An improved multi innovation adaptive robust dual kalman filter algorithm for estimating battery state
An improved multi innovation adaptive robust dual kalman filter algorithm for estimating battery state
复制标题
DOI：
10.1007/s11581-023-05314-2
复制
发表时间：
2023-12
期刊：
影响因子：
2.8
通讯作者：
Zhe Guan;Fa Zhi Yang;Tao hua Yu;Aimin An
中科院分区：
文献类型：
--
作者：
Zhe Guan;Fa Zhi Yang;Tao hua Yu;Aimin An
针对当前动力电池管理系统中对荷电状态（State of Charge，SOC）估计不准确的问题，并考虑到单独估计SOC时可能受到健康状态（State of Health，SOH）的偏移约束。提出了一种自适应鲁棒卡尔曼滤波与多新息理论相结合的对偶卡尔曼滤波参数在线辨识方法。这种控制方法是基于自适应鲁棒卡尔曼控制。它在不同的时间使用多个新息值和卡尔曼增益来校正估计值。双卡尔曼滤波用于在线参数辨识和电池健康状态的联合估计，增加了误差信息量，为SOC和SOH的精确估计提供了优化方法。为验证算法的合理性，采用二阶RC等效电路模型表征电池的动态特性，并在不同工况下进行了实验验证。实验结果表明，在UDDS、FUDS和US06三种工况下，SOC的平均误差分别为0.56%、0.31%和1.23%。稳定后的SOH估计误差小于1.73%。在五种估计算法中，该算法的估计误差最小。该算法具有良好的精度和收敛性。多新息自适应鲁棒对偶卡尔曼滤波算法为锂电池的精确状态估计和广泛应用提供了理论基础。
In response to the inaccurate estimation of State of Charge (SOC) in current power battery management systems, and considering that SOC may be subject to offset constraints from State of Health (SOH) when estimated separately. A method combining adaptive robust Kalman filtering with multiple innovation theories and online identification of dual Kalman filtering parameters is proposed. This control method is based on adaptive robust Kalman control. It corrects the estimated values using multiple innovation values and Kalman gains at different times. Dual Kalman filtering is used for online parameter identification and joint estimation of battery health status, which increases the amount of error information and provides an optimized method for accurate estimation of SOC and SOH. To verify the rationality of the algorithm, a second-order RC equivalent circuit model is used to characterize the dynamic characteristics of the battery, and experimental verification is carried out under different operating conditions. The experimental results show that the average error of SOC under three operating conditions: UDDS, FUDS, and US06 is 0.56%, 0.31%, and 1.23%, respectively. The estimated error of SOH after stabilization is less than 1.73%. The estimation error is the lowest among the five estimation algorithms. The proposed algorithm has been verified to have good accuracy and convergence. The multi-innovation adaptive robust dual Kalman filtering algorithm provides a theoretical basis for accurate state estimation and widespread application of lithium batteries.