--2024, 39(5)宋显华 姚全正实时准确地评估电动汽车锂电池健康状态(SOH)对电动汽车的稳定行驶至关重要.因此,该文提出一种基于锂电池日常片段充电数据和双扩展卡尔曼滤波-小波神经网络-小波长短时记忆神经网络(DEKF-WNN-WLSTM)的电池全充时间估计...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
基于片段充电数据和DEKF-WNN-WLSTM的锂电池健康状态实时估计 Real-Time State of Health Estimation for Lithium-Ion Batteries Based on Daily Segment Charging Data and Dual Extended Kalman Filters-Wavelet Neural Network-Wavelet Long Short-Term Memory Neural Network
基于片段充电数据和DEKF-WNN-WLSTM的锂电池健康状态实时估计.pdf
- 文件大小:
- 1.03 MB
- 下载次数:
- 60
-
高速下载
|
|
