孤立森林算法在大坝监测数据异常识别中的应用  

Application of Isolated Forest in Abnormal Identification of Dam Monitoring Data

作　　者：张海龙[1] 范振东 陈敏 ZHANG Hailong;FAN Zhendong;CHEN Min(Shaanxi Key Laboratory of Safety and Durability of Concrete Structures,Xijing University,Xi’an 710123,China;Huadong Engineering Corporation Limited,Power China,Hangzhou 311122,China;Huaian Water Conservancy Survey Design and Research Institute Co.,Ltd.,Huaian 223200,China)

机构地区：[1]西京学院陕西省混凝土结构安全与耐久性重点实验室,陕西西安710123 [2]中国电建集团华东勘测设计研究院有限公司,浙江杭州311122 [3]淮安市水利勘测设计研究院有限公司,江苏淮安223200

出　　处：《人民黄河》2020年第8期154-157,168,共5页Yellow River

基　　金：陕西省教育厅科学研究项目(15JK2171)。

摘　　要：大坝安全监测数据的可靠性和完整性是分析监测效应量变化规律、评判大坝运行性态的前提,有必要对大坝安全监测异常数据进行识别和清除,以提高监测数据质量。将适用于大数据处理的孤立森林算法应用于大坝监测异常数据识别中,首先通过小波变换提取监测数据的趋势项,然后采用孤立森林算法对扣除趋势项的剩余量进行异常值识别。通过实例对比分析发现该方法能有效识别大坝监测数据的异常值。The reliability and integrity of the dam safety monitoring data is the premise for analyzing the variation and judging the operational behavior of the dam.For improving the quality of monitoring data,it is necessary to identify and clear the abnormal data.Therefore,the isolated forest algorithm was applied to the abnormal identification of dam monitoring data.Firstly,the trend change data was extracted by wavelet transform,then the residual amount was identified by the isolated forest method.Through the practical application,the method could effectively identify the outliers of the dam monitoring data.

关 键 词：孤立森林 大坝监测数据 异常识别 小波变换 

分 类 号：TV698.1[水利工程—水利水电工程]