学位论文-—水库泥沙淤积的不确定因素分析--外文翻译.doc

外文翻译
Uncertainty Analysis Of Reservoir Sedimentation
Abstract: Significant advances have been made in understanding the importance of the factors involved in reservoir sedimentation. However, predicting the accumulation of sediment in a reservoir is still a complex problem. In estimating reservoir sedimentation and accumulation, a number of uncertainties arise. These are related to quantity of streamflow, sediment load, sediment particle size, and specific weight, trap efficiency, and reservoir operation。In this study, Monte Carlo simulation and Latin hypercube sampling are used to quantify the uncertainty of annual reservoir sedimentation and accumulated reservoir sedimentation through time. In addition, sensitivity analysis was performed to examine the importance of various factors on the uncertainty of annual reservoir sedimentation. The proposed procedures have been applied to the Kenny Reservoir at the White River Basin in Colorado.The uncertainty of annual reservoir sedimentation and the effect of each uncertain factor, taken individually and in combinations, on the uncertainty of accumulated reservoir sedimentation through time have been examined. The results show that annual streamflow and sediment load are the most important factors determining the variability of annual reservoir sedimentation and accumulated reservoir sedimentation.In the case of Kenny Reservoir, the uncertainty expressed by the coefficient of variation can be on the order of 65% for annual reservoir sedimentation and 39% for accumulated reservoir sedimentation volume.
Introduction
Reservoir sedimentation varies with several factors such as sediment production, sediment transportation rate, sediment type, mode of sediment deposition, reservoir operation, reservoir geometry, and streamflow variability. Sediment is transported as suspended and bed loads by streams and rivers coming into a reservoir. Due to flow deceleration when a river approaches a reservoir, the sediment transport capacity decreases,and some of the incoming sediment is trapped and deposited in the reservoir. In addition, the deposited sediments may consolidate by their weight and the weight of overlying water through time. Predicting the sediment coming into a reservoir,its deposition, and its accumulation throughout the years, after construction of the dam, have been important problems in hydraulic engineering. Despite the advances made in understanding several of the factors involved in reservoir sedimentation, predicting the accumulation of sediment in a reservoir is still a complex problem. Empirical models, based on surveys and field observations, have been developed and applied to estimate annual reservoir sedimentation load (RSL), accumulated reservoir sedimentation load, (ARSL), and accumulated reservoir sedimentation volume (ARSV) after a given number of years of reservoir operation. Likewise, several mathematical models for predicting reservoir sedimentation have been developed based on the equations of motion and continuity for water and sediment.However,empirical methods are still widely used in actual engineering practice.
In estimating resevoir sediment inflow, reservoir sedimentation,and reservoir sediment accumulation, either by empirical or analytical approaches, a number of uncertainties arises.The main factors affecting reservoir sedimentation are (1)quantity of streamflow; (2) quantity of sediment inflow into a reservoir;(3) sediment particle size; (4) specific weight of the deposits; and (5) reservoir size and operation. Depending on the particular case at hand, some factors may be more important than others. All of these factors are uncertain to some degree and, as a consequence, reservoir sedimentation will be an uncertain quantity too.In addition, which model (or procedure) is applicable to estimate some of the foregoing quantities and, in fact, which model is to be used to estimate the amount of sediment that will be trapped in a reservoir are questions that cannot be answered with certainty. For instance, Fan (1988) obtained information on 34 stream-,18 watershed-, and 20 reservoir-sedimentation models and stated that different models may give significantly different results even when using the same set of input data. Such an additional factor, known as ‘‘model uncertainty,’’ may be quite a large component of the overall uncertainty. In any case, the planner and manager of a reservoir may be interested in quantifying how the uncertainty of some of the factors affecting reservoir sedimentation translate into the uncertainty of annual sediment deposition and accumulated sediment deposition through time.In this paper, we address the issue quantifying the effect of parameter uncertainty on reservoir sedimentation based on a set of predefined models as will be described below.The effect of model uncertainty is not considered in this study.
