Identification of lithium-ion battery equivalent circuit model parameters based on the multi-innovation identification algorithm
LIN Peng,1, LIU Tao2, JIN Peng,3,4,5, WANG Zhenpo3, WANG Shengjie1, YUAN Hongsheng1, MA Ze1, DI Yu1
1.Beijing Mechanical Equipment Institut, Beijing 100854, China
2.Vehicle Research Institutee, Beijing 100024, China
3.National Engineering Research Center of Electric Vehicles, School of Mechatronical Engineering, Beijing Institute of Technology, Beijing 100081, China
4.School of Electrical and Control Engineering, North China University of Technology
5.Collaborative Innovation Center of Electric Vehicle in Beijing, Beijing 100144, China
实时、准确地获得电池模型的参数可提高电池状态估计的精度。常用的系统辨识算法和智能优化算法不仅实时性差,而且辨识精度低。为了解决等效电路模型的参数辨识及提高等效电路模型参数的辨识精度,本文通过直接离散的方法建立了能够同时辨识二阶RC(resistance-capacitance)等效电路模型和PNGV(partnership for a new generation of vehicles)模型参数的差分方程。基于多新息算法辨识理论,提出了带遗忘因子的多新息辅助模型扩展递推最小二乘(FMIAELS)算法。FMIAELS算法只需利用电池的电流及端电压即可实现等效电路模型参数的实时、精确辨识。实验验证结果表明,在不同温度、工况和老化程度下,FMIAELS算法可精确地辨识电池的模型参数,误差约为常用的系统辨识算法和智能优化算法的1/3。FMIAELS算法也能实现开路电压(OCV)的精确辨识,在不同脉冲下辨识的OCV的精度也明显优于常用的系统辨识算法和智能优化算法,其平均误差仅有0.22%。
关键词:等效电路模型
;
模型参数辨识
;
多新息辨识算法
;
锂离子电池
Abstract
This study aims to obtain the battery model parameters in real time and effectively improve the battery state estimation accuracy. The commonly used system identification and intelligent optimization algorithms have poor real-time performances and low identification accuracies. To address the issue on the equivalent circuit model identification and improve the identification accuracy of the equivalent circuit model parameters, this study establishes a difference equation that identifies the parameters of the second-order resistance-capacitance equivalent circuit model and the Partnership for a New Generation of Vehicles model through a direct discretization method. A multi-innovation auxiliary model extended recursive least squares algorithm with a forgetting factor (FMIAELS) is proposed based on the identification theory of the multi-information algorithm. The FMIAELS algorithm realizes a real-time and accurate identification of the equivalent circuit model parameters by using only the current and the terminal voltage of a battery. The experimental verification results demonstrate that the FMIAELS algorithm accurately identifies the battery model parameters under different temperatures, working conditions, and states of health. The error is about 1/3 that of the common system identification and intelligent optimization algorithms. Moreover, the FMIAELS algorithm accurately identifies the open-circuit voltage (OCV). Under various working conditions, its OCV identification accuracy is significantly better than that of the common system identification and intelligent optimization algorithms, yielding only a 0.22% average error.
