基于SR-UKF算法的锂电池SOH预测

Lithium battery SOH Prediction based on square root unscented Kalman filter algorithm

传统的扩展卡尔曼滤波(EKF)算法估计精度较低,无迹卡尔曼滤波(UKF)算法不能确保滤波过程中状态误差协方差矩阵非负性,因而易出现滤波发散问题。为此,文中通过平方根无迹卡尔曼滤波(SR-UKF)算法预测锂电池健康状态(SOH)。首先,利用UT变换对系统进行线性化处理;然后,利用状态误差协方差矩阵的平方根替代状态误差协方差矩阵,保证状态误差协方差矩阵非负性;其次,构建二阶RC等效电路模型,根据最小二乘法辨识模型初始参数;最后,将代表SOH的欧姆内阻作为状态变量,使用SR-UKF实时估计欧姆内阻,并根据欧姆内阻与SOH的关系获取锂电池的SOH。为验证SRUKF算法在不同放电情况下的适应性,通过恒流放电工况和HPPC放电工况对SR-UKF算法进行仿真。结果表明,相比于传统的EKF算法、UKF算法,SR-UKF算法预测欧姆内阻的效果更好。

As the traditional extended Kalman filter(EKF) algorithm has low estimation accuracy,and the unscented Kalman filter(UKF)algorithm cannot ensure the non-negativity of the state error covariance matrix during the filtering process and is easy to cause the problem of filtering divergence,the square root unscented Kalman filter(SR-UKF)algorithm is adopted in this paper to predict the state of health(SOH) of the lithium battery. In this method,the UT transformation is used to linearize the system,the square root of the state error covariance matrix is utilized to replace the state error covariance matrix to ensure that the state error covariance matrix is non-negative,a second-order RC equivalent circuit model is construct,and the initial parameters of the model according are identified according to the least square method. The ohmic internal resistance representing SOH is taken as a state variable to estimate the ohmic internal resistance in real time by means of SR-UKF algorithm and obtain the SOH of the lithium battery according to the relationship between the ohmic internal resistance and SOH.In order to verify the adaptability of the SR-UKF algorithm under different discharge conditions,the SR-UKF algorithm was simulated under the constant current discharge conditions and HPPC discharge conditions. The simulation results show that,in comparison with the traditional EKF algorithm and UKF algorithm,the ohmic internal resistance predictive effect of the SR-UKF algorithm is much better.