共查询到19条相似文献，搜索用时 421 毫秒
1.
2.
3.
4.
针对单一电离层总电子含量(TEC)预报模型存在的缺陷，如受外界因素干扰较大、预报精度随预报时间的增加明显降低等，本文提出一种基于补充集合经验模态分解(CEEMD)电离层TEC组合预报模型。该模型实现电离层TEC预报的关键途径为：首先，利用CEEMD对TEC原始序列进行自适应分解，得到具有不同频率的分量并依据分量复杂度分析结果进行重构；其次，使用广义回归神经网络(GRNN)模型对高频分量进行建模与预报，使用Holt-Winters模型对低频分量进行建模与预报；最后，重构高频分量预报结果与低频分量预报结果得到电离层TEC预报值。根据太阳活动选取两段不同年积日、不同纬度电离层TEC序列进行实验，结果表明本文提出组合预报模型较单一的Holt-Winters模型、GRNN模型预报精度更高，在太阳活动平静期预报结果的平均相对精度为92.83%,在太阳活动剧烈期预报结果的平均相对精度为84.35%,对于长时间TEC预报也具有较好的效果，稳定性高。 相似文献
5.
利用小波变换分解变形监测数据,重构分离得到低频线性项和高频非线性项,线性特征强的子序列使用ARMA模型进行建模预报,非线性特征强的子序列使用神经网络模型进行建模预报,最后合成获得预报值。文中使用隔河岩大坝GPS变形监测数据进行建模预报,分析小波分解层数对预报精度的影响,提出平滑测度确定分解层数的方法,比较ARMA模型或神经网络模型和小波分解下的组合模型的预报精度,结果表明组合模型的预报精度优于单一模型,其精度在0.6mm,最大偏差为1.6mm,平均改善76%和77%。 相似文献
6.
提出了一种基于小波变换和最小二乘支持向量机的卫星钟差预报方法。首先通过小波变换把钟差时间序列分解成具有不同频率特征的分量,然后根据各分量的特点构建不同的最小二乘支持向量机模型进行预报,最后将各分量的预报结果进行叠加得到最终的钟差预报值。实验结果表明,该方法的预报效果优于单一的最小二乘支持向量机模型以及常规的二次多项式模型和灰色系统模型。 相似文献
7.
8.
针对卫星钟差呈现出非线性、非平稳变化的特性,提出了一种基于双树复小波和广义回归神经网络的卫星钟差预报方法。该方法充分利用双树复小波变换的完全重构、近似平移不变性和较好的方向选择性等优良特性,首先把钟差时间序列分解成具有不同频率特征的分量,然后根据各分量的特点构建高频和低频广义回归神经网络进行预报,最后将各分量的预报结果进行叠加得到最终的预报值。以IGS提供的钟差数据为例进行试验,并与单一的灰色模型、最小二乘支持向量机和广义回归神经网络的预报结果进行对比分析。结果表明,该方法具有较高的预报精度,对于较长时间的钟差预报也能保证较好的预报效果,应用于卫星钟差预报可行、有效。 相似文献
9.
基于卫星双向时间频率传递进行钟差预报的方法研究 总被引：1，自引：0，他引：1
用多项式拟合、谱分析、改进的AR模型三种方法对由卫星双向时间频率传递得出的钟差时间序列进行了拟舍和预报分析。为了抵制钟差时间序列中异常值的影响，引入了“抗差等价权”，利用卫星双向时间频率传递得到的1d的钟差，按不问采样率、不同时间跨度进行计算分析。结果表明，抗差估计的预报精度明显高于最小二乘估计；平滑值的预报精度高于采样值；由于钟差时间序列中有明显的周期变化，多项式进行钟差预报的精度不可靠；用谱分析进行钟差预报的精度不高，但可以发现钟差时间序列中的主要周期变化；改进的AR模型预报精度最高，预报6h钟差的RMs在1ns左右。 相似文献
10.
为了提高卫星钟差预报的精度,针对钟差数据中量级较小的误差,提出了一种基于中位数的小波阈值法钟差数据预处理策略。首先,利用小波阈值方法将钟差数据进行分解,得到分解后的高频系数和低频系数。然后,利用中位数法处理各层影响阈值设置的高频系数,通过处理后的高频系数计算阈值,从而提高小波阈值法剔除小异常值的能力。最后,用北斗二号卫星钟差数据进行了验证,结果表明,利用所提方法处理后的钟差数据建模,小波神经网络（wavelet neural network,WNN）模型预报的精度提高约14.1%,预报稳定性提高约19.7%。该方法可以有效剔除钟差历史观测序列中量级较小的误差,改善钟差数据质量,从而提高模型钟差预报的精度。 相似文献
11.
This article presents the application of a multivariate prediction technique for predicting universal time (UT1–UTC), length of day (LOD) and the axial component of atmospheric angular momentum (AAM χ 3). The multivariate predictions of LOD and UT1–UTC are generated by means of the combination of (1) least-squares (LS) extrapolation of models for annual, semiannual, 18.6-year, 9.3-year oscillations and for the linear trend, and (2) multivariate autoregressive (MAR) stochastic prediction of LS residuals (LS + MAR). The MAR technique enables the use of the AAM χ 3 time-series as the explanatory variable for the computation of LOD or UT1–UTC predictions. In order to evaluate the performance of this approach, two other prediction schemes are also applied: (1) LS extrapolation, (2) combination of LS extrapolation and univariate autoregressive (AR) prediction of LS residuals (LS + AR). The multivariate predictions of AAM χ 3 data, however, are computed as a combination of the extrapolation of the LS model for annual and semiannual oscillations and the LS + MAR. The AAM χ 3 predictions are also compared with LS extrapolation and LS + AR prediction. It is shown that the predictions of LOD and UT1–UTC based on