中文摘要：

水库入库流量,通常通过实测库水位、出库流量和库容曲线反推求得,这个过程常存在大的、不服从正态分布的误差,即所谓的＂粗差＂,大大影响实时校正效果。抗差理论提出了误差的污染分布概念,利用等价权构造具有抗御粗差与极值误差影响的目标函数,使得在抗差理论基础上发展起来的估计方法,具有较强的抗粗差与抗极值误差影响的能力。为此,把抗差理论引入实时校正中,提出用光滑曲线拟合入库流量,通过抗差权函数抵御异常值影响,求得一条光滑的＂实测＂入流过程线,据此再做实时校正。利用了抗差系统具有的抗差能力,有效阻止了非正常因素进入系统,保证了洪水预报的精度。

英文摘要：

The input flow of a reservoir, the ;observed input flow, is usually calculated by the observed water stage, out flow stage--storage relation line. In this process, coarse errors, which does not have normal distribution characteristics, are often caused. In the robust estimation theory, the concept of contaminated distribution of error is put forward, through the equivalent weight, the objective function able to resist the effects of coarse errors is constructed. Based on these, the estimation method shows stronger anti-error ability. In this paper, a new robust correction method combined with the robust estimation theory is introduced： Before the real-time correction, the input flow is smoothed through the curve fitting method, the outliers are resisted by the robust weighting function, then a smooth observed input flow which would he used in the real-time correction comes out.