题名 | 基于神经网络技术的地球自转变化预报 |
作者 | 王琪洁 |
学位类别 | 博士 |
答辩日期 | 2007-06-13 |
授予单位 | 中国科学院上海天文台 |
授予地点 | 上海天文台 |
导师 | 廖新浩 |
关键词 | 地球自转 人工神经网络 大气角动量 实时预报 ENSO |
其他题名 | Studies on the prediction of Earth’s variable rotation by Artificial Neural Networks |
中文摘要 | Prediction of the variations of the Earth’s variable rotation is of great scientific and practical importance. However, due to the complicated time-variable characteristics of the Earth’s variable rotation, it’s usually difficult to obtain satisfied predictions by conventional linear time series analysis methods. This study employs the non-linear artificial neural networks (ANN) to predict the variations of the Earth’s rotation. As the solid Earth and its surrounding fluid layers form an approximately close dynamic system, changes of atmospheric or oceanic angular momentum will result in variations in the solid Earth's rotation, based on the law of conversation of angular momentum. The high-accuracy Earth rotation observations and researches on global atmospheric models reveal that the axial atmospheric angular momentum (AAM) X3 correlates strongly with the LOD changes and the equatorial AAM X1 and X2 correlate with the polar motion excitation. Thus, when the AAM series is incorporated into the prediction of the Earth’s variable rotation, it will add a physical constraint to the prediction. The present study focuses on incorporating the AAM series into the prediction of the Earth’s variable rotation to improve accuracies of the ERP predictions by ANN. In addition, the technology is also applied to predict the El Nino/ Southern Oscillation (ENSO) event. The main research work of this thesis can be summarized as follows: (1) The algorithms of determining the topology of an ANN is analyzed, and the Root Mean Squared Error (RMSE) is chosen as the criterion to determine the topology of the network. The network algorithm flow that is suitable for our work is investigated. (2) Based on the ANN technique, the variations of the Earth’s rotational rate (i.e., length of day, LOD) are predicted in three ways, i.e., using LODR only, using both LODR and AAM data, and real-time rapid approach, respectively. a) The results show that ANN has effective non-linear prediction ability. b) As the atmosphere is the main excitation source of the LOD change, the accuracies of predictions are significantly improved after introducing the AAM into the LOD prediction, especially for the long prediction intervals. c) Real-time rapid prediction is of great scientific and practical importance. In this thesis we introduce the operational prediction series of , which is from National Centers for Environmental Prediction (NCEP), to the prediction set of ANN model for the first time. The results show that our work about real-time rapid prediction is successful. (3) Based on the ANN technique, the polar motion (PM) are predicted in two ways, i.e., using PM only and using both PM and AAM data. The results show that for 1 to 7 days forward prediction, the accuracy is not improved after introducing the AAM data, but for 8 to 40 days forward prediction the accuracy is significantly improved after introducing the AAM data. Because the ocean and the ground water are also the polar motion excitation sources except for the atmosphere, it awaits further investigations for incorporating the atmosphere, oceans and ground water into the polar motion prediction. (4) Since June of 2006, the sea surface temperature anomaly (SSTA) of the west Pacific has exceeded 0.5oC for 4 continuous months, it maybe evolve into a new ENSO event. We apply the up-to-date SSTA data to predict this event by AAN. We predict the SSTA of Nino3.4 sea area, based on the monthly data released by the Climate Prediction Center (CPC) of NECP. We conducted 8 predictions from September, 2006 to April, 2007. Every prediction was made 12 months forward. We compare our results with those of CPC. Our work is closer to the average of all dynamical and statistical models. This demonstrates that our prediction has certain reliability and reference values. |
英文摘要 | 地球自转变化的预报具有重要的科学意义和实际应用价值。然而由于地球自转变化复杂的时变特性,传统的线性时间序列分析方法往往难以取得良好的预报效果。本文采用非线性的人工神经网络技术预报地球自转变化。 由于固体地球及环绕着它的流体圈层构成一个近似封闭的动力学系统,角动量守恒原理表明,大气或海洋角动量的任何变化都会影响固体地球的自转变化。现代测地技术获得的高精度地球自转变化和全球大气环流模式的研究结果表明,与日长变化成强相关的是大气角动量函数的轴向分量X3 ,与极移激发相关的是大气角动量函数的赤道向分量X1 、X2 。将大气角动量时间序列引入到地球自转变化预报中,相当于增加一个物理约束条件。正是基于此,本文着重研究和探索应用非线性的神经网络技术,将大气角动量时间序列引入到地球自转变化预报中,改善地球自转参数(ERP)的预报精度,以及应用神经网络技术预报El Nino/南方涛动(ENSO)事件。 本文主要研究内容如下: (1)分析了神经网络的拓扑结构算法,提出选用最小均方误差法确定网络的拓扑结构。研究探讨了适合于本研究的网络算法流程。 (2)基于神经网络技术,对地球自转速率变化分别进行了单独、联合和实时快速预报。a)结果证实了神经网络具有良好的非线性预报能力;b)大气是日长变化主要的激发源,联合大气角动量序列作日长变化预报,结果表明,预报精度得到显著的提高,特别是对于时间跨度较大的预报;c)实时快速预报具有重要的科学意义和实际应用价值,本文首次在网络的预报过程中引入美国环境预报中心(NCEP)全球分析与预报系统的 实时预报序列,结果表明,本工作所作的联合实时预报是成功的。 (3)基于神经网络技术,对极移分别进行了单独和联合预报。联合预报和单独预报相比,对于时间跨度为1-7日的预报,联合预报的精度略低于单独预报;但从8-40日,联合预报的精度均明显高于单独预报。由于大气仅是极移的主要激发源之一,海洋和地下水等激发源由于资料的原因目前尚未计及,所以,综合考虑大气、海洋和地下水等激发因素对极移进行联合预报,值得作深入的研究和探讨。 (4)2006年6月起,西太平洋海表水温的异常变化(SSTA)已连续4个月超过0.5oC,极有可能演化成为新的ENSO事件。我们利用最新的SSTA观测资料,应用神经网络技术,对此次ENSO事件进行试验预报。本文根据CPC每月发布的最新观测数据,在2006年9月至2007年4月期间,对赤道太平洋Nino3.4海区的SSTA进行神经网络的非线性预报,每次都预报12个月的SSTA值。通过与CPC网站发布的预报结果相比较,我们的结果接近于所有动力学模型和统计学模型的平均值。这表明我们的预报具有一定的可靠性和参考价值。 |
语种 | 中文 |
公开日期 | 2011-07-01 |
页码 | 115 |
内容类型 | 学位论文 |
源URL | [http://119.78.226.72//handle/331011/14658] ![]() |
专题 | 上海天文台_中国科学院上海天文台学位论文 |
推荐引用方式 GB/T 7714 | 王琪洁. 基于神经网络技术的地球自转变化预报[D]. 上海天文台. 中国科学院上海天文台. 2007. |
个性服务 |
查看访问统计 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论