基于信号分解高频分量再处理策略的大坝变形预测方法  

Dam Deformation Prediction Method Based on Signal Decomposition High-frequency Component Reprocessing Strategy

作　　者：余代广 方勇 周煜 YU Daiguang;FANG Yong;ZHOU Yu(Nanjing Water Conservancy Planning and Design Institute Co.,Ltd.,Nanjing 210014,Jiangsu,China)

机构地区：[1]南京市水利规划设计院股份有限公司,江苏南京210014

出　　处：《水力发电》2024年第11期93-99,共7页Water Power

基　　金：江苏省水利科技项目(2021068,2022011)。

摘　　要：准确预测大坝变形有助于掌握变形的变化规律和发展趋势,是保证大坝安全稳定运行的关键。由于大坝变形具有波动性,将给变形预测精度带来不利影响,为此,提出了基于信号分解技术的大坝变形预测方法。首先利用VMD方法将原始变形数据分解为不同频率的变形分量,以降低变形数据的波动性;由于初次分解得到的高频变形分量仍具有一定的不确定性,其后采用EMD方法对所有的高频分量进行再处理,从而得到相对稳定的变形分量集合;最后采用长短期记忆神经网络对所有分量进行预测并叠加,以得到最终的变形预测结果。实例分析表明,基于信号分解技术的大坝变形预测精度高于单次分解对应的变形预测结果,且其预测精度显著高于传统变形预测方法。Accurate prediction of dam deformation is helpful to grasp the change rule and development trend of dam deformation,and is the key to ensure the safe and stable operation of dam.Because the dam deformation has a certain degree of fluctuation,it will adversely affect the prediction accuracy of deformation.Therefore,a dam deformation prediction method based on signal decomposition technology is proposed.Firstly,the original deformation data is decomposed into deformation components of different frequencies using VMD method to reduce the fluctuation of deformation data.Then,the EMD method is used to reprocess all the high-frequency components since the high-frequency deformation components obtained from the initial decomposition still have a certain degree of uncertainty,so as to obtain the relatively stable set of deformation components.Finally,the long and short-term memory neural network is used to predict all the components and superimpose them to obtain the final deformation prediction.The example analysis shows that the deformation prediction accuracy of the proposed method is higher than that of single decomposition,and its prediction accuracy is significantly higher than that of the traditional deformation prediction methods.

关 键 词：大坝变形预测 信号分解 高频分量 再处理策略 VMD EMD LSTM 

分 类 号：TV698.11[水利工程—水利水电工程] TV642