变形分量信息随机森林分析法在缺失数据处理中的应用  

Application of Random Forest Analysis of Deformation Component Information in Missing Data Processing

作　　者：季骏 鲍中秋 张晓阳 贾玉豪 JI Jun;BAO Zhongqiu;ZHANG Xiaoyang;JIA Yuhao(Nanjing Water Conservancy Planning and Design Institute Co.,Ltd.,Nanjing 210022,Jiangsu,China;Shanghai Investigation,Design&Research Institute Co.,Ltd.,Shanghai 200335,China;College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210024,Jiangsu,China)

机构地区：[1]南京市水利规划设计院股份有限公司,江苏南京210022 [2]上海勘测设计研究院有限公司,上海200335 [3]河海大学水利水电学院,江苏南京210024

出　　处：《水力发电》2023年第7期95-100,共6页Water Power

基　　金：国家自然科学基金资助项目(51579085)。

摘　　要：完整的变形监测数据是评估大坝运行状态的前提条件,数据缺失将给相关分析带来很大难度。为此,提出了基于随机森林(RF)算法对不同尺度变形分量信息提取缺失值的处理方法。首先,将待处理测点和相似变形测点监测数据变换为不同尺度的分量;然后,以待处理测点各分量为因变量,使用RF法从其他测点分量中提取对其影响显著的分量,记为一组样本,依次提取直至完成所有分量的信息提取;最后,对得到的样本组构建BP模型,并组合得到原始变形缺失数据补全结果。经实例分析可知,相比于其他方法,基于随机森林算法的缺失数据处理方法的变形补全结果更准确,在大坝变形缺失值补全中有明显优势。该方法可为坝工领域其他缺失监测量的处理方法提供参考。Complete deformation monitoring data is a prerequisite for assessing the operational status of a dam,and the data missing will bring great difficulty to the related analysis.To this end,the missing value processing method based on random forest(RF)algorithm for deformation component information extraction at different scales is proposed.First,the monitoring data of the measurement points to be treated and similar deformation measurement points are transformed into components of different scales.Then,each component of the measurement points to be treated is used as the dependent variable,and the components with significant influence on them are extracted from other components using RF method,which are recorded as a group of samples until all the components are extracted in turn.Finally,the BP models are constructed for the obtained sample sets and combined to obtain the missing data completion results in the original deformation.As shown by the example analysis,the deformation completion results of the proposed missing data processing method based on RF algorithm are more accurate and have obvious advantages in dam deformation missing value completion compared with other methods.This method can also be used as a reference for the processing of other missing monitoring data in the field of dam engineering.

关 键 词：大坝变形 监测数据 缺失值补全 变形分量 信息提取 随机森林 BP模型 

分 类 号：TV698.1[水利工程—水利水电工程]