基于Sage-Husa的WMI-SARCKF的锂电池SOC估计  

SOC Estimation of Lithium-ion Batteries Based on WMI-SARCKF Using Sage-Husa

作　　者：凌六一[1,2] 张婷 张虎[1] 杨翀 祁靓 LING Liuyi;ZHANG Ting;ZHANG Hu;YANG Chong;QI Liang(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China;School of Artificial Intelligence,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001 [2]安徽理工大学人工智能学院,安徽淮南232001

出　　处：《安徽理工大学学报（自然科学版）》2024年第4期20-28,共9页Journal of Anhui University of Science and Technology:Natural Science

基　　金：安徽省高校自然科学基金资助项目(KJ2019A0106)。

摘　　要：目的为能够精确估计锂电池的荷电状态(SOC),避免容积卡尔曼滤波计算过程中矩阵易负定的问题,提高算法的鲁棒性和精度。方法提出了一种基于Sage-Husa的加权自适应鲁棒容积卡尔曼滤波(WMI-SARCKF)算法,并结合遗忘因子递推最小二乘法(AFFRLS)形成联合算法,实现了模型参数和SOC的交替更新。AFFRLS在线算法能够自适应调节遗忘因子大小,解决了离线参数辨识因难以适应复杂工况而导致辨识精度下降的问题。在传统的容积卡尔曼滤波基础上,使用对角化变化取代Cholesky分解,采用噪声自适应算法来降低观测噪声对SOC估计精度的影响,通过粒子滤波的权重的思想赋予每个新息不同的权重,提高算法收敛速度。为保证输出的残差序列正交,在时间方程和量测方程中引入渐消因子,增强了容积卡尔曼滤波对电池突变状态的跟踪能力。结果将改进的容积卡尔曼算法(Improved-CKF)与离线CKF和AFFRLS-CKF算法分别在DST和FUDS工况下进行对比,改进后算法的SOC估计平均绝对误差(MAE)和均方根误差(RMSE)均明显降低。结论仿真结果表明,改进后的算法具有更高的滤波器稳定性和SOC估计精度,在实际中具有较高的应用价值。Objective To accurately estimate the state of charge(SOC)of lithium batteries,avoid the problem of matrix negative definiteness during cubature Kalman Filtering calculations and improve the robustness of the algorithm.Methods A Weighted Multi-Innovation and Sage-Husa Adaptive Robust CKF(WMI-SARCKF)was proposed and a joint algorithm was formed by combining it with the adaptive forgetting factor recursive least squares algorithm(AFFRLS)method so that the alternating updates of model parameters and state of charge(SOC)were achieved.The AFFRLS online algorithm adjusted adaptively the forgetting factor size,solving the problem of decreased identification accuracy due to the difficulty of offline parameter identification to adapt to complex working conditions.In addition to traditional cubature Kalman Filtering,diagonalization transformation was used instead of Cholesky decomposition,and a noise adaptive algorithm was adopted to reduce the impact of observation noise on SOC estimation accuracy.By adopting the concept of particle filtering weights,different weights were assigned to each new observation,improving the algorithm convergence speed.To ensure the orthogonality of the output residual sequence,a fading factor was introduced in the time equation and measurement equation,enhancing the capacity of the cubature Kalman Filter to track battery sudden state changes.Results When the improved cubature Kalman Filtering algorithm(Improved-CKF)was compared with the offline CKF and AFFRLS-CKF algorithms under DST and FUDS working conditions,the SOC estimation mean absolute error(MAE)and root mean square error(RMSE)of the improved algorithm were significantly reduced.Conclusion The algorithm has higher filter stability and SOC estimation accuracy,with a practical application value.

关 键 词：荷电状态 容积卡尔曼 在线参数辨识 多新息 最小二乘法 

分 类 号：TM912[电气工程—电力电子与电力传动]