基于级联效应的水库调度决策多维综合评估

Multi-dimensional comprehensive evaluation of reservoir operation decision-making based on cascade effect

水库防洪、兴利调度的核心为水库水位的科学控制。受复杂现实环境因素影响,如应急突发事件、水资源供需不确定性等,实际调度过程中各时段水库水位往往偏离初始优化路径,进而影响后续调度决策以及防洪、兴利效益目标,即水库水位控制决策受扰动后的级联效应现象。文中提出考虑影响持续性、全局损失性以及系统失效性3个方面的水库调度决策级联效应评估指标体系,以此为基础,采用TOPSIS方法量化月尺度水库调度决策级联效应,应用K-means算法划分级联效应强度等级,并在泾河流域东庄水库供水调度中进行应用。结果表明:1)相对于影响持续性指标与系统失效性指标,全局损失性指标对东庄水库调度决策的变化更加敏感,其中在枯水期尤为显著;2)水库调度决策扰动产生的综合级联效应在10月份最为显著,5-7月份相对较弱,重点管控10月水库水位有利于避免水库水位偏离初始优化路径时大规模负向级联效应的产生。

The core of reservoir flood control operation and joint utilizable regulation is the scientific control of reservoir water level.Affected by complex realistic environment,such as emergencies,uncertain water supply and demand,etc.,the reservoir water level in actual operation often deviates from initial expected path.As a result,the deviation affects subsequent scheduling and the related flood control and benefit objectives,that is the cascading effect phenomenon embedded in reservoir operation.This article proposes an evaluation index for quantifying the cascading effect of reservoir operation,which contains the persistence of impact,the loss of global state,and the failure of reservoir operation system.Based on this index system,the TOPSIS method is used to quantify the cascading effect of monthly scale reservoir operation.The K-means algorithm is applied to classify the cascading effect intensity levels.The Dongzhuang Reservoir in the Jinghe River Basin is selected as the study.Obtained results show that:1)Compared to the impact sustainability index and system failure index,the global loss index is more sensitive to the change of reservoir operation,especially during the dry season.2)The comprehensive cascade effect caused by disturbance in reservoir operation is the most significant in October,and relative weaker from May to July.Focusing on controlling the reservoir water level in October can help avoid the occurrence of large-scale negative cascade effects when the reservoir water level deviates from the initial optimization path.