黃夢濤 胡禮芳 張齊波
摘 要:針對鋰離子電池剩余壽命(remaining useful life,RUL)難以精準(zhǔn)預(yù)測的問題,建立單指數(shù)經(jīng)驗(yàn)容量衰退模型,提出能夠有效解決電池非線性問題的平方根求積分卡爾曼濾波(square-root quadrature kalman filtering,SQKF)算法。現(xiàn)有的最優(yōu)估計方法中,求積分卡爾曼濾波(quadrature kalman filtering,QKF)是一種高精度采樣算法。研究發(fā)現(xiàn),QKF的估計誤差易引起非對稱、非正定協(xié)方差的傳播,影響算法穩(wěn)定性。在QKF算法上進(jìn)行平方根擴(kuò)展,并對單變量求積節(jié)點(diǎn)進(jìn)行多維擴(kuò)展,將SQKF算法應(yīng)用于電池容量跟蹤估計;另外,從理論上證明SQKF的穩(wěn)定性。使用NASA公開數(shù)據(jù)集對算法進(jìn)行仿真驗(yàn)證,并與現(xiàn)有的擴(kuò)展卡爾曼濾波、無跡濾波、QKF算法對比。結(jié)果表明,在一定條件下,SQKF的RUL預(yù)測誤差在6%以內(nèi),數(shù)值精度以及數(shù)值穩(wěn)定性有很大提高,并且研究發(fā)現(xiàn)SQKF受鋰離子電池個體差異性的影響較小,文中方法在鋰離子電池RUL預(yù)測的實(shí)際應(yīng)用方面具有參考價值。
關(guān)鍵詞:鋰離子電池;剩余使用壽命;經(jīng)驗(yàn)容量衰退模型;平方根求積分卡爾曼濾波
中圖分類號:TP 391.9
文獻(xiàn)標(biāo)志碼:A
文章編號:1672-9315(2022)05-0994-09
DOI:10.13800/j.cnki.xakjdxxb.2022.0519開放科學(xué)(資源服務(wù))標(biāo)識碼(OSID):
Square-root quadrature Kalman filtering for remaining useful life prediction in lithium-ion battery
HUANG Mengtao,HU Lifang,ZHANG Qibo
(College of Electrical and Control Engineering,Xi’an University of Science and Technology,Xi’an 710054,China)
Abstract:In order to improve the prediction accuracy ?of Remaining Useful Life(RUL)of lithium-ion battery,a single-exponential empirical capacity degradation model is established,and a Square-Root Quadrature Kalman Filtering(SQKF)algorithm is proposed to solve the nonlinear estimation problem of battery.Among the existing optimal estimation methods,Quadrature Kalman Filtering(QKF)is a high-precision sampling algorithm.According to the researches,the estimation error of QKF tends of trigger the propagation of asymmetric and non-positive covariance,which affects the stability of the algorithm.In this paper,the square root extension is carried out on QKF,the multi-dimensional extension of the univariate integral point is accomplished,and SQKF is applied to battery capacity tracking estimation.Meanwhile,the stability of the approach is theoretically proved.The algorithm is simulated and verified using NASA’s public dataset,and compared with Extended Kalman Filtering,Unscented? Filtering,and QKF.The results show that under certain conditions,the prediction error of SQKF on RUL is within 6%,both the numerical accuracy and stability are greatly improved,and SQKF is less affected by the individual differences of lithium-ion batteries.SQKF has reference value in the practical application of RUL prediction for lithium-ion battery.