Keywords:equivalent circuit model
;
model parameter identification
;
multi-innovation identification algorithm
;
lithium ion battery
LIN Peng. Identification of lithium-ion battery equivalent circuit model parameters based on the multi-innovation identification algorithm[J]. Energy Storage Science and Technology, 2023, 12(10): 3155-3169
常用的电池模型分为经验模型、电化学模型和等效电路模型。其中经验模型只能反映电压与荷电状态(state of charge,SOC)、电流之间的近似关系,无法准确描述电池的瞬态特性,应用范围较小。而电化学模型虽能精确地描述电池的静动态特性,但其结构复杂、求解困难,这限制了它的应用场景。RC(resistance-capacitance)等效电路模型能够较好地描述电池的静动态特性,在电池仿真、分析和估计电池状态等领域应用广泛[11-12]。根据RC环节的个数,RC等效电路模型主要分为:零阶RC等效电路模型、一阶RC等效电路模型、二阶RC等效电路模型和多阶RC等效电路模型。二阶模型包含二阶RC等效电路模型和PNGV(partnership for a new generation of vehicles)模型。虽然二阶RC等效电路模型比Thevenin模型(即一阶RC等效电路模型)多一个RC环节,但二阶RC等效电路模型并不复杂,其精度比Thevenin模型高。PNGV模型能够反映电流的累积效应对开路电压(open-circuit voltage,OCV)的影响,该模型也常应用于电池特性分析和状态估计[13-18]。
由于锂离子动力电池为非线性时变系统,表现在模型参数上就是时变参数。虽然递推增广最小二乘算法(recursive extended least squares algorithm,RELS)具有递推最小二乘算法的所有优势,在有色噪声下可实现模型参数的无偏估计和一致性估计[35],但存在辨识精度低、收敛速度慢等缺点,不能解决数据饱和问题,进而无法实现时变参数的实时跟踪。因此,RELS并不适用于锂离子动力电池的模型参数辨识。
因此,带遗忘因子的多新息辅助模型扩展递推最小二乘算法(multi-innovation auxiliary model extended recursive least squares algorithm with forgetting factor,FMIAELS)可表示为:
详细推导过程见附录。
2.2 常用的系统辨识算法
为了说明FMIAELS在电池模型参数辨识上的优势,利用常用的系统辨识算法与该算法进行对比分析。常用的系统辨识算法有RELS[38]、带遗忘因子的递推最小二乘算法(recursive least squares algorithm with forgetting factor,FRLS)[39-41]、递推随机牛顿梯度校正算法(recursive stochastic newton gradient correction algorithm,RSNA)[42-43]、模型参考自适应系统(model reference adaptive system,MRAS)[44-45]和智能优化算法。
为了评价系统辨识算法的精度,采用平均绝对误差(mean absolute error,MAE)、均方根误差(root mean square error,RMSE)和加权平均绝对百分比误差(weighted mean absolute percentage error,WMAPE)作为性能评价的指标。MAE、RMSE和WMAPE由式(24)~式(26)给出。由这些指标的定义可知,它们的值越小,说明系统辨识算法的辨识效果越好,辨识精度越高。
SUN Z Y, WANG Z P, LIU P, et al. Overview of fault diagnosis in new energy vehicle power battery system[J]. Journal of Mechanical Engineering, 2021, 57(14): 87-104.
WANG Z P, WANG Q S, LIU P, et al. Review on techniques for power battery state of health estimation driven by big data methods[J]. Journal of Mechanical Engineering, 2023, 59(2): 151-168.
WANG F F, FENG X M, ZHAO G J, et al. Identification of retired power lithium-ion batteries of chemical systems by electrochemical impedance spectroscopy[J]. Energy Storage Science and Technology, 2023, 12(2): 609-614.
HU X S, TANG X L. Review of modeling techniques for lithium-ion traction batteries in electric vehicles[J]. Journal of Mechanical Engineering, 2017, 53(16): 20-31.
LIN P, JIN P, ZHANG H Y. Identification and fast measurement method of open-circuit voltage[J]. Journal of the Electrochemical Society, 2023, 170(3): 030525.
WANG C Y, CUI N X, LI C L, et al. State of power prediction based on electro-thermal battery model and multi-parameter constraints for lithium-ion battery[J]. Journal of Mechanical Engineering, 2019, 55(20): 28-35.
LIN P, JIN P, HONG J C, et al. Battery voltage and state of power prediction based on an improved novel polarization voltage model[J]. Energy Reports, 2020, 6: 2299-2308.
ZHU R, DUAN B, WEN F Z, et al. Lithium-ion battery modeling and parameter identification based on decentralized least squares method[J]. Journal of Mechanical Engineering, 2019, 55(20): 85-93.
SHENG J, FU Y M, YU H G. Structure simulation of large soft pack module for energy storage[J]. Energy Storage Science and Technology, 2023, 12(2): 579-584.
LI J H, KLEE BARILLAS J, GUENTHER C, et al. Sequential Monte Carlo filter for state estimation of LiFePO4 batteries based on an online updated model[J]. Journal of Power Sources, 2014, 247: 156-162.
LAI X, GAO W K, ZHENG Y J, et al. A comparative study of global optimization methods for parameter identification of different equivalent circuit models for Li-ion batteries[J]. Electrochimica Acta, 2019, 295: 1057-1066.
