基于SVD-SUKF的水下机器人电池SOC估计

SOC Estimation of Underwater Vehicle Battery Basedon SVD-SUKF

荷电状态(SOC)的准确估计关系到水下机器人的电池使用效率与任务规划。针对传统SOC估计算法存在的准确性、稳定性和鲁棒性不足等问题,提出一种奇异值分解增强的球型无迹卡尔曼滤波(SVD-SUKF)SOC估计算法。建立2阶Thevenin电路模型,并使用遗忘因子递推最小二乘法对模型参数进行在线辨识;在无迹卡尔曼滤波算法的基础上引入球型无迹变换和奇异值分解,避免繁琐的调参过程、减少算法计算量以及解决算法的协方差矩阵非正定问题;采用城市道路循环工况对SVD-SUKF算法进行验证。结果表明:SVD-SUKF算法收敛速度较快,平均绝对值误差为0.006 8、均方根误差为0.005 6,算法相较于扩展卡尔曼滤波和无迹卡尔曼滤波有更高的估计精度、更好的稳定性和更强的鲁棒性。

The accurate estimation of state of charge(SOC)is related to the battery usage efficiency and mission planning of underwater vehicle.Aiming at the lack of accuracy,stability and robustness of traditional SOC estimation algorithms,a singular value decomposition enhanced spherical unscented Kalman filter(SVD-SUKF)SOC estimation algorithm is proposed.A second-order Thevenin circuit model is established,and a forgetting factor recursive least square is used to identify the model parameters online.The spherical unscented transform and singular value decomposition are introduced on the basis of the unscented Kalman filtering algorithm,which avoids the cumbersome parameter tuning process,reduces the computational amount of the algorithm,and solves the problem of non-positive determination of the covariance matrix of the algorithm.And the urban road cycling conditions are used to validate the SVD-SUKF algorithm.The SVD-SUKF algorithm is validated using urban road cycle conditions.The results show that the SVD-SUKF algorithm converges faster,the average absolute value error is 0.0068,the root mean square error is 0.0056,and the algorithm has higher estimation accuracy,better stability and stronger robustness than the extended Kalman filter and the unscented Kalman filter.