项目名称: 基于经验模式分解和跳步-广义回归神经网络的地球定向参数预报
项目编号: No.U1531128
项目类型: 联合基金项目
立项/批准年度: 2016
项目学科: 天文学、地球科学
项目作者: 王琪洁
作者单位: 中南大学
项目金额: 47万元
中文摘要: 针对地球自转变化复杂的时变特性,充分考虑和顾及地球定向参数(EOP)的高频和低频组分对预报的影响,引入经验模式分解方法,采用高效的跳步神经网络组合模型对EOP进行综合预报;研究和探索将大气和海洋角动量时间序列同时引入到地球自转变化预报中,进一步改善EOP的预报精度。本项目:(1)充分考虑和顾及EOP的高频和低频组分,将经验模式分解法引入,减弱或消除高频组分对EOP短期预报的影响;(2)利用跳步时间序列分析模型能够显著地削弱序列两端的端部效应,特别是极大地改善低频组分的分辨率的优良特性,构造跳步神经网络组合模型,发挥两者的优势,对提高预报效率具有重要意义;(3)将大气、海洋影响同时引入EOP预报,对提高EOP预报精度和稳定性具有重要的科学意义。本项目对于丰富EOP预报理论,提高深空探测器导航与追踪精度、维持坐标系统和时间系统,建立我国自主的高精度EOP预报方法和系统具有重要的科学和现实意义。
中文关键词: 地球自转;大气;海洋;经验模式分解;跳步时间序列分析
英文摘要: In view of the complex time-variable characteristics of the Earth’s variable rotation, this proposal pays attention to the influence of high frequency and low frequency components in EOP on its prediction. Therefore, this proposal introduces the Empirical Mode Decomposition(EMD) method and integrates it with the efficient Leap-Step Time Series Analysis - General Regression Neural Network(LSTSA-GRNN) to predict the EOP. The proposal focuses on incorporating the AAM and OAM series into the prediction of the Earth’s variable rotation to improve accuracies of the EOP predictions. It introduces the EMD method to weaken or eliminate the influence of high-frequency components on the EOP short-term forecast. And LSTSA significantly weaken the edge-effect of the EOP series, and greatly improve the resolution of low-frequency components. This proposal will contribute to improve the efficiency, accuracy and stability of EOP prediction by fusing LSTSA and GRNN, and incorporating the AAM and OAM series. This will enrich the theory of EOP prediction, improve the accuracy of deep-space navigation, tracking, and coordinate and time system maintainment, and benefit the construction of our own high accuracy EOP prediction method and system.
英文关键词: Earth Rotation;Atmosphere;Ocean;Empirical Mode Decomposition ;Leap Step Time Series Analysis