AR模型参数的抗差递推估计  

Robustified Recursive Estimation of Auto-Regressive(AR) Model Parameter

作　　者：王玉丽[1] 包为民[1] 沈丹丹[1] 周俊伟[1] 孙逸群[1] 

机构地区：[1]河海大学水文水资源学院,南京210098

出　　处：《中国农村水利水电》2017年第6期74-77,共4页China Rural Water and Hydropower

基　　金：国家自然科学基金面上基金项目(51279057);水利部公益性行业科研专项项目(201501059)

摘　　要：将抗差理论引入到自回归模型(AR)的参数估计中,在实时洪水预报的误差修正中,可以利用抗差理论估计AR模型的参数,抵御异常观测值对参数估计的影响,以保证实时洪水的预报精度。选取闽江水吉流域的5场实测洪水资料,通过递推最小二乘法和抗差递推最小二乘法来率定AR模型参数,结果表明:当观测资料无异常值时,递推最小二乘法和抗差递推最小二乘的结果相近,都比较满意;当观测资料中有异常值时,递推最小二乘的结果明显不合理,抗差递推最小二乘法能获得最优估值。The robust theory is introduced into the Auto-Regressive model （AR） parameter estimation. In the real-time correction of the floods, the robust theory is used to estimate the parameter of Auto-Regressive （ AR）, withstand the observations of interference in outliers. Thus the accuracy of real-time flood forecasting is guaranteed.Selecting the 5 floods in Shuiji Basin, calibrating the AR model parameter by the method of the recursive least-squares and robust recursive least-squares. The calculation results show that the result of recursive least square method and robust recursive least squares are similar when there are no outliers. And when the observations have outliers, robust least square results are better than the least squares method.

关 键 词：AR模型 实时洪水预报 水吉流域 递推最小二乘法 抗差递推最小二乘法 

分 类 号：TV122[水利工程—水文学及水资源]