基于WA-LSTM-ARIMA的混凝土坝变形组合预测模型  

Concrete Dam Deformation Combination Prediction Based on WA⁃LSTM⁃ARIMA

作　　者：周兰庭[1] 柳志坤 徐长华 ZHOU Lanting;LIU Zhikun;XU Changhua(College of Water Conservancy and Hydraulic Engineering,Hohai University,Nanjing 210098,China;Jingzhou Yangtze River Channel Administration Bureau,Jingzhou 434000,China)

机构地区：[1]河海大学水利水电学院,江苏南京210098 [2]荆州市长江河道管理局,湖北荆州434000

出　　处：《人民黄河》2022年第1期124-128,共5页Yellow River

基　　金：国家自然科学基金资助项目(51209078)。

摘　　要：针对混凝土坝变形实测数据序列的不规律性和预测精度欠佳等问题,基于复合建模思想提出一种基于WA-LSTM-ARIMA的大坝变形组合预测模型。首先通过小波多分辨率分析对原始监测序列进行多尺度分解,从中提取高频周期性分量、低频趋势性分量和高频随机性分量;然后将去噪处理后的随机分量与高频周期性分量融合得到综合高频序列,并使用LSTM进行建模预测,对于低频趋势性分量则应用ARIMA模型进行预测,将两组预测结果叠加后即可得到最终的坝体变形预测结果;最后通过工程实例证明该模型所得预测值与实测值拟合较好,与传统的静态模型预测结果对比表明,该模型的预测精度更高。In view of the concrete dam deformation of the observed sequence regularity and poor precision issues,this article which was based on composite modeling thought put forward based on WA⁃LSTM⁃ARIMA combination forecast method of dam deformation.Firstly,the wavelet multi⁃resolution analysis was carried out on the original monitoring sequence to extract the high⁃frequency periodical component,low⁃frequen⁃cy trend component and high⁃frequency random component;secondly,the random component after denoising was fused with the high⁃frequen⁃cy periodical component to obtain the comprehensive high⁃frequency sequence and LSTM was used for modeling and prediction.For the low⁃frequency trend component,ARIMA model was used for prediction and the final dam deformation prediction result could be obtained after the superposition of the two groups of prediction results.Finally,an engineering example proved that the predicted value and measured value of the model fitted well and the comparison with the traditional static model showed that the model had better fitting prediction ability.

关 键 词：混凝土坝变形 小波分解 LSTM ARIMA 组合预测 预测精度 

分 类 号：TV698[水利工程—水利水电工程]