摘要
由于大坝变形监测数据为非平稳、非线性的时间序列,因此采用经验模态分解法(EMD)、相关向量机理论(RVM)以及Arima误差修正模型对大坝变形监测数据进行分析预测。首先利用EMD分解法对原始时间序列进行分解和重构,使原始序列平稳化,得到若干本征模态函数(IMF)以及残差序列,再以RVM预测模型对上述结果进行分析预测,最后利用Arima误差修正模型对预测残差进行误差修正,从而建立了以RVM为基础预测模型的EMD-RVM-Arima大坝变形预测模型。以某双曲拱坝为例,采用该模型对其变形监测数据进行分析预测,得到的平均残差为2. 89 mm,同时计算出SVM、RVM法的平均残差为11. 62 mm、9. 30 mm。可以看出,EMD-RVM-Arima模型大大提高了预测精度,该模型在大坝变形预测中具有可行性。
As the time series of the dam deformation monitoring data is non-stationary and non-linear,the data is analyzed and predicted herein with the method of empirical mode decomposition(EMD),relevant vector machine theory(RVM)and Arima error correction model,and then an EMD-RVM-Arima-based prediction model of dam deformation is established on the basis of RVM.By taking a double-curvature arch dam as the study case,the model is adopted to analyze and predict its deformation monitoring data,from which the mean residual of 2.89 mm is obtained,while the mean residuals calculated from the methods of SVM and RVM are 11.62 mm and 9.30 mm respectively as well.It can be seen that the EMD-RVM-Arima-based prediction model largely enhances the deformation prediction accuracy,thus is feasible to be applied to the prediction of dam deformation concerned.
作者
曹恩华
包腾飞
刘永涛
李慧
CAO Enhua;BAO Tengfei;LIU Yongtao;LI Hui(State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering,Hohai University,Nanjing210098,Jiangsu,China;National Engineering Research Center of Water Resources Efficient Utilization and Engineering Safety,Hohai University,Nanjing210098, Jiangsu,China;College of Water-Conservancy and Hydropower,Hohai University,Nanjing210098,Jiangsu,China)
出处
《水利水电技术》
CSCD
北大核心
2018年第12期59-64,共6页
Water Resources and Hydropower Engineering
基金
国家重点研发计划(2016YFC0401601)
国家自然科学基金资助项目(51579086
51739003
51479054
51379068
41323001)
江苏省杰出青年基金项目(BK20140039)
江苏高校优势学科建设工程资助项目(水利工程)(YS11001)