基于变窗口自适应无迹卡尔曼滤波的锂离子电池荷电状态预测  

Lithium-Ion Battery State of Charge Estimation Based on Variable-Window Adaptive Untraceable Kalman Filtering Algorithm

作　　者：范兴明[1] 吴润玮 封浩 张鑫[1] Fan Xingming;Wu Runwei;Feng Hao;Zhang Xin(School of Mechanical and Electrical Engineering Guilin University of Electronic and Technology,Guilin 541004 China)

机构地区：[1]桂林电子科技大学机电工程学院,桂林541004

出　　处：《电工技术学报》2025年第6期1974-1983,共10页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(61741126);广西自然科学基金(2022GXNSFAA035533)资助项目。

摘　　要：基于噪声协方差匹配方法的自适应无迹卡尔曼滤波(AUKF)算法,其固定长度的时间窗影响算法噪声统计量。且AUKF中匹配窗口长度常由经验法确定,在复杂工作条件下容易引起噪声协方差估算的不确定。为了进一步提高算法的噪声协方差估算精度,提出一种由自适应遗传算法(AGA)确定初始窗口长度的变窗口自适应无迹卡尔曼滤波(VAUKF)。并引入Allan方差分析法识别误差序列的波动,再基于更新规则适当调整窗口长度,提高VAUKF对时变噪声的抗干扰能力。通过FUDS、US06工况验证所提出的VAUKF算法可行性。仿真结果表明,VAUKF相比AUKF在不同工况下都实现了荷电状态(SOC)预测精度和鲁棒性的提高。In the lithium-battery charge state prediction,the Kalman filter algorithm is independent of a large number of dataset training.It can predict the state quantities with the observed data to obtain the optimal estimation of the system in the form of extended Kalman(extended Kalman filter(EKF)),untraceable Kalman filter(UKF),and other extended forms.However,the Kalman filtering algorithm and its extended forms for lithium battery nonlinear time-varying system obtain a fixed noise covariance,easily leading to the prediction error.Therefore,this paper proposes a variable-window adaptive untraceable Kalman(VAUKF)to determine the adaptive untraceable Kalman time window length,avoiding the prediction error caused by improper window length selection.The adaptive genetic algorithm(AGA)has been proven to achieve good parameter computation ability in avoiding local optimization and convergence speed problems.Thus,AGA calculates the optimal time window length,the overlapping grouped Allan ANOVA identifies the error sequence fluctuation,and the iterative process adjusts the window length appropriately.The VAUKF improves the SOC’s prediction accuracy and robustness compared to the AUKF.First,based on the second-order RC lithium-ion battery equivalent circuit model,simulation modeling is carried out under the Federal Urban Driving Schedule(FUDS)and US06 high-speed cycling condition(US06)data.The noise level of the prediction process is obtained through the Allan variance,the variable window adjustment rule is determined,and the impact of noise fluctuations on prediction performance is analyzed.Then,the SOC prediction performance of VAUKF under different multiplicities is explored for the noise-matching window update rule,which provides more reasonable parameter conditions for VAUKF.Finally,the tracking ability and convergence speed of VAUKF and AUFK under different working conditions are analyzed,and the simulation results are discussed.Compared with AUKF,the VAUKF decreases MAE by 25.3%and RMSE by 24.4%in RMSE under the FUDS

关 键 词：SOC预测 自适应无迹卡尔曼 变窗口自适应无迹卡尔曼 

分 类 号：TM912[电气工程—电力电子与电力传动]