Key words:lithium-ion battery;remaining useful life;empirical capacity degradation model;square-root quadrature Kalman filtering
0 引 言
近年來,新能源汽車憑借節(jié)能、低排放、經(jīng)濟(jì)等優(yōu)勢,迅速占據(jù)汽車市場。鋰離子電池作為新能源汽車的主要動力源,隨著充放電循環(huán)次數(shù)的增加,電池容量會不同程度的衰退,導(dǎo)致電池使用壽命縮短,電池性能退化。鋰離子電池剩余壽命(remaining useful life,RUL)的精準(zhǔn)預(yù)測,對于電池狀態(tài)估計、預(yù)測性維護(hù)、自主健康管理具有重大意義,對保證新能源汽車安全穩(wěn)定運(yùn)行起到重要作用。
目前,主流的鋰離子電池RUL預(yù)測方法分為
2類。一類是基于數(shù)據(jù)驅(qū)動的預(yù)測方法。使用鋰離子電池全壽命周期的退化實(shí)驗(yàn)數(shù)據(jù),利用機(jī)器學(xué)習(xí)算法運(yùn)算,得出電池的退化狀態(tài)。馬里蘭大學(xué)高級生命周期工程中心學(xué)者ZHANG等人使用彈性均方反向傳播誤差的方法自適應(yīng)優(yōu)化長短期記憶遞歸神經(jīng)網(wǎng)絡(luò),并使用Dropout技術(shù)解決神經(jīng)網(wǎng)絡(luò)模型的過擬合問題。實(shí)驗(yàn)驗(yàn)證,該方法相對于傳統(tǒng)的循環(huán)神經(jīng)網(wǎng)絡(luò)(recurrent neural network,RNN)可以更快速、更準(zhǔn)確地預(yù)測出電池的RUL。徐東輝等提出一種非線性組合預(yù)測模型,將Elman神經(jīng)網(wǎng)絡(luò)和非線性自回歸神經(jīng)網(wǎng)絡(luò)(nonlinear auto-regressive model with exogenous inputs,NARX)預(yù)測得到的電池性能退化特征量,用最小二乘支持向量機(jī)和RBF神經(jīng)網(wǎng)絡(luò)進(jìn)行非線性組合,預(yù)測電池的RUL。實(shí)驗(yàn)結(jié)果表明,2種方法的預(yù)測精度均比單一化的Elman神經(jīng)網(wǎng)絡(luò)和NARX神經(jīng)網(wǎng)絡(luò)精度高。馬里蘭大學(xué)HU等將遞歸神經(jīng)網(wǎng)絡(luò)、支持向量機(jī)、相關(guān)向量機(jī)、貝葉斯線性回歸模型等方法進(jìn)行加權(quán)組合,發(fā)現(xiàn)加權(quán)組合的數(shù)據(jù)驅(qū)動方法預(yù)測精度高于單一的數(shù)據(jù)驅(qū)動預(yù)測方法的精度。
另一類是基于機(jī)理模型的預(yù)測方法。通過建立機(jī)理模型來表征鋰離子電池的容量退化過程,再使用基于Bayesian估計的濾波方法精確估計電池的退化狀態(tài)。同濟(jì)大學(xué)DAI等在Thevenin模型上增加一個RC回路(極化電阻和極化電容并聯(lián)組成一個RC回路)組成等效電路模型,在線辨識模型參數(shù),完成RUL預(yù)測。但在近似過程中,電池內(nèi)部參量之間的一些重要隱含關(guān)系易被忽略,等效電路模型難以完全考慮復(fù)雜的外界條件,模型對電池的動、靜態(tài)特性描述較弱。馬里蘭大學(xué)高級生命周期工程中心HE等使用大量退化實(shí)驗(yàn)數(shù)據(jù),建立雙指數(shù)經(jīng)驗(yàn)容量衰退模型,使用粒子濾波算法辨識模型參數(shù),并不斷更新參數(shù),實(shí)現(xiàn)RUL估計。美國國家航空航天局艾姆斯研究中心的SAHA等通過電池的集中參數(shù)模型建立單指數(shù)經(jīng)驗(yàn)容量衰退模型,使用粒子濾波算法有效地估計RUL。
基于電池經(jīng)驗(yàn)容量衰退模型的預(yù)測方法依據(jù)動力電池參數(shù)之間的關(guān)聯(lián)性建立模型,模型的適用性強(qiáng),易于獲取。文中采用基于電池的經(jīng)驗(yàn)容量衰退模型的方法預(yù)測鋰離子電池的RUL,并設(shè)計平方根求積分卡爾曼濾波(square-root quadrature kalman filtering,SQKF)算法求解模型。
1 鋰離子電池經(jīng)驗(yàn)容量衰退模型建立
美國國家航空航天局(NASA)的一些學(xué)者研究發(fā)現(xiàn),鋰離子電池的電荷轉(zhuǎn)移電阻與電解質(zhì)電阻之和與容量C具有線性關(guān)系。SAHA等據(jù)此提出一個能反映電池容量衰退過程的電池機(jī)理模型。模型考慮庫倫效率對電池衰退的影響,也考慮電池靜置引起的容量再生,模型見式(1)。
C=ηC+βexp(-β/Δt) ???(1)
式中 k為周期索引,代表第k次充放電周期;Ck表示第k個充放電周期的放電容量;η為庫倫效率;Δtk-1為第k-1個充放電周期到第k個充放電周期的休息時間。β1和β2為待確定參數(shù),受溫度、電流倍率等因素影響,文中采用模型狀態(tài)擴(kuò)張的方法將未知參數(shù)辨識問題轉(zhuǎn)為估計問題。為模擬外界干擾對模型的影響,引入系統(tǒng)誤差和測量誤差。采用如下方程描述β1,β2以及Ck的變化。
