基于经验值分解及Elman神经网络的桥址区风速预测  

Wind Speed Forecasting for Bridge Sites Based on Empirical Mode Decomposition and Elman Neural Network

作　　者：陶齐宇[1] 余传锦 李永乐[2] 张明金[2] 蒋劲松[1] TAO Qiyu YU Chuanjin LI Yongle ZHANG Mingjin JIANG Jinsong(Institute of Highway Planning and Design of Sichuan Provincial Department of Communications and Transportation, Chengdu 610041, China Department of Bridge Engineering, Southwest Jiaotong University, Chengdu 610031, China)

机构地区：[1]四川省交通运输厅公路规划勘察设计研究院,四川成都610041 [2]西南交通大学桥梁工程系,四川成都610031

出　　处：《灾害学》2017年第4期85-89,共5页Journal of Catastrophology

基　　金：交通运输部建设科技计划项目(2014318800240);四川省创新研究团队(2015TD0004)

摘　　要：准确的风速预测对于保障强风区的桥梁及行车安全是十分必要的。但因风速波动性大,非平稳性质强,准确预测较为困难。为提高预测精度,研究中采用EMD-Elman预测模型。将预测性能良好的Elman神经网网络融入经验值分解技术,以降低风速时程的非平稳性质。以大渡河大桥桥址区的实测风速作为算例验证。系统地研究了EMD-Elman模型的预测效果,并将其与Elman神经网络及被广泛采用的持续法和差分自回归移动平均模型进行对比。结果显示,融入经验值分解技术后,EMD-Elman模型预测性能有大幅提升;较Elman神经网络、持续法和差分自回归移动平均模型而言,EMD-Elman模型预测性能最为优越,可用于桥址区风速预测。It is necessary to make accurate wind speed forecasting to ensure safety of bridges and vehicles under strong wind. However,resulting from the great fluctuations and non-stationary of wind speeds,it is difficult to achieve precise predictions. To improve forecasting accuracy,EMD-Elman model is proposed,combined with Empirical Mode Decomposition（ EMD） and Elman neural network,to reduce the non-stationary nature. Real wind speed series collected from the Dadu River bridge site are taken as the experiment subjects. The prediction performance of EMD-Elman model is systematically studied. It is compared with that of the Elman neural network,the persistence method and the autoregressive integrated moving average model,which are all generally used. The results show that the performance of EMD-Elman has a significant enhancement after EMD employed; compared with the others,EMDElman is the best and can be employed for wind speed forecasting for bridge sites.

关 键 词：经验值分解 ELMAN神经网络 桥址区 风速 预测模型 

分 类 号：TU398[建筑科学—结构工程] X43[环境科学与工程—灾害防治]