摘要
精准的锂电池建模是保证电池储能系统可靠性至关重要的手段。荷电状态(state of charge,SOC)的准确估计保证了特定应用程序的安全高效运行。为了提高SOC的估计精度,首先建立等效电路模型,利用遗忘因子的偏差补偿最小二乘法(bias compensation recursive least squares,BCRLS)对电池模型进行参数辨识。然后,利用自适应无迹卡尔曼滤波(adaptive unscented Kalman filter,AUKF)算法来估计SOC。由于无迹无迹卡尔曼滤波算法易受非线性因素的干扰,因此提出了利用权重量定义AUKF算法提高SOC的估计精度。由于电池在放电过程中,电池内部特性会发生变化,而电池欧姆内阻会对SOC估计结果产生直接影响。基于此,本工作提出了双自适应无迹卡尔曼滤波来进一步提高SOC的估计精度。通过和不同算法进行比较,实验结果表明,所提算法估计SOC的误差控制在2%以内,验证了算法的有效性。
An accurate lithium-ion battery model is very important to ensure the reliability of the battery. Accurate estimation of state of charge(SOC) ensures the safety and efficient operation of specific applications. To improve the estimation accuracy of SOC, an equivalent circuit model is established and the parameters are identified using bias compensation recursive least squares(BCRLS) of the forgetting factor. The SOC is then estimated using the adaptive unscented Kalman filter(AUKF) algorithm. The AUKF algorithm defined by weight vectors was proposed to improve the estimation accuracy of SOC because of the vulnerability of the unscented Kalman filter technique to nonlinear variables. However, the internal characteristics of the battery will change during the discharge process, and the ohmic internal resistance of the battery will have a direct effect on the SOC estimations. Based on this, we propose a dual AUKF to further improve the estimation accuracy of SOC. Compared with other algorithms, the experimental results show that the proposed algorithm’s error in estimating SOC is less than 2%, demonstrating the effectiveness of the algorithm.
作者
黄鹏超
鄂加强
HUANG Pengchao;E Jiaqiang(Liuzhou Vocational Technical College School of Automotive Engineering,Liuzhou 545616,Guangxi,China;College of Mechanical and Vehicle Engineering,Hunan University,Changsha 410082,Hunan,China)
出处
《储能科学与技术》
CAS
CSCD
北大核心
2022年第2期660-666,共7页
Energy Storage Science and Technology
基金
2019年广西高校中青年项目:电动汽车动力电池均衡控制技术研究(2019KY1257)。
关键词
锂离子电池
荷电状态
偏差补偿最小二乘法
权重向量
双自适应无迹卡尔曼滤波
lithium ion battery
state of charge(SOC)
bias compensation recursive least squares
weight vectors
dual adaptive unscented Kalman filter