WEI X Z, ZHU B, XU W. Internal resistance identification in vehicle power lithium-ion battery and application in lifetime evaluation[C]//2009 International Conference on Measuring Technology and Mechatronics Automation. April 11-12, 2009, Zhangjiajie, China. IEEE, 2009: 388-392.
CASTANO S, GAUCHIA L, VONCILA E, et al. Dynamical modeling procedure of a Li-ion battery pack suitable for real-time applications[J]. Energy Conversion and Management, 2015, 92: 396-405.
YAO S G, LIAO P, XIAO M, et al. Equivalent circuit modeling and simulation of the zinc nickel single flow battery[J]. AIP Advances, 2017, 7(5): 055112.
LIU C Z, LIU W Q, WANG L Y, et al. A new method of modeling and state of charge estimation of the battery[J]. Journal of Power Sources, 2016, 320: 1-12.
YANG Y, TANG T F, QIN D T, et al. PNGV equivalent circuit model and SOC estimation algorithm of lithium batteries for electric vehicle[J]. Journal of System Simulation, 2012, 24(4): 938-942.
ZHANG X, WANG Y J, WU J, et al. A novel method for lithium-ion battery state of energy and state of power estimation based on multi-time-scale filter[J]. Applied Energy, 2018, 216: 442-451.
ZHANG C, ALLAFI W, DINH Q, et al. Online estimation of battery equivalent circuit model parameters and state of charge using decoupled least squares technique[J]. Energy, 2018, 142: 678-688.
SONG Q, MI Y X, LAI W X. A novel variable forgetting factor recursive least square algorithm to improve the anti-interference ability of battery model parameters identification[J]. IEEE Access, 2019, 7: 61548-61557.
HUANG K, GUO Y F, LI Z G. High precision parameter identification of lithium-ion battery model based on feedback particle swarm optimization algorithm[J]. Transactions of China Electrotechnical Society, 2019, 34(S1): 378-387.
LIU F, MA J, SU W X, et al. Model parameter online identification based SOC estimation method[J]. Journal of Northeastern University (Natural Science), 2020, 41(11): 1543-1549.
LIU Y Q, LU J Y, LONG X L, et al. Dynamic identification of model parameters for energy storage batteries[J]. Journal of National University of Defense Technology, 2019, 41(5): 87-92.
LIU Z C, ZHANG Y H. Parameter identification and state of charge estimation of lithium-ion batteries[J]. Energy Storage Science and Technology, 2022, 11(11): 3613-3622.
DU B H, ZHANG Y, WU T Z, et al. An online identification method for equivalent model parameters of aging lithium-ion batteries[J]. Energy Storage Science and Technology, 2021, 10(1): 342-348.
LIU X Y, LI W L, ZHOU A G. PNGV equivalent circuit model and SOC estimation algorithm for lithium battery pack adopted in AGV vehicle[J]. IEEE Access, 2018, 6: 23639-23647.
PAI K J. A reformatory model incorporating PNGV battery and three-terminal-switch models to design and implement feedback compensations of LiFePO4 battery chargers[J]. Electronics, 2019, 8(2): 126.
WANG W Q, SUN Y Q, HE Y, et al. PNGV model based on variable parameter for lithium iron phosphate Li-ion battery[J]. Battery Bimonthly, 2020, 50(1): 40-44.
CHEN Z, WANG Z D, MOU W B, et al. State-of-charge estimation of lead-carbon batteries based on the PNGV model and an adaptive Kalman filter algorithm[J]. Energy Storage Science and Technology, 2023, 12(3): 941-950.
XIAO R X, SHEN J W, LI X Y, et al. Comparisons of modeling and state of charge estimation for lithium-ion battery based on fractional order and integral order methods[J]. Energies, 2016, 9(3): 184.
MENG J W, BOUKHNIFER M, DIALLO D. Comparative study of lithium-ion battery open-circuit-voltage online estimation methods[J]. IET Electrical Systems in Transportation, 2020, 10(2): 162-169.
LIU W L, WANG L F, LIAO C L, et al. Parameters identification method of battery model for electric vehicles under the charging mode[J]. Transactions of China Electrotechnical Society, 2017, 32(11): 198-207.
ABOUDA S E, ELLOUMI M, KOUBAA Y, et al. Design of parametric estimation algorithm for Hammerstein-Wiener mathematical models[C]//2019 19th International Conference on Sciences and Techniques of Automatic Control and Computer Engineering (STA). March 24-26, 2019, Sousse, Tunisia. IEEE, 2019: 371-375.
