基于CEEMDAN-GMDH-ARIMA的大坝变形预测模型研究  

Research on Dam Deformation Prediction Model Based on CEEMDAN⁃GMDH⁃ARIMA

作　　者：程小龙 张斌 刘相杰 刘陶胜[1] CHENG Xiaoong;ZHANG Bin;LIU Xiangjie;LIU Taosheng(College of Civil and Surveying and Mapping Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China)

机构地区：[1]江西理工大学土木与测绘工程学院,江西赣州341000

出　　处：《人民黄河》2024年第1期146-150,共5页Yellow River

基　　金：国家自然科学基金青年科学基金项目(42004158)。

摘　　要：为提高大坝变形预测精度,针对大坝变形数据的复杂性和非线性等特征,基于自适应噪声完备集成经验模态分解(CEEMDAN)、数据处理群集法(GMDH)和差分自回归移动平均模型算法(ARIMA)进行大坝变形预测研究。采用CEEMDAN将大坝变形原始数据分解为高频随机分量、中频周期分量和低频趋势分量,再分别采用GMDH模型、ARIMA模型对高中频分量、低频分量进行预测,建立基于CEEMDAN-GMDH-ARIMA的大坝变形预测模型。以江西上犹江水电站为例,将该模型预测结果与反向传播(BP)、径向基函数(RBF)、GMDH和CEEMDAN-GMDH模型的预测结果进行对比分析。结果表明:CEEMDAN-GMDH-ARIMA模型的均方根误差(E_(RMS))、平均绝对误差(E_(MA))、相关系数(r)分别为0.048 mm、0.035 mm、0.994,均优于BP、RBF、GMDH、CEEMDAN-GMDH模型,模型预测效果最好,能够很好地体现监测点水平位移变化趋势。In order to improve the accuracy of dam deformation prediction,in view of the complexity and nonlinear characteristics of dam de-formation data,a Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN),Group Method of Data Handling(GMDH)and Autoregressive Integrated Moving Average Model(ARIMA)were used to conduct research on dam deformation prediction.CEEMDAN was used to decompose the original dam data deformation into high-frequency random components,medium-frequency periodic components and low-frequency trend components,and then the GMDH model and ARIMA model were used to predict the high-medium fre-quency components and low-frequency components respectively.Dam Deformation Prediction Model Based on CEEMDAN-GMDH-ARIMA was established.Taking Jiangxi Shangyoujiang Hydropower Station as an example,the prediction results of this model were compared with the prediction results of back propagation(BP),radial basis function(RBF),GMDH and CEEMDAN-GMDH models.The results show that the root mean square error(ERMS),mean absolute error(EMA),and correlation coefficient(r)of the CEEMDAN-GMDH-ARIMA model are 0.048 mm,0.035 mm,and 0.994,respectively,which are superior to the BP,RBF,GMDH,and CEEMDAN-GMDH models.The model has the best prediction performance and can well reflect the trend of horizontal displacement changes at monitoring points.

关 键 词：自适应噪声完备集成经验模态分解 数据处理群集法 差分自回归移动平均模型算法 大坝 变形预测 江西上犹江水电站 

分 类 号：TV698.11[水利工程—水利水电工程]