摘要
大坝变形监测数据通常呈强波动性,针对大坝变形信息中的高频分量,提出了EMD-EEMD-LSTM模型对其中的大坝变形信息进行分析,预测大坝变形趋势。首先,选取EMD对原始大坝变形序列进行分解,得到若干分量;针对高频分量,使用EEMD对其进行分解,以挖掘蕴含其中的有效变形信息;最后以LSTM为预测模型,对上述得到的各分量进行建模分析。分析表明,EMD-EEMD模型可以有效解决了原始变形序列及高频分量的强波动性,结合LSTM在时序预测中的优越性,基于EMD-EEMD-LSTM的大坝变形预测模型具有较高精度。
As the monitoring data of dam deformation usually show strong volatility and there are high-frequency components in these deformation information,an EMD-EEMD-LSTM model is proposed to analyze the dam deformation information and predict the dam deformation trend.First,the EMD is selected to decompose the original dam deformation sequence to obtain several components.Then,for the high-frequency components,the EEMD is used to decompose them to explore the effective deformation information embedded inside.Finally,the LSTM is used as the prediction model to model and analyze each component.The analysis shows that the EMD-EEMD model can effectively solve the strong volatility of the original deformation sequence and high-frequency components,and combined with the superiority of LSTM in time series prediction,the proposed method can greatly improve the prediction accuracy of dam deformation.
作者
董泳
刘肖峰
李云波
贾玉豪
DONG Yong;LIU Xiaofeng;LI Yunbo;JIA Yuhao(Nanjing Water Conservancy Planning and Design Institute Co.,Ltd.,Nanjing 210022,Jiangsu,China;School of Water Resources and Hydropower,Hohai University,Nanjing 210024,Jiangsu,China;Shanghai Survey Design and Research Institute Co.,Ltd.,Shanghai 200335,China)
出处
《水力发电》
CAS
2022年第10期68-71,112,共5页
Water Power
基金
国家自然科学基金项目(51579085)。