流域水文分析与水文预报方法研究

发布时间：2017-12-27 00:18

本文关键词：流域水文分析与水文预报方法研究　出处：《华中科技大学》2016年博士论文　论文类型：学位论文

【摘要】：流域水文分析和水文预报是水文学领域的两个主要研究方向,是水利工程规划建设、水资源优化配置及安全高效可持续利用的重要支撑。由于我国特殊的地理位置和气候条件,导致水资源在时空分布上极度不均,此外受气候变化与人类活动的强烈影响,流域水循环过程和水资源时空分布规律发生了深远变化,加剧了流域水文特性的复杂程度和水资源的不安全性,尤其是大型水利工程、跨流域调水工程等人类活动对水文系统产生了重要影响,水利工程胁迫下流域自然径流破碎化导致水文系统偏离自然条件下的演变规律,使得流域水资源系统的时空变异规律更加复杂,对流域水文分析和水文预报研究提出了更高要求。本文围绕变化环境下流域水资源演变及高效利用中面临的关键科学问题和技术难题,以长江上游为主要研究对象,研究流域水文分析和水文预报的先进理论方法与技术手段,研究工作对于减轻早涝灾害损失及实现水资源可持续利用具有重要的理论指导意义和工程实用价值。相关研究成果可供流域管理机构参考和借鉴,应用前景广阔。本文的主要研究内容和创新性成果包括：(1)为克服单变量趋势分析方法无法检验出整体水文事件是否存在显著变化趋势的局限,本文引入多变量Mann-Kendal趋势分析方法,分别对长江上游干支流水文控制站的年最大洪峰、年最大7d洪量和年最低月平均径流、年最低3个月平均径流的单变量与联合变量进行变化趋势分析。实例研究表明,长江上游洪水过程整体呈现减小趋势,而低径流过程则整体呈增大趋势。在整体水文事件趋势分析方面,多变量Mann-Kendal趋势分析方法展现出明显优势,能够检验多个相互关联的水文变量是否具有显著变化趋势,而单变量Mann-Kendal趋势分析方法仅能单独检验某一水文变量否具有显著变化趋势。因此,对包含多个相互关联变量的水文事件进行趋势分析时,需要同时进行单变量和多变量趋势分析才能全面掌握水文事件中单个水文变量变化趋势和整体变化趋势。(2)在水文模型参数率定中,基于残差的整体性评价指标能够定量评价模型模拟结果和实测水文资料的差别,因而被广泛选作目标函数来指导水文模型的参数率定。然而,基于整体性评价指标的参数率定方法仅能保障水文模拟结果和实测水文资料的残差尽可能小,无法满足水文模拟结果和实测水文资料有较高的水文一致程度。因此,本文将能够量化流域水文特征的水文特性签名及整体性评价指标一起作为参数率定的目标函数,极大提高了径流预报的水文一致程度,尤其是提出的流量信息熵差指标能够度量模拟和实测径流的静态统计信息差异,从而提高了预报径流拟合流量历时曲线的能力。同时,针对目标函数个数过多而出现的支配保留现象,研究工作对水文特性签名目标函数进行离散化,既有效缓解了支配保留效应,又能兼顾目标函数的区分性能,提高了水文模型参数率定可包含的目标函数个数。(3)针对确定性预报仅给出变量在未来时刻的一个单点预报值,没有提供与预报相关的内在不确定程度,本文通过建立区间预报来度量洪水预报的不确定性。然而,传统区间预报建立方法需要对水文数据或预报误差的概率分布进行人为假设,而且建立区间预报的计算量较大,阻碍了区间预报的实际应用。因此,本文提出两种改进LUBE区间预报方法,第一种较大程度地改进了原始单目标LUBE区间预报方法存在的问题,第二种则是扩展单目标LUBE区间预报方法到多目标框架。长江上游洪水区间预报的研究结果表明,两种方法均明显提高了区间预报效果,在相似的区间预报覆盖率下,利用本文所提方法产生的区间预报宽度更窄。同时,本文采用区间预报平均相对宽度指标计算区间预报的相对宽度,提高了高流量时段的洪水预报效果。此外,多目标LUBE区间预报方法能够帮助研究人员根据需要选择覆盖率和宽度合适的区间预报,极大地减少了单目标LUBE方法中CWC目标函数参数选择带来的工作量。(4)为提高中长期径流预报精度,本文研究偏互信息输入因子选择方法,从前期实测降水、径流和遥相关气候因子中选择合适的输入,通过增加与流域径流相关性高的遥相关气候因子作为数据驱动模型的输入,实现了流域中长期径流高精度预报。偏互信息输入因子选择方法能够度量输入变量和输出变量的线性与非线性相关关系,且可以避免冗余输入变量的选入,函数测试结果表明,偏互信息方法对时间序列模型和非线性模型均十分有效,能够准确按照与输出变量相关程度顺序挑选出相关变量。同时,金沙江流域实例研究结果也表明偏互信息方法适用于水文气象变量的输入因子选择,基于偏互信息方法的预报结果优于线性相关系数法。
[Abstract]:Watershed hydrological analysis and hydrological prediction are two main research directions in the field of hydrology. They are important support for water conservancy project planning and construction, optimal allocation of water resources, and safe, efficient and sustainable utilization. Due to China's special geographical location and climatic conditions, the water resources is extremely uneven distribution in space and time, is strongly affected by the climate change and human activities, a profound change occurred in the water cycle process and the spatial and temporal distribution of water resources law, heightened the safety complexity of watershed hydrologic characteristics and water resources. Especially the large water conservancy project, inter basin water diversion project and other human activities have an important impact on the hydrological system, water conservancy project stress natural runoff basin fragmentation lead to deviations from the evolution of natural conditions makes the hydrological system, spatial and temporal variability of watershed water resources system is more complex, put forward higher requirements on watershed hydrologic analysis and forecast research. The key scientific and technical problems in the evolution of basin water resources around the changing environment and efficient utilization of this paper, in the upper reaches of the Yangtze River as the main research object, research on watershed hydrologic analysis and prediction of the advanced theoretical methods and technical means, research work to alleviate the drought and flood disaster losses and has important theoretical significance and practical value to realize the sustainable utilization of water resources. The relevant research results can be used for reference and reference for river basin management institutions, and the application prospects are wide. The main research contents and innovative results are as follows: (1) in order to overcome the single variable trend analysis method to test the overall hydrological events if there is a significant trend of limitations, this paper introduces the multi variable Mann-Kendal trend analysis method, analysis the change trend of single variable and joint variable respectively on the upper reaches of the Yangtze River Hydrological control station of the annual maximum the annual maximum flood volume and flood peak, 7d years minimum monthly runoff and annual minimum 3 months of average runoff. The case study shows that the flood process in the upper reaches of the Yangtze River tends to decrease, while the low runoff process is increasing. In the overall trend of hydrological event analysis, multivariate Mann-Kendal trend analysis method shows obvious advantages, to test whether the multiple interrelated hydrological variables have significant changes in trend, while the single variable Mann-Kendal trend analysis method can only test a single variable has significant change. Therefore, trend analysis of hydrological events involving multiple interrelated variables requires simultaneous univariate and multivariate trend analysis to fully grasp the trend and overall trend of single hydrological variables in hydrological events. (2) in the calibration of hydrological model parameters, the holistic evaluation index based on residuals can quantitatively evaluate the difference between the simulated results and the observed hydrologic data. Therefore, it is widely selected as the objective function to guide the parameter calibration of hydrological models. However, the method of parameter calibration based on holistic evaluation index can only ensure that the residual of hydrological simulation results and measured hydrologic data is as small as possible, which can not meet the high hydrological consistency of hydrological simulation results and measured hydrological data. Therefore, this paper will be able to sign the hydrological characteristics and overall evaluation index quantification of watershed hydrological characteristics as objective function parameters set, which greatly improves the consistency of hydrological runoff forecasting, especially the static statistical information flow difference information entropy the index can measure the simulated and measured runoff, thus improving the ability to predict runoff flow duration curve fitting. At the same time, according to the objective function of a large number of emerging dominant retention phenomenon, the discretization of the hydrological characteristics of the objective function signature research, can effectively alleviate the dominant reserve effects, and can also distinguish the performance objective function, the objective function improves a number of hydrological model parameters calibration can be contained. (3) for a deterministic prediction, we only give a single point prediction value of variables in the future time, and do not provide the intrinsic uncertainty related to prediction. In this paper, we set up interval prediction to measure the uncertainty of flood forecasting. However, the traditional interval prediction method needs artificial hypothesis for the probability distribution of hydrological data or forecast error, and the computation of interval prediction is large, which hinders the practical application of interval prediction. Therefore, two improved LUBE interval prediction methods are proposed in this paper. The first one improves the existing problems of the original single target LUBE interval prediction method, and the second one is to extend the single target LUBE interval prediction method to the multi-objective framework. The results of flood interval prediction in the upper reaches of Yangtze River indicate that the two methods can obviously improve the effect of interval prediction. Under the similar interval prediction coverage, the interval prediction produced by the proposed method is narrower. At the same time, in this paper, the relative width of interval prediction average relative width is used to calculate the relative width of the interval forecast, which improves the effect of flood forecast in high flow time period. In addition, multi-objective LUBE interval prediction method can help researchers choose interval prediction with appropriate coverage and width according to needs, which greatly reduces the workload of parameter selection of CWC target function in single target LUBE method. (4) in order to improve the long-term runoff forecast accuracy, this paper studies partial mutual information input factor selection method, selecting the appropriate input from the measured rainfall, runoff and teleconnection climate factor, the increase in runoff and watershed high correlation teleconnection climate factors as data driven model input, to achieve the long-term runoff high basin the accuracy of prediction. Partial mutual information input factor selection method can measure the linear and nonlinear correlation between input variables and output variables, and avoid redundant input variables selection. The result of function test shows that partial mutual information method is very good for time series models and nonlinear models.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：P333;P338

【相似文献】