【作者】 李娟；

【导师】 陈元芳； 王栋；

【作者基本信息】 河海大学 ， 水文学及水资源， 2006， 硕士

【摘要】 水文频率分析是给各种水利工程提供具有概率含义的水文设计数据,以确定工程的规模、投资和效益。本论文主要探讨最大熵原理在P-Ⅲ型分布参数估计中的应用。 论文在比较系统的总结国内外水文频率计算研究进展的基础上,对熵及最大熵原理进行了阐述,总结了最大熵原理在水文频率分析中的应用现状。在此基础上提出一种基于线性矩法的熵估计方法(POMELM)。考虑到过去的熵估计方法(V.P.Singh教授提出的熵估计方法POMEMOM和熵适线法POMEFIT)仅考虑了简单样本,故本论文中还提出考虑历史洪水的熵估计公式。进行统计试验比较了三种熵估计方法,结果表明本论文建议的POMELM方法要优于POMEMOM和POMEFIT两种方法,所提出的考虑历史洪水的熵估计公式是有效可行的。为了进一步了解熵估计方法的统计特性,还与传统参数估计方法(矩法、权函数法、线性矩法、适线法)作了比较。统计试验结果表明:POMELM方法明显好于矩法,比权函数法和适线法略好,与线性矩法相当,但在多数统计试验方案中,其有效性比线性矩法略好一些。更多还原

【Abstract】 It’s necessary to provide different hydrologic design values for various kinds of hydraulic works through hydrologic frequency analysis in order to determine the scale, investment and benefit of the project. The main purpose of the thesis is to study parameter estimation method for Pearson-III distribution based on the principle of maximum entropy (POME).On the basis of the systematic summary of the study progress in hydrologic frequency analysis in China and abroad, and brief introduction of current application situation of the principle of maximum entropy in hydrologic frequency analysis. A new parameter estimation method (POMELM) based on POME and L-moment is proposed in the thesis. Considering that the actual two parameter estimation methods (POMEMOM and POMEFIT) may not utilize historical flood information, so a series of formula with considering historical flood information for POMELM. POMEMOM and POMEFIT are also proposed. Through the comparison by Monte-Carlo experiment, the results show that the new method POMELM is much better than POMEMOM and POMEFIT both in bias or effectiveness of quantiles and parameters. It also shows that the formula considering historical flood proposed are effective and feasible by Monte-Carlo method. In order to understand further the statistic performance of the new method POMELM, the comparison with conventional estimation methods, such as method of Moments, method of Weighted Function Moments, method of L-Moments and Curve-Fitting method has been done by Monte-Carlo experiment too. The results show that POMELM is obvious match better than Method of Moments, and it’s almost equal to the method of L-Moments, but in majority of experiment schemes, it leans toward to be a little bit better than method of L-Moments, especially in the effectiveness of quantiles.更多还原