设计洪水不确定性下的水库防洪风险调度

水库防洪调度中，设计洪水的推求对于水库防洪安全至关重要，而在其实际的推求过程中存在着诸多不确定性因素影响。本文针对设计洪水洪峰、洪量以及洪水过程线多重不确定性，提出了一种水库防洪风险调度模型并进行求解，探求设计洪水不确定性下的水库防洪调度过程。应用Copula函数建立洪峰与最大三日洪量的联合分布模型，采用蒙特卡洛重抽样方法，获取一系列峰量联合设计值；针对传统设计洪水过程线选择的单一性问题，对水库实测洪水过程进行了分类，并对不同类型的洪水过程线进行随机模拟，生成多维不确定性下的设计洪水过程；引入经济学指标条件风险值CVaR(Conditional Value at Risk)来衡量水库防洪调度过程中超过调洪最高洪水位的风险，建立了考虑CVaR的水库防洪调度模型，通过不同的风险系数取值获得不同的水库防洪调度过程。以安康水库洪水过程为例，二次重现期标准下百年一遇设计洪水联合设计值存在着较大的不确定性；采用K-means聚类法把洪水过程分成了三类，并应用蒙特卡洛法对不同类型洪水过程进行随机模拟，获取了考虑不确定性的大量水库设计洪水过程，推求了安康水库不同风险系数下水库防洪调度规则。本文所建立的考虑设计洪水不确定性的水库防洪风险调度模型可为不确定性条件下水库调度规则的制定提供指导。

Reservoir Risk Operation for Flood Control Under Design Flood Uncertainty

Design flood determines reservoir safety in the context of reservoir flood control operation. However, many uncertainties come with design flood calculation process. Therefore, this paper proposes reservoir risk operation model for flood control under design flood uncertainty and explores reservoir flood control operations. Specifically, the copula function is employed to establish the joint distribution model of flood peak and maximum three-day flood volume. Following, the Monte Carlo method is used to resample the bivariate design flood under specific return period. Moreover, in order to lower the impact from selection singleness of traditional design flood hydrograph, we classify the measured flood processes into different types and perform random simulation to generate large amount of design flood processes. Finally, the economic index conditional value at risk (CVaR) is introduced to measure the risk of maximum reservoir level exceeding the flood control water level in the process of reservoir flood control. A reservoir risk operation model for flood control employing CVaR is established. Different reservoir flood control operations can be obtained under different risk coefficient values. The AnKang reservoir is selected as the case study. It can be seen that design flood uncertainty is considerably large with secondary type return period of one hundred years. The K-means clustering method divides the measured flood processes into three categories. Following, the Monte Carlo method is used to simulate different types of flood processes and obtain a large amount of design flood processes. Facing the uncertain design flood processes, we derive reservoir flood control operations under different risk coefficients. This paper establishes the reservoir risk operation model for flood control which can effectively cope with the uncertain design flood. Related results can provide support for making reservoir flood control operations under design flood uncertainty.