基于改进3-Sigma混合模型的水库大坝监测数据智能分析与预警技术研究  

Intelligent analysis and early warning technology study on monitoring data of reservoir dam based on improved 3-sigma hybrid model

作　　者：袁小松 许小华 王小笑 王海菁 YUAN Xiaosong;XU Xiaohua;WANG Xiaoxiao;WANG Haijing(Jiangxi Academy of Water Science and Engineering,Jiangxi Provincial Technology Innovation Center for Ecological Water Engineering in Poyang Lake Basin,Jiangxi Key Laboratory of Flood and Drought Disaster Defense,Nanchang Jiangxi,330029,China)

机构地区：[1]江西省水利科学院,江西省鄱阳湖流域生态水利技术创新中心,水旱灾害防御江西省重点实验室,江西南昌330029

出　　处：《江西水利科技》2024年第5期343-348,共6页Jiangxi Hydraulic Science & Technology

基　　金：江西水利科技项目(202224ZDKT19);江西省技术创新引导类计划项目(20223AEI91008).

摘　　要：水库大坝监测数据的安全阈值范围确定及预警体系构建是大坝安全监测的关键,拟定预警指标阈值的主要任务是根据坝体抵御经历荷载的能力,来评估和预测抵御可能发生荷载的能力,从而确定在该荷载组合下,监控效应量的极值和警戒值。针对水库大坝渗流、渗压、变形等长序列监测数据,本文提出了一种基于改进3-Sigma的混合模型,通过设计离群因子对安全阈值范围进行调控,融入DBSCAN模型,从序列分布和密度两个方向进行分析,提取出序列数据的异常值,确定大坝安全承载阈值范围,然后通过计算自定义预警系数,最终确定预警阈值范围,从而指导水库大坝安全运行管理。试验结果表明:相较于常用的异常值检测模型,混合模型查准率(99%)、F1分数(0.97)、AUC-PR(0.99)三项指标均表现优秀,在大坝安全监测领域具有可行性。The determination of the safety threshold range for monitoring data of reservoir dams and the construction of an early warning system are key factors for dam safety monitoring.The main task of formulating early warning indicator thresholds is to evaluate and predict the ability to resist potential loads based on the dam's ability to withstand experienced loads,in order to determine the extreme and warning values of the monitoring effect under this load combination.This technology is mainly aimed at monitoring data of long sequences such as seepage,seepage pressure,and deformation of reservoir dams.A hybrid model based on improved 3-Sigma is proposed,which regulates the safety threshold range by designing outlier factors.The DBSCAN model is integrated to analyze the sequence distribution and density,extract outliers from the sequence data,determine the safety bearing threshold range of the dam,and then calculate the custom warning coefficient to ultimately determine the warning threshold range,thereby guiding the safe operation and management of reservoir dams.The experimental results show that compared to commonly used outlier detection models,the three indicators of precision(99%),F1 score(0.97),and AUC-PR(0.99)all perform well and are feasible in the field of dam safety monitoring.

关 键 词：大坝安全监测 异常监测 离群因子 混合模型 

分 类 号：TV697.21[水利工程—水利水电工程]