DING F, LIU P X, LIU G J. Multiinnovation least-squares identification for system modeling[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 2010, 40(3): 767-778.
LIU Y J, XIE L. Performance analysis of RELS identification for non-uniformly sampled mutlirate ARMAX systems[J]. Control Engineering of China, 2009, 16(4): 409-411, 415.
GUO L, LJUNG L, PRIOURET P. Performance analysis of the forgetting factor RLS algorithm[J]. International Journal of Adaptive Control and Signal Processing, 1993, 7(6): 525-537.
CAMPI M. Performance of RLS Identification Algorithms with Forgetting Factor: A -Mixing Approach[J]. Journal of Mathematical Systems Estimation and Control, 1997, 7(1): 29-54.
PALEOLOGU C, BENESTY J, CIOCHINĂ S. A practical variable forgetting factor recursive least-squares algorithm[C]//2014 11th International Symposium on Electronics and Telecommunications (ISETC). November 14-15, 2014, Timisoara, Romania. IEEE, 2015: 1-4.
BERCU B, GODICHON A, PORTIER B. An efficient stochastic Newton algorithm for parameter estimation in logistic regressions[J]. SIAM Journal on Control and Optimization, 2020, 58(1): 348-367.
CAO P F, LUO X L. Auxiliary-model-based stochastic Newton recursive identification for dual-rate system[J]. Control and Decision, 2015, 30(1): 117-124.
LANDAU I. Unbiased recursive identification using model reference adaptive techniques[J]. IEEE Transactions on Automatic Control, 1976, 21(2): 194-202.
RAVI TEJA A V, VERMA V, CHAKRABORTY C. A new formulation of reactive-power-based model reference adaptive system for sensorless induction motor drive[J]. IEEE Transactions on Industrial Electronics, 2015, 62(11): 6797-6808.
QIN A K, HUANG V L, SUGANTHAN P N. Differential evolution algorithm with strategy adaptation for global numerical optimization[J]. IEEE Transactions on Evolutionary Computation, 2009, 13(2): 398-417.
CUI L Z, LI G H, LIN Q Z, et al. Adaptive differential evolution algorithm with novel mutation strategies in multiple sub-populations[J]. Computers & Operations Research, 2016, 67: 155-173.
LIN P, JIN P, ZOU A X, et al. Real-time identification of partnership for a new generation of vehicles battery model parameters based on the model reference adaptive system[J]. International Journal of Energy Research, 2021, 45(6): 9351-9368.
HE H W, XIONG R, GUO H Q, et al. Comparison study on the battery models used for the energy management of batteries in electric vehicles[J]. Energy Conversion and Management, 2012, 64: 113-121.
... 常用的电池模型分为经验模型、电化学模型和等效电路模型.其中经验模型只能反映电压与荷电状态(state of charge,SOC)、电流之间的近似关系,无法准确描述电池的瞬态特性,应用范围较小.而电化学模型虽能精确地描述电池的静动态特性,但其结构复杂、求解困难,这限制了它的应用场景.RC(resistance-capacitance)等效电路模型能够较好地描述电池的静动态特性,在电池仿真、分析和估计电池状态等领域应用广泛[11-12].根据RC环节的个数,RC等效电路模型主要分为:零阶RC等效电路模型、一阶RC等效电路模型、二阶RC等效电路模型和多阶RC等效电路模型.二阶模型包含二阶RC等效电路模型和PNGV(partnership for a new generation of vehicles)模型.虽然二阶RC等效电路模型比Thevenin模型(即一阶RC等效电路模型)多一个RC环节,但二阶RC等效电路模型并不复杂,其精度比Thevenin模型高.PNGV模型能够反映电流的累积效应对开路电压(open-circuit voltage,OCV)的影响,该模型也常应用于电池特性分析和状态估计[13-18]. ...
