首页 | 期刊简介 | 编委会 | 投稿指南 | 常用下载 | 联系我们 | 期刊订阅 | In English
引用本文:顾小丽,李培良,谭海涛,张婷婷,李 磊,王雪竹,于宜法.基于RBF 神经网络的EMD 方法在海平面分析中的应用.海洋与湖沼,2009,40(5):532-539.
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  下载PDF阅读器  关闭
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 2172次   下载 2995 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于RBF 神经网络的EMD 方法在海平面分析中的应用
顾小丽1, 李培良1, 谭海涛2, 张婷婷1, 李 磊1, 王雪竹1, 于宜法1
1.中国海洋大学海洋环境学院;2.中海石油研究中心
摘要:
采用径向基函数神经网络法延拓原始数据序列, 有效抑制了EMD 分解中出现的端点发散效应, 从而实现准确的EMD 分解。利用该方法对中国近海验潮站的月平均海平面资料进行处理, 分解得到的内在模函数分量代表了海平面各种周期性变化。通过EMD 分解得到的总体自适定趋势项为非线性变化, 比以往趋势项提取方法更有优势, 它反映了在资料长度内海平面的长期升降情况。数据序列越长, 该方法所能分解出来的IMF 成分越多, 可分辨的频率越小。
关键词:  EMD, 径向基函数, IMF 分量, 海平面变化
DOI:10.11693/hyhz200905002002
分类号:
基金项目:教育部科学技术重点项目,108159 号;国家自然科学基金资助项目,40506006 号和40676013 号;国家重点基础研究发展计划(973)项目, 2007CB411807 号。
附件
SEA LEVEL VARIATION ANALYSIS WITH RBF NEURAL NETWORK BASED EMD METHOD
GU Xiao-Li1, LI Pei-Liang1, TAN Hai-Tao2, ZHANG Ting-Ting1, LI Lei1, WANG Xue-Zhu1, YU Yi-Fa1
1.Oceanography Department, Ocean University of China;2.China National Offshore Oil Research Center
Abstract:
A new empirical mode decomposition (EMD) method was developed based on radial basic function (RBF) neural network and applied to analyze sea level variability. A temporal signal series is extended with the network before a new round of EMD starts. The result of analysis with the new method agrees better with the result of 5 year running mean than that using traditional EMD. The new method decomposes a series of sea level temporal signal into a set of intrinsic mode functions representing the periodic components of signal and an overall adaptive trend. It was shown that the annual variability was very remarkable in 47 stations along coast of China, and the semi-annual variability was also clear but less remarkable, with the amplitude decreased north to south. In the last 40 years, the sea level fluctuation of the area has been increasing in a non-linear pattern generally but varied geographically. The maximum fluctuation occurs in estuary zones. The results also reveal that the longer the temporal signal series is, the more IMF can be obtained from the data using the new method and a more exact overall adaptive trend can be extracted from signal. A shorter time series could achieve a complete decomposition easily but some of the IMF components may be hidden in overall adaptive trend, resulting in obviously nonlinear sea level variability, which may not be actually true.
Key words:  EMD, Radial Basic Function, IMF, Sea level variations
版权所有 海洋与湖沼 Oceanologia et Limnlolgia Sinica Copyright©2008 All Rights Reserved
主管单位:中国科协技术协会 主办单位:中国海洋湖沼学会
地址:青岛市福山路32号  邮编:266071  电话:0532-82898753  E-mail:ols@qdio.ac.cn
技术支持:北京勤云科技发展有限公司