多监测项目联合的大坝安全监控模型  

Dam Safety Monitoring Model based on Multiple Monitoring Items

作　　者：王丽蓉 郑东健[2] WANG Lirong;ZHENG Dongjian(PowerChina Northwest Engineering Corporation Limited,Xi'an 710065,China;College of Water-Conservancy & Hydropower Engineering, Hohai University, Nanjing 210098,China)

机构地区：[1]中国电建集团西北勘测设计研究院有限公司,西安710065 [2]河海大学水利水电学院,南京210098

出　　处：《西北水电》2021年第6期8-14,25,共8页Northwest Hydropower

基　　金：国家重点研发计划课题(2018YFC1508603);国家自然科学基金重点项目(51739003).

摘　　要：大坝安全监测布置的各种监测项目,从不同侧面反映大坝的工作状态,各监测项目之间既有区别又有联系,有时还存在相互影响。将各监测项目观测资料作为反映大坝工作状态信息的一个整体进行分析,有助于判断监测量是否在合理区间内有规律地波动,可以更加有效揭示各种内外因素对大坝运行状态整体的影响。但各监测项目测点间存在的内在规律难以获取,如何确定大坝联合监控模型的因子表达模式、构造模型函数,存在极大困难。因此融合动态时间规整算法(DTW)和卷积长短时记忆网络(ConvLSTM),构建了大坝安全监测整体信息联合监控模型,该模型具有提取测点最新相关性信息和处理时序数据的能力。将该模型应用于某重力坝同坝段6个监测点的测值预测,监测项目包括水平位移、垂直位移、扬压力,考虑上游水位、气温、降雨量3种环境量的影响,模型预测精度明显高于LSTM模型和常规统计模型,可为大坝安全监控及预警的可靠性提供保障。The various monitoring items arranged for the dam safety monitoring reflect the working state of the dam from different angles,and all the monitoring items are independent and interrelated,and sometimes there is mutual influence.Analyzing the observation data of each monitoring item as a whole reflecting the dam's working state can help to judge whether the monitoring amount fluctuates regularly within a reasonable range and understand the overall influence of various internal and external factors on the dam's working state.However,it is difficult to obtain the inherent law among the monitoring points of each monitoring item,and it is extremely difficult to determine the factor expression mode and the constructing model function of the multi-item monitoring model of the dam.Therefore,the Dynamic Time Warping(DTW)algorithm and Convolutional Long Short-Term Memory network(ConvLSTM)are combined to construct a joint monitoring model for the overall safety data of the dam.This model is capable of extracting the latest correlation information of monitoring points and processing time series data.The model is applied to predict the monitoring data of six monitoring points in the same dam section of a gravity dam and the monitoring items include horizontal displacement,vertical displacement and uplift pressure.Considering the influence of upstream water level,temperature and rainfall,the prediction accuracy of the model is significantly higher than that of the LSTM model and the conventional statistical model,which can provide a guarantee for the reliability of dam safety monitoring and early warning.

关 键 词：大坝安全监测 联合监控模型 深度神经网络 动态时间规整算法 

分 类 号：TV698.1[水利工程—水利水电工程]