1
... 常用的电池模型分为经验模型、电化学模型和等效电路模型.其中经验模型只能反映电压与荷电状态(state of charge,SOC)、电流之间的近似关系,无法准确描述电池的瞬态特性,应用范围较小.而电化学模型虽能精确地描述电池的静动态特性,但其结构复杂、求解困难,这限制了它的应用场景.RC(resistance-capacitance)等效电路模型能够较好地描述电池的静动态特性,在电池仿真、分析和估计电池状态等领域应用广泛[11-12].根据RC环节的个数,RC等效电路模型主要分为:零阶RC等效电路模型、一阶RC等效电路模型、二阶RC等效电路模型和多阶RC等效电路模型.二阶模型包含二阶RC等效电路模型和PNGV(partnership for a new generation of vehicles)模型.虽然二阶RC等效电路模型比Thevenin模型(即一阶RC等效电路模型)多一个RC环节,但二阶RC等效电路模型并不复杂,其精度比Thevenin模型高.PNGV模型能够反映电流的累积效应对开路电压(open-circuit voltage,OCV)的影响,该模型也常应用于电池特性分析和状态估计[13-18]. ...
1
... 常用的电池模型分为经验模型、电化学模型和等效电路模型.其中经验模型只能反映电压与荷电状态(state of charge,SOC)、电流之间的近似关系,无法准确描述电池的瞬态特性,应用范围较小.而电化学模型虽能精确地描述电池的静动态特性,但其结构复杂、求解困难,这限制了它的应用场景.RC(resistance-capacitance)等效电路模型能够较好地描述电池的静动态特性,在电池仿真、分析和估计电池状态等领域应用广泛[11-12].根据RC环节的个数,RC等效电路模型主要分为:零阶RC等效电路模型、一阶RC等效电路模型、二阶RC等效电路模型和多阶RC等效电路模型.二阶模型包含二阶RC等效电路模型和PNGV(partnership for a new generation of vehicles)模型.虽然二阶RC等效电路模型比Thevenin模型(即一阶RC等效电路模型)多一个RC环节,但二阶RC等效电路模型并不复杂,其精度比Thevenin模型高.PNGV模型能够反映电流的累积效应对开路电压(open-circuit voltage,OCV)的影响,该模型也常应用于电池特性分析和状态估计[13-18]. ...
... 常用的电池模型分为经验模型、电化学模型和等效电路模型.其中经验模型只能反映电压与荷电状态(state of charge,SOC)、电流之间的近似关系,无法准确描述电池的瞬态特性,应用范围较小.而电化学模型虽能精确地描述电池的静动态特性,但其结构复杂、求解困难,这限制了它的应用场景.RC(resistance-capacitance)等效电路模型能够较好地描述电池的静动态特性,在电池仿真、分析和估计电池状态等领域应用广泛[11-12].根据RC环节的个数,RC等效电路模型主要分为:零阶RC等效电路模型、一阶RC等效电路模型、二阶RC等效电路模型和多阶RC等效电路模型.二阶模型包含二阶RC等效电路模型和PNGV(partnership for a new generation of vehicles)模型.虽然二阶RC等效电路模型比Thevenin模型(即一阶RC等效电路模型)多一个RC环节,但二阶RC等效电路模型并不复杂,其精度比Thevenin模型高.PNGV模型能够反映电流的累积效应对开路电压(open-circuit voltage,OCV)的影响,该模型也常应用于电池特性分析和状态估计[13-18]. ...
1
... 常用的电池模型分为经验模型、电化学模型和等效电路模型.其中经验模型只能反映电压与荷电状态(state of charge,SOC)、电流之间的近似关系,无法准确描述电池的瞬态特性,应用范围较小.而电化学模型虽能精确地描述电池的静动态特性,但其结构复杂、求解困难,这限制了它的应用场景.RC(resistance-capacitance)等效电路模型能够较好地描述电池的静动态特性,在电池仿真、分析和估计电池状态等领域应用广泛[11-12].根据RC环节的个数,RC等效电路模型主要分为:零阶RC等效电路模型、一阶RC等效电路模型、二阶RC等效电路模型和多阶RC等效电路模型.二阶模型包含二阶RC等效电路模型和PNGV(partnership for a new generation of vehicles)模型.虽然二阶RC等效电路模型比Thevenin模型(即一阶RC等效电路模型)多一个RC环节,但二阶RC等效电路模型并不复杂,其精度比Thevenin模型高.PNGV模型能够反映电流的累积效应对开路电压(open-circuit voltage,OCV)的影响,该模型也常应用于电池特性分析和状态估计[13-18]. ...