钠离子电池健康状态预测  

State of health prediction for sodium-ion batteries

作　　者：冯一峰 沈佳妮[1] 车海英[1,2] 马紫峰 贺益君[1] 谈文 杨庆亨 FENG Yifeng;SHEN Jiani;CHE Haiying;MA Zifeng;HE Yijun;TAN Wen;YANG Qingheng(Department of Chemical Engineering,Shanghai Electrochemical Energy Devices Research Center,Shanghai Jiao Tong University,Shanghai 200240,China;Zhejiang NaTRIUM Energy Co.Ltd.,Shaoxing 312000,Zhejiang,China;Shanghai Pylontech Energy Technology Co.,Ltd.,Shanghai 200240,China)

机构地区：[1]上海交通大学化学化工学院,上海电化学能源器件工程技术中心,上海200240 [2]浙江钠创新能源有限公司,浙江绍兴312000 [3]上海派能能源科技股份有限公司,上海201203

出　　处：《储能科学与技术》2021年第4期1407-1415,共9页Energy Storage Science and Technology

基　　金：国家重点研发计划(2016YFB0901500);国家自然科学基金(21938005)项目。

摘　　要：钠离子电池健康状态(SOH)预测对于电池优化管理有重要意义,但由于钠离子电池老化机理复杂,影响因素众多,精准SOH预测挑战巨大。为此,本研究从健康状态时序测量数据出发,提出了基于双指数模型的粒子滤波法(DEM-PF)和基于小波分析的高斯过程回归法(WA-GPR),以实现钠离子电池单步SOH和剩余可用寿命(RUL)预测。前者直接采用双指数函数构建时序SOH数据模型,并结合PF算法进行模型参数更新;后者采用小波分析实现时序SOH数据多尺度解耦,采用GPR构建各尺度数据模型并进行融合后实施预测。实验结果表明,相比DEM-PF方法,WA-GPR方法的单步SOH和RUL预测效果更好,单步SOH预测均方根误差为0.8%,RUL预测误差最小为3次循环,从而为钠离子电池管理提供有效支撑。Sodium-ion batteries(SIBs)show promising application prospect in large-scale energy storage,due to the abundant and low-cost sodium resources.Most of the research focuses on the development of new SIB materials such as electrodes and electrolytes.Engineering manufacturing technologies and demonstration applications are still in the exploration stage.To ensure high safety,long life,and high efficiency operation,the battery management systems(BMSs)based on the characteristics of SIBs need to be developed.Accurate state of health(SOH)prediction is one of the core functions of BMS,and single-step ahead and multi-step-ahead SOH prediction are important for the estimation of state of charge and the prediction of remaining useful life(RUL),respectively.Compared to lithium-ion batteries,SIBs have similar operation mechanism,but the larger sodium ions result in more complicated battery characteristics and aging mechanism,which may make it difficult for accurate SOH prediction for the full SIBs.Based on the SOH time series,a double exponential model-based Particle filter(DEM-PF)method and a wavelet-analysis-based Gaussian process regression(WA-GPR)method are proposed.In the DEM-PF method,the DEM is utilized to model SOH time series.The PF is used to update the model parameters.In the WA-GPR method,WA is used to decouple the global degradation trend and local capacity regeneration and fluctuations of SOH time series.The GPR with time index input is used to prediction the global degradation trend.The GPR with lag vector input realizes the autoregression of the local capacity regeneration and fluctuation.The two methods are validated and compared in the 1 C charge/discharge aging test of a 1 A·h pouch-type SIB.The results indicates that the WA-GPR method shows better accuracy and stability both in the one-step-ahead SOH and RUL prediction,with the prediction root mean square error of 0.8%for one-step-ahead SOH and minimum error of 3 cycles for RUL.

关 键 词：钠离子电池 健康状态 粒子滤波算法 高斯过程回归 小波分析 

分 类 号：TM912[电气工程—电力电子与电力传动]