Several methods of uncertainty analysis have been developed and applied in water resources engineering. The most widely used methods are first-order analysis (FOA) and Monte Carlo simulation (MCS). FOA is based on linearizing the functional relationship that relates a dependent random variable and a set of independent random variables by Taylor series expansion. This method has been applied in several water resources and environmental engineering problems involving uncertainty. Examples include storm sewer design; ground-water-flow estimation , prediction of dissolved oxygen;and subsurface-flow and contaminant transport estimation . In MCS, stochastic inputs are generated from their probability distributions and are then entered into empirical or analytical models of the underlying physical process involved in generating stochastic outputs. Then, the generated outputs are analyzed statistically to quantify the uncertainty of the output. Many examples of uncertainty analysis by MCS can be found in water resources and environmental engineering. Some examples include steady-state ground-water-flow estimation and water-quality modeling . Scavia et al. (1981) made a comparison of MCS and FOA for determining uncertainties associated with eutrophication model outputs such as phytoplankton, zooplankton, and nitrogen forms.They indicated that both MCS and FOA agree well in estimating the mean and variance of model estimates. However, MCS has the advantage of providing better information about the output frequency distribution.
Latin hypercube sampling (LHS) is an alternative simulation procedure that has been developed for uncertainty analysis of physical and engineering systems.The basic idea behind LHS is to generate random stochastic inputs in a stratified manner from the probability distributions. In this way the number of generated inputs can be reduced considerably as compared to MCS.They pointed out that the point estimate method yields a larger mean and variance than those obtained by the FOA and LHS methods. Furthermore, in studying the importance of stochastic inputs on the output by sensitivity analysis, LHS yields more information than the other two methods.
In this study, uncertainty analysis based on MCS and LHS methods are conducted to estimate the probability distribution of annual reservoir sedimentation volume (RSV). In addition,sensitivity analysis is performed to see the relative importance of stochastic inputs in estimating the variability of RSV. Furthermore,uncertainty analysis of ARSV throughout time is performed using MCS.In this procedure, annual streamflows are generated by a stochastic time series model. The effect of parameter uncertainty in the stochastic model on the output (i.e.,ARSV) is also considered.
Estimation Of Annual And Accumulated
Reservoir Sediment Load(Mass) And Volume
Reservoir sedimentation volume depends, among other factors,on the quantity of sediment inflow, the percentage of sediment inflow trapped by the reservoir, and the specific weight of the deposited sediment considering the effect of compaction with time.The incoming sediment load and the streamflow discharge are usually measured at hydrometric gauging stations, and a sediment rating curve is constructed.The sediment rating curve expresses the relationship between the rate of sediment discharge and the rate of streamflow discharge and is usually represented graphically on logarithmic coordinates.Incoming sediment is generally composed of suspended sediment and bed load. When the bed load cannot be obtained by measurements, it can be estimated by formulas.
In estimating annual sediment load, it has been common practice to use annual sediment rating curves for both suspended sediment and bed load. The annual sediment rating curve is the relation between annual sediment load and annual streamflow discharge.Two methods can be considered for determining annual sediment rating curves . A simple method involves the following steps: (1) For a given year calculate daily sediment loads from daily sediment rating curves; (2) add all daily sediment loads and divide the sum by the number of days in the year, then this value represents the annual average sediment load in tons per day; (3) repeat Steps 1 and 2 for all years of record; and (4) plot the annual average sediment load versus the annual average streamflow for each year in the record. An alternative method is based on estimating annual sediment loads using flow duration curves. In any case, an annual sediment rating curve can be constructed by simple regression analysis after logarithmic transformation of annual average streamflow discharges and annual average sediment loads. Colby (1956) stated that in actual practice daily sediment rating curves could be assumed to be equivalent to instantaneous sediment rating curves.
Daily rating curves of suspended sediment and bed load may be represented as
(1)
(2)
where QSD = daily suspended sediment load (tons/day); QBD= daily bed load (tons/day); QWD = daily average streamflow discharge (m3/s);and , and , = rating curve coefficients for suspended sediment and bed load, respectively. Cs is a rating curve correction factor needed to avoid underestimating the estimate of suspended load. Such a correction factor Cs is equal to exp(2.65) where s is the residual standard error (Ferguson 1986). Likewise, Cb is the correction factor for bed load. The corresponding annual rating curves of suspended sediment and bed load are
(3)
(4)
where QSt = (annual average) suspended sediment load (tons/day) in year t; QBt = (annual average) bed load (tons/day) in year t; QWt = (annual average) streamflow discharge (m3/s) in year t; and , and , = rating curve coefficients for annual average suspended-sediment and bed loads, respectively. Then, the (average) total sediment inflow in year t, QTt, is simply QTt = QSt + QBt.

