【作者】 李娜；

【导师】 王祥三；

【作者基本信息】 武汉大学 ， 水文学及水资源， 2005， 硕士

【摘要】 本文结合广西北海市南流江流域河库洪水预报问题对水库洪水预报和河段洪水预报的基本理论及模型进行了探讨。鉴于参数识别的好坏直接影响到模型的精度和效率,本文重点探讨和研究了一种高效混合遗传算法并用其来优选概念性水文模型的参数,取得了良好的效果。论文主要进行了下述工作: 一.概括了洪水预报的必要性和意义,叙述了国内外洪水预报的研究现状和发展趋势。详尽的阐述了水文预报方法的发展历史。 二.介绍了水文预报的基本概念和各种水文模型的特点。详细阐述流域水文预报产汇流模型的理论和结构以及河段洪水预报的原理和技术。叙述了洪水实时校正模型和技术。 三.对参数识别的基本方法及各种方法的特点、发展和应用情况进行了分析。介绍了遗传算法的基本概念,论述了基本遗传算法及其实现过程;鉴于基本遗传算法和传统的模型参数率定方法各有利弊,在介绍了Rosenbrock算法的基础上,将它们进行了有机的结合形成了混合遗传算法,并对遗传算子进行了设计和改进,采用了小生境技术;利用典型数学函数对混合遗传算法的性能进行了测试分析,证明了混合遗传算法与基本遗传算法相比具有更好的收敛性能和更高的收敛速度。并对多目标决策混合遗传算法进行了初步的探讨。 四.结合广西北海合浦水库的工程实例,选用西塘站以上控制面积采用新安江三水源模型对其参数进行了率定;对比分析了单目标决策的遗传算法和多目标决策的混合遗传算法的优选效果;将优选得到的参数应用于整个小江流域进行了洪水预报。在此基础上应用改进后的可变遗忘因子递推最小二乘算法对模拟结果进行了实时校正,显著提高了预报精度。最后采用三种方案进行了河库联合洪水预报,其中方案二利用分段马斯京根法和最小二乘回归方法进行河道洪水演算,计算效果较好。更多还原

【Abstract】 The thesis studies the theory and the model of runoff forecasting and river forecasting, as well as the application in the Nanliujiang watershed. Considering the parameter estimation influencing the precision and efficiency of model directly, this thesis proposes a Hybrid Genetic Algorithm to estimate the parameter of the conceptional hydrologic model, which results show that it works efficiently and is rather satisfactory. The main content of this dissertation are as follows.1. The necessity and significance of the flood forecasting is summarized. The history of hydrology forecasting is reviewed, and the current situation and the development trend of flood forecasting in domestic and international are presented.2. The conception of hydrology forecasting and the characteristic of each model are introduced. The theory and key technology of runoff-concentration model and river flood forecasting are explained in detail. Besides, the real-time calibrating model is reviewed.3. The development, application and deficiency of the parameter estimation methods are analyzed. Following, both the basic conception of Genetic Algorithm and the Standard Genetic Algorithm are introduced. After the recommend of the Rosenbrock model, the thesis combines the GA with Rosenbrock to propose a new method called Hybrid Genetic Algorithm. The dissertation also design and improve the Genetic Algorithm operator, adopting the niche technology. The Hybrid Genetic Algorithm is evaluated by typical mathematic functions and the results prove that the Hybrid Genetic Algorithm has a stronger convergence capability and faster convergence velocity than the SGA. Moreover, the multiple-objects Hybrid Genetic Algorithm is also studied.4. Hepu reservoir in Guangxi province is taken for example. The catchment area which is located in the upstream of Xitang rainfall-station is chosen to estimate the parameter of Xin’anjiang model. The results of single-object Hybrid Genetic Algorithm and multiple-objects Hybrid Genetic Algorithm are compared. By using the estimated parameter for whole reservoir, the dissertation studies the watershed flood forecasting of Hepu. The improved recursive least-squares estimated with variable forgotten factorsmethod is used to improve the model efficiency evidently. Finally, three methods are used to forecast the flood of reservoir and reach respectively. With the second method, Muskingun method and least-squares regress model are used to forecast river flood, which gets the content results.更多还原