基于AGNES和K-Medoids算法的河网实时控制规则优化方法

Optimization method for real-time river network control rules based on AGNES and K-Medoids algorithms

摘要

【目的】针对城市内部河网对闸泵控制规则响应相对滞后的问题，急需对现有河网实时控制规则进行优化，以减轻暴雨期间河网对排水管网的顶托，进而降低洪涝风险。【方法】首先，采用InfoWorks ICM构建洪涝模型识别了不同设计暴雨下的顶托排水口，并定量分析了顶托排水口现状和最大排水能力下的陆面淹没响应。然后，基于AGNES和K-Medoids算法构建了双层聚类模型，根据顶托排水口的特征确定了河网调控点。最后，以河网调控点为调控参考对象优化了河网实时控制规则，通过观测暴雨事件评估了所提方法的性能。以福州市仓山区双向感潮片区为案例进行分析。【结果】结果显示：河网顶托的排水口比例是23.73%,确定了3个河网调控点。“苏迪罗”超强台风暴雨下，对比原有规则，优化规则下河道断面水头降低了3.51%,溢流雨水井数量减少了1.53%,陆面淹没面积和深度分别减少了6.18%和9.25%。【结论】识别顶托排水口可初步揭示洪涝风险区，计算陆面淹没响应可为确定河网调控点和优化河网实时控制规则提供可靠依据；通过对比模拟观测暴雨事件，证实了优化规则对降低河道断面水头的有效性和优越性，表明了更新闸泵调控参考对象是提高城市内部河网响应速度、缓解洪涝风险的关键途径；所提方法为优化河网运行、提升城市洪涝管理能力提供参考。

Abstract

[Objective] To address the delayed response of urban river networks to sluice and pump control rules, existing real-time control rules should be optimized to mitigate the backwater effect on drainage networks during storm events, thereby reducing flood risks. [Methods] First, a flood model was established using InfoWorks ICM to identify backwater-affected drainage outlets under different design storm events, followed by a quantitative analysis of surface inundation responses under both existing conditions and maximum drainage capacity. Then, a two-layer clustering model integrating AGNES and K-Medoids algorithms was established to determine river network control points based on the characteristics of backwater-affected drainage outlets. Finally, using these control points as control references, real-time control rules were optimized, and the performance of the proposed method was evaluated through observed storm events. A bidirectional tidal-influenced area in Cangshan District, Fuzhou City, was selected as a case study. [Results] The result showed that 23.73% of drainage outlets experienced backwater, and three river network control points were determined. Under the Super Typhoon Soudelor storm event, the optimized rules reduced the water head at river cross-sections by 3.51%, decreased the number of overflowing stormwater inlets by 1.53%, and lowered the surface inundation area and depth by 6.18% and 9.25%, respectively, compared to the original rules. [Conclusion] Identifying backwater-affected drainage outlets helps preliminarily locate flood-prone areas, and analyzing surface inundation responses provides a reliable basis for determining river network control points and optimizing real-time control rules. The comparative simulation of observed storm events verifies the effectiveness and advantages of the optimized rules in reducing river cross-section water head. Furthermore, it shows that updating sluice and pump control references is crucial for enhancing the responsiveness of urban river networks and mitigating flood risks. The proposed method provides a practical reference for optimizing river network operations and improving urban flood management.

关键词

城市洪涝 / 河网实时控制 / 顶托排水口 / 双层聚类模型 / 河网调控点 / 优化规则

Key words

urban flooding / real-time river network control / backwater-affected drainage outlet / two-layer clustering model / river network control point / optimized rules

基金资助

国家重点研发计划(2022YFC3800102)

水利部重大科技项目(SKS-2022014)