译文
水库泥沙淤积的不确定因素分析
摘要：尽管在理解关于水库中所涉及的几个因素的重要性方面取得了很大的进展，然而，预测水库泥沙的淤积仍然是一个复杂的问题。水库泥沙的沉淀和淤积，出现了一些不确定因素。这些涉及到流量，泥沙，泥沙颗粒大小，比重，拦沙效率和水库运行等。在这项研究中，蒙特·卡洛模拟法和拉丁超立方抽样法是用来量化水库年度泥沙淤积和水库泥沙淤积过程的不确定性。此外，通过敏感度分析来确定水库年度泥沙淤积各种不确定性因素的重要性。这个程序已经应用到科罗拉多州白河流域的肯尼水库。水库年度泥沙淤积的不确定性和影响水库泥沙淤积的不确定性因素，采取单独和组合的方法审查水库泥沙淤积进程中的每个不确定因素。结果表明，年流量和输沙量是确定水库年度泥沙淤积和泥沙积累过程变化最重要的因素。在肯尼水库的情况下，变差系数的不确定性可以大概表达为年度水库泥沙淤积的65%和累积的水库泥沙淤积量的39%。
简介
水库泥沙淤积因几个因素的不同而变化，例如产沙、 输沙率、 沉积物类型、 泥沙淤积的模式、 水库调度、 水库的几何形状，河川径流变化的模式等。泥沙作为悬移质和河床质通过小溪和河流进入水库。 由于水流临近水库，流速下降，携沙能力下降，泥沙下沉并淤积在水库中。此外，淤积的泥沙可能会因泥沙自身重量和它上面的水的质量的压力而变得坚固。预测进入水库的泥沙通过时间积累所造成的淤积问题，将会是大坝施工以来一直存在在水利水电工程中的重要问题。尽管在理解关于水库中所涉及的几个因素的重要性方面取得了很大的进展，然而，预测水库泥沙的淤积仍然是一个复杂的问题。以调查和实地观察为基础的实证模型得到发展和应用于估计每年水库泥沙淤积负载 (RSL)、 累积的水库泥沙淤积负载 （ARSL），并且计算运行一定年限之后的水库泥沙淤积的库容(ARSV)。同样，预测水库泥沙淤积的几个基于运动方程和水沙连续性的数学模型已经发展。然而，在实际工程实践中仍然广泛使用实证方法。
在估计水库泥沙流入、水库泥沙沉淀和水库泥沙淤积时，只有通过实证或分析的方法，一些不确定因素才能出现。影响水库泥沙淤积的主要因素是 （1）流速 ；(2) 入库泥沙量 ；(3) 泥沙颗粒大小 ；(4) 淤积的泥沙重量 ；(5) 水库的大小和运行方式。根据眼前的一些特定情况，一些因素可能会比其他因素更重要。所有这些因素都有某种程度的不确定性，因此，水库泥沙淤积也将有很大程度的不确定性。此外，一些模型 （或过程） 适用于一些上述某些因素的估计，但事实上，另一些用于估计淤积的模型，在一座水库中，泥沙量可能是不能肯定地回答的问题。例如，Fan（1988 年） 关于对34条溪流、 18个流域和 20 个水库的泥沙淤积模型有关资料的分析，指出不同的模型可能会有明显不同的结果，即使使用相同的输入数据。这种附加因素，称为模型的不确定性，可能是整体不确定性中相当大的组成部分。在一些情况下，设计者和水库管理者可能会有兴趣量化一些影响水库泥沙淤积因素的不确定性，把它转化为每年的泥沙淤积和泥沙淤积通过时间变化的不确定性。在一些论文中，我们通过基于一组预定义的模型参数来量化对水库泥沙淤积的不确定性的影响来处理问题。在这些研究中，是不考虑模型不确定性因素的影响。
不确定性分析的几种方法已被发展和运用于水资源工程中。第一阶分析 （FOA） 和蒙特卡罗模拟 (MCS) 是最广泛使用的方法。FOA 基于线性关系，涉及一个从属的随机变量和一套独立的随机变量的函数关系，由泰勒级数展开。这种方法涉及的不确定性已应用在几个水资源和环境工程的问题中。例子包括风暴下水道设计，地面水流估计，溶解氧的预测，潜流和污染物的运输估计等。在 MCS中，随机输入通常从其概率分布分析，然后进入实证或分析基于随机产出的潜在物理过程的模型。然后，生成的产出分析统计是用以量化输出的不确定性。有很多通过MCS 所做的不确定性分析，运用于水资源和环境工程的例子。这些例子包括稳态地面水流估计和水质量建模。斯卡维亚(1981 年) 作了一个 MCS 和FOA的比较，用来确定浮游植物、 浮游动物、 和氮形式等水体富营养化模型产生的不确定性。他们表示 MCS 和FOA都非常接近估计的平均值和方差模型的估计数。然而，MCS 在提供更好地分发信息的输出频率方面更具有优势。
