水文频率分析的未确知数随机模拟模型  

Unascertained number model based on stochastic simulation for hydrological frequency analysis

作　　者：张明[1] 周润娟[1] 金菊良[2] 

机构地区：[1]安徽工程大学建筑工程学院,安徽芜湖241000 [2]合肥工业大学土木与水利工程学院,合肥230009

出　　处：《水力发电学报》2012年第5期14-18,共5页Journal of Hydroelectric Engineering

基　　金：安徽工程大学引进人才科研启动基金(2009YQQ012);国家自然科学基金项目(51079037);安徽高校省级自然科学研究项目(KJ2012B016)

摘　　要：水文频率分析中存在着多组统计参数对应着相同的适线目标函数值的"异参同效"现象,该现象与适线方法、分布线型选择等因素有关。如何选择合适的参数组进行频率分析,构成了一个未确知系统问题。在利用随机模拟方法获得大量较优目标函数值对应的参数组基础上,分别构建了均值、离差系数和偏态系数的未确知数形式变量,并采用未确知数随机模拟模型进行模拟求解。实例分析结果表明:未确知数随机模拟模型计算结果与权函数法计算结果接近,优于目估适线法的分析结果,且计算结果较为稳定。该未确知数随机模拟模型在具有随机性、模糊性和数据资料不精确情况下的频率分析、洪水灾害风险管理等各种水资源不确定性系统中,具有推广应用价值。Phenomenon that a few groups of statistic parameters correspond to the same fitting functions, always exists in hydrological frequency analysis, namely parameters equifinality, which could be caused by either different fitting methods or different distribution types. How to select a suitable group of parameters to progress frequency analysis forms an unascertained mathematics problem. In this paper, a method based on unascertained number using Motel Carlo （UNMC） is proposed to deal with parameters cquifinality. First, a few groups of statistic parameters of equivalent fitting effect are.generated by a large number of optimization searching. Then, unascertained variables for each of the statistic parameters, mean, coefficient of variation and coefficient of skewness, are constructed according to the parameterg distribution. Finally, a UNMC model is used to calculate the peak flood of each design frequency. Results of case studies indicate that UNMC can obtain results closer to those by the weighted function method, and both methods are superior to the traditional curve fitting methods. UNMC could be widely used in other frequency analysis problems in hydrology and water resources system.

关 键 词：水文学 水文频率分析 异参同效 未确知数 随机模拟 蒙特卡洛方法 

分 类 号：TV122[水利工程—水文学及水资源]