4 結(jié) 論
1)單指數(shù)經(jīng)驗(yàn)容量衰退模型適合描述鋰離子電池的容量衰退過程。
2)SQKF估計鋰離子電池容量的過程穩(wěn)定。
3)在不同預(yù)測起始點(diǎn)下,SQKF的精準(zhǔn)度均高于EKF,UF,QKF。
4)相比于EKF,UF,QKF,SQKF預(yù)測算法的精度受電池個體差異影響較小,穩(wěn)定性好。
5)基于單指數(shù)經(jīng)驗(yàn)容量衰退模型的SQKF算法為RUL預(yù)測提供了一個良好的選擇,若進(jìn)一步考慮電池壽命和工作環(huán)境溫度的關(guān)系,可以提高算法的適用性和精確性。
參考文獻(xiàn)(References):
[1]馬憲民.電動汽車高頻交流功率分布系統(tǒng)的諧波分析[J].西安科技大學(xué)學(xué)報,2001,21(2):145-147,151.
MA Xianmin.Harmonic analysis of high frequency AC power distribution system in electric vehicle[J].Journal of Xi’an University of Science and Technology,2001,21(2):145-147,151.
[2]YANG J,XIA B,HUANG W,et al.Online state-of-health estimation for lithium-ion batteries using constant-voltage charging current analysis[J].Applied Energy,2018,212(15):1589-1600.
[3]WANG Y,TIAN J,SUN Z,et al.A comprehensive review of battery modeling and state estimation approaches for advanced battery management systems[J].Renewable and Sustainable Energy Reviews,2020(131):110015-110033.
[4]HASIB S A,ISLAM S,CHAKRABORTTY R K,et al.A comprehensive review of available battery datasets,RUL prediction approaches,and advanced battery management[J].IEEE Access,2021(9):86166-86193.
[5]王常虹,董漢城,凌明祥,等.車用鋰離子電池剩余使用壽命預(yù)測方法[J].汽車工程,2015,37(4):476-479.WNG Changhong,DONG Hancheng,LING Mingxiang,et al.Remaining useful life prediction of automotive lithium-ion battery[J].Automotive Engineering,2015,37(4):476-479.
[6]JIN S,SUI X,HUANG X,et al.Overview of machine learning methods for lithium-ion battery remaining useful lifetime prediction[J].Electronics,2021(10):3126-3144.
[7]WEI J,DONG G,CHEN Z.Remaining useful life prediction and state of health diagnosis for lithium-ion batteries using particle filter and support vector regression[J].IEEE Transactions on Industrial Electronics,2018,65(7):5634-5643.
[8]ZHANG Y,XIONG R,HE H,et al.Long short-term memory recurrent neural network for remaining useful life prediction of lithium-ion batteries[J].IEEE Transactions on Vehicular Technology,2018,67(7):5695-5705.
[9]徐東輝.車用鋰離子動力電池剩余壽命非線性組合預(yù)測研究[J].北京師范大學(xué)學(xué)報:自然科學(xué)版,2021,57(5):571-576.XU Donghui.Nonlinear combination prediction of remaining useful life of automotive lithium-ion batteries[J].Journal of Beijing Normal University(Natural Science),2021,57(5):571-576.
[10]HU C,YOUN B D,WANG P,et al.Ensemble of data-driven prognostic algorithms for robust prediction of remaining useful life[J].Reliability Engineering and System Safety,2012(103):120-135.
[11]蔡艷平,陳萬,蘇延召,等.鋰離子電池剩余壽命預(yù)測方法綜述[J].電源技術(shù),2021,45(5):678-682.CAI Yanping,CHEN Wan,SU Yanzhao,et al.Review of remaining useful life prediction for lithium ion batteries[J].Chinese Journal of Power Sources,2021,45(5):678-682.
[12]劉月峰,張公,張晨榮,等.鋰離子電池RUL預(yù)測方法綜述[J].計算機(jī)工程,2020,46(4):11-18.LIU Yuefeng,ZHANG Gong,ZHANG Chenrong,et al.Review of RUL prediction method for lithium-ion batteries[J].Computer Engineering,2020,46(4):11-18.
[13]DAI H,WEI X,SUN Z.A new SOH prediction concept for the power lithium-ion battery used on HEVs[C]//IEEE Vehicle Power and Propulsion Conference.Washington,USA:IEEE Computer Society,2009:1649-1653.
[14]HE W,WILLIARD N,OSTERMAN M,et al.Prognostics of lithium-ion batteries based on dempster-shafer theory and the Bayesian Monte Carlo method[J].Journal of Power Sources,2011,196(23):10314-10321.
[15]SAHA B,GOEBEL K.Modeling Li-ion battery capacity depletion in a particle filtering framework[C]//Annual Conference of the Prognostics and Health Management Society.San Diego,CA,USA:Prognostics and Health Management Society,2009:1-10.[16]
ITO K.Gaussian filters for nonlinear filtering problems[J].IEEE Transactions on Automatic Control,2000,45(5):910-927.
[17]丁勁濤,羅美君,咼曉兵,等.航空鋰離子電池剩余容量及RUL預(yù)測建模[J].電池,2019,49(4):329-333.DING Jintao,LUO Meijun,GUO Xiaobing,et al.Available capacity and RUL prediction model for aviation Li-ion battery[J].Battery,2019,49(4):329-333.
[18]HUANG M,ZHANG Q.Prediction of remaining useful life of lithium-ion ba-ttery based on UKF[C]//China Automation Congress.Shanghai,China:Institute of Electrical and Electronics Engineers Inc.,2020:4502-4506.
[19]冉昌艷,程向紅,王海鵬.基于Gauss-Hermite求積分卡爾曼濾波的SINS非線性初始對準(zhǔn)方法[J].東南大學(xué)學(xué)報(自然科學(xué)版),2014,44(2):266-271.RAN Changyan,CHENG Xianghong,WANG Haipeng.SINS nonlinear initial alignment using Gauss-Hermite quadrature Kalman filter[J].Journal of Southeast University(Natural Science Edition),2014,44(2):266-271.
[20]巫春玲,韓崇昭.平方根求積分卡爾曼濾波器[J].電子學(xué)報,2009,39(12):1826-1839.WU Chunling,HAN Chongzhao.Square-root quadrature Kalman filter[J].Acta Electronica Sinica,2009,39(12):1826-1839.
[21]ARASARATNAM I,HAYKIN S,ELLIOTT R J.Discrete-time nonlinear filtering algorithms using Gauss-Hermite quadrature[J].Proceedings of the IEEE,2007,95(5):953-977.
[22]ZHOU J,JIAN W.An estimation strategy of lithium battery SOC based on Gauss-Hermite filter[J].Applied Mechanics & Materials,2014,492:147-150.
[23]HUANG M,WANG C,LIU B,et al.Quadrature Kalman filter based on state of charge estimation for lithium-ion battery[J].Advances in Mechanical Engineering,2020,12(7):1-10.
[24]XIONG K,ZHANG H,CHAN C.Performance evaluation of UKF-based nonlinear filtering[J].Automatica,2006,42(2):261-270.
[25]BOLE B,KULKARNI C S,DAIGLE M J.Adaptation of an electrochemistry-based Li-ion battery model to account for deterioration observed under randomized use[C]//Conference of the Prognostics & Health Management Society.Fort Worth,TX,USA:Prognostics and Health Management Society,2014:502-510.