拉丁立方体抽样 (LHS) 是替代模拟程序，已经开发出物理和工程系统的不确定性分析。LHS 背后的基本思想是以分层方式从概率分布生成随机输入。以这种方式生成输入数目可以与 MCS 相比大大减少。他们指出的点估计方法产生更大的平均值和方差比是通过FOA 和 LHS 的方法获得的。此外，在研究随机输入输出灵敏度分析的重要性上，LHS 能比其他两种方法收获更多详细的信息。
在此研究中，基于 MCS 和 LHS 方法的不确定性分析用以估计每年水库泥沙淤积库容 (RSV) 的概率分布。此外，通过灵敏度分析来分析随机输入在估计水库泥沙淤积库容变化中的相对重要性。并且，使用 MCS 来执行 ARSV 的整个时间过程的不确定性分析。在此过程中，每年河川流量用随机时间序列模型所生成。随机参数的不确定性影响在输出的模型（ARSV)中也被考虑。
年度淤积的估计
水库泥沙负载 (质量) 和库容
在其他因素中，水库泥沙淤积量取决于进入水库的泥沙入流量和泥沙淤积率，和考虑时间过程与压实效应的淤积泥沙的具体重量的百分比。进入的泥沙负载与河川径流的排出量的测量通常在水文测站和泥沙评级曲线中构造。泥沙评级曲线表示的输沙率与河川径流排放率之间的关系，通常在对数坐标上以图形方式表示。流入的泥沙一般由悬移质泥沙及河床质泥沙组成。当河床质泥沙不能通过直接测量时，它可以通过公式估计。
在估计每年的输沙量时，很常见的做法是通过每年的泥沙评级曲线用于估计悬移质泥沙及河床质泥沙。每年的泥沙评级曲线是每年输沙量和年河川径流流量之间的关系。有两种方法可以用来确定每年的泥沙评级曲线 。一种简单方法包括以下步骤：(1) 对于给定年份，每日的泥沙淤积量可以从日常的泥沙评级曲线中查出 ；(2) 把所有日常的泥沙淤积量相加，总和除以一年中的天数，然后此值代表年平均每天泥沙淤积量；(3) 对于所有年份的记录，重复步骤 1 和 2； (4) 计算记录每一年年度平均泥沙负载与年度平均河川径流量。另一种方法是基于使用流量时间曲线估计每年泥沙荷载。在任何情况下，每年的泥沙评级曲线可以在每年平均径流排放和每年平均泥沙量对数变换之后做简单回归分析。柯比 (1956 年) 指出在实际做法中，日常泥沙评级曲线可以假定等价于瞬时泥沙评级曲线。
悬移质泥沙及河床质泥沙的每日评级曲线可表示为
(1)
(2)
其中 QSD 代表年平均悬移质泥沙负荷 （吨 / 天） ；QBD 代表年平均河床质泥沙负载 （吨 / 天） ；QWD代表每年平均河川径流流量 (立方米/秒) ；, , 分别代表评级曲线的悬移质泥沙和河床质泥沙的系数。Cs代表悬移质泥沙负载评级曲线的变差系数。变差系数 Cs 等于 2.65 的平方，其中s 是标准差 (弗格森 1986年）。同样，Cb 是河床质泥沙负载的变差系数。悬移质泥沙及河床质泥沙相应的年度评级曲线是
(3)
(4)
其中 QSt 代表平均每年悬移质输沙量 （吨 / 天）;QBt 代表平均每年河床质输沙量（吨 / 天）;QWt 代表平均每年河川径流流量 （立方米/秒)，, , , 分别代表评级曲线的悬移质泥沙和河床质泥沙的系数。然后，平均每年河川径流流量 QTt，可简单的用式 QTt = QSt + QBt 计算得到。