基于中低频阻抗谱的锂离子电池容量快速估计方法

A Rapid Capacity Estimation Method for Lithium-ion Batteries Based on Medium-low Frequency Electrochemical Impedance Spectroscopy

针对批量电动汽车退役电池的梯次利用,电池评估专用设备的技术研发面临迫切需求,但评估软件的算法策略开发难度大。兼顾快速性与准确性的阻抗谱方法在专用设备实现中具有良好的可行性。因此选择测试时间短且对采样频率要求低的中低频阻抗谱,在硬件实现中既节约了高频采样成本,又避免了低频正弦难以精准实现的难题。通过拟合电荷转移阻抗特征圆,提取顶点虚部、拐点虚部、圆心横坐标、圆与实轴的交点及其模态分解残值5个健康特征,运用皮尔逊相关系数验证健康特征与容量的相关性,利用高斯过程回归指数模型进行模型训练并验证,实现了锂离子电池容量的快速估计。首先应用实验室测试数据进行方法验证,试验值均在估计值的95%置信区间内;然后应用公开数据集进一步验证,该方法建立估计模型决定系数R2为0.92,估计结果的均方根误差为0.490 8,平均绝对百分比误差为1.343 1%。此外,分别选取减少阻抗谱拟合圆数据点、选取中低频阻抗谱顶点和拐点阻抗、提取实轴以上全频段固定频率点阻抗3种方法,对比验证了所提方法的精度优势和有效性。结果表明:通过拟合特征圆提取关键参数,融合拐点和顶点特征,在保证较高精度的条件下,能够实现电池容量的快速估计。

For the stepwise utilization of retired batteries in bulk electric vehicles,the technological development of specialized equipment for battery evaluation is in urgent demand,but the development of algorithmic strategies for evaluation software is difficult.The electrochemical impedance spectroscopy(EIS)method that balances speed and accuracy has good feasibility in the implementation of specialized equipment.This paper selects the medium-low frequency EIS that require short testing time and low sampling frequency,which not only saves the cost of high-frequency sampling but also avoids the problem of low-frequency sinusoidal difficult to achieve accurately in the hardware implementation.By fitting the characteristic circle of charge transfer impedance,five health characteristics were extracted,the imaginary part of the vertex,the imaginary part of the inflection point,the abscissa of the circle center,the intersection of the circle with real axis,and its modal decomposition residual value.Pearson correlation coefficient was used to verify the correlation between health characteristics and capacity,and Gaussian process regression index model was used to train and verify the model,realization of rapid estimation of lithium-ion battery capacity.Firstly,the laboratory test data are applied to validate the method,and the experimental values are all within the 95%confidence interval of the estimated values;then the public dataset is applied to further validate the method,which establishes the estimation model with a coefficient of determination of R~2 of 0.92,and the estimation results in an RMSE of 0.4908,and a MAPE of 1.3431%.In addition,three methods were selected to reduce the EIS fitting circle data points,to select the impedance at the top and inflection points of the medium-low frequency EIS,to extract the impedance at fixed frequency points in the full frequency band above the real axis,comparing and verifying the accuracy advantages and efficiency of the proposed methods.The results show that by fitting feature circles to extract key parameters as well as fusing inflection point and vertex features,rapid estimation of battery capacity can be achieved while ensuring high accuracy.