摘要
为提高大坝变形预测精度,提出了一种基于经验模态分解(EMD)和支持向量机(SVM)的大坝变形预测新算法(EMD-SVM)。该算法先对大坝位移序列进行经验模态分解,有效分离出隐含在时序中的非线性高频波动成分和低频趋势成分;然后应用支持向量机对各分量进行建模预测;最后叠加各分量预测值得到预测结果。通过算例验证,并与BP神经网络、支持向量机对比分析表明,该算法具有较强的泛化能力和自适应拟合能力,能在一定程度上保证较优的局部预测值和较好的全局预测精度,在大坝变形预测中具有一定的实用价值。
Aiming at improving the prediction accuracy of dam deformation, based on empirical mode decomposition（EMD） and support vector machine （SVM）, a new algorithm of dam deformation prediction was presented. Firstly, thedam deformation sequence is decomposed with EMD and the nonlinear trend of volatility of high frequency component andlow frequency one are separated effectively. And then, the SVM was applied to build a prediction model for each compo-nent. Finally, each component prediction were superposed to determine the forecast value. Compared with BP neural net-work and SVM, the results show that the proposed method has better generalization ability and adaptive fitting effect. Itensures the optimal local prediction and possesses the higher precision forecasting in a certain extent, which can be appliedto dam deformation prediction practically.
出处
《水电能源科学》
北大核心
2014年第12期67-70,共4页
Water Resources and Power
基金
国家自然科学基金项目(41461089)
广西"八桂学者"岗位专项经费资助项目
广西空间信息与测绘重点实验室资助课题(桂科能130511402
桂科能130511407)
关键词
大坝变形
经验模态分解
支持向量机
精度评定
dam deformation
empirical mode decomposition
support vector machine
precision evaluation
作者简介
任超（1974-），男，博士、副教授，研究方向为GPS高精度数据处理及其应用，E-mail：renchao@glut．edu．cn
