基于EEMD-SE-LSTM的混凝土坝变形监测模型  

Deformation monitoring model of concrete dams based on EEMD-SE-LSTM

作　　者：侯回位 郑东健[1] 刘永涛 黄寒冰 HOU Huiwei;ZHENG Dongjian;LIU Yongtao;HUANG Hanbing(College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098,China)

机构地区：[1]河海大学水利水电学院,江苏南京210098

出　　处：《水利水电科技进展》2022年第1期61-66,共6页Advances in Science and Technology of Water Resources

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

摘　　要：为提高混凝土坝变形监测数据的预测精度,构建了一种基于集成经验模态分解(EEMD)与样本熵重构(SE)的长短期记忆网络(LSTM)预测模型。模型利用EEMD对原始数据序列进行分解,并计算每个分量序列的样本熵,以原始序列样本熵作为基准进行重构,再对重构后的各序列建立LSTM模型进行预测,最后把各预测值叠加以得到最终预测结果。以某混凝土拱坝为例,将该模型预测结果与EMD-LSTM、LSTM和SVM模型的预测结果进行对比,结果表明EEMD-SE-LSTM模型具有更高的预测精度,在混凝土坝的变形预测中具备更好的可行性与优越性。To improve the prediction accuracy of concrete dam deformation monitoring data,a long and short-term memory network(LSTM)prediction model was constructed based on integrated empirical mode decomposition(EEMD)and sample entropy reconstruction(SE).In this model,EEMD is firstly used to decompose the original data sequence,and the sample entropy of each component sequence is then calculated.The sample entropy of the original sequence is used as the reference for reconstruction,and then the LSTM model is established to predict the reconstructed sequences.Finally,the predicted values are superimposed to obtain the final prediction result.Taking a concrete arch dam as an example,the prediction results of the EEMD-SE-LSTM model were compared with those of EMD-LSTM,LSTM and SVM models.The results show that the EEMD-SE-LSTM model has higher prediction accuracy and better feasibility and superiority in the prediction of concrete dam deformation.

关 键 词：集成经验模态分解 长短期记忆神经网络 样本熵 变形预测 混凝土坝 

分 类 号：TV698.1[水利工程—水利水电工程]