Adaptive estimation of state of charge and capacity with online identified battery model for vanadium redox flow battery
Adaptive estimation of state of charge and capacity with online identified battery model for vanadium redox flow battery
复制标题
DOI：
10.1016/j.jpowsour.2016.09.123
复制
发表时间：
2016-11
影响因子：
9.2
通讯作者：
Zhongbao Wei;K. Tseng;N. Wai;T. Lim;M. Skyllas-Kazacos
中科院分区：
文献类型：
--
作者：
Zhongbao Wei;K. Tseng;N. Wai;T. Lim;M. Skyllas-Kazacos
可靠的状态估计在很大程度上取决于准确的电池模型。然而，电池模型的参数是时变的，随着操作条件的变化和电池老化。现有的联合估计方法通过将在线模型辨识与状态估计相结合来解决模型的不确定性问题，并已显示出更高的精度。然而，交叉干扰可能会出现从集成框架，妥协的数值稳定性和精度。为此，本文提出了模型辨识与状态估计的解耦，以消除交叉干扰的可能性。采用递推最小二乘（RLS）方法在线自适应模型参数，并在此基础上提出了一种基于扩展卡尔曼滤波（EKF）的SOC和容量联合估计方法。所提出的联合估计器有效地压缩了滤波器的阶数，从而大大提高了计算效率和数值稳定性。实验室规模的钒氧化还原液流电池的实验表明，所提出的方法是高度可信的，具有良好的鲁棒性，以不同的操作条件和电池老化。该方法与现有的一些方法进行了比较，并显示出上级的精度，收敛速度和计算成本。
Reliable state estimate depends largely on an accurate battery model. However, the parameters of battery model are time varying with operating condition variation and battery aging. The existing co-estimation methods address the model uncertainty by integrating the online model identification with state estimate and have shown improved accuracy. However, the cross interference may arise from the integrated framework to compromise numerical stability and accuracy. Thus this paper proposes the decoupling of model identification and state estimate to eliminate the possibility of cross interference. The model parameters are online adapted with the recursive least squares (RLS) method, based on which a novel joint estimator based on extended Kalman Filter (EKF) is formulated to estimate the state of charge (SOC) and capacity concurrently. The proposed joint estimator effectively compresses the filter order which leads to substantial improvement in the computational efficiency and numerical stability. Lab scale experiment on vanadium redox flow battery shows that the proposed method is highly authentic with good robustness to varying operating conditions and battery aging. The proposed method is further compared with some existing methods and shown to be superior in terms of accuracy, convergence speed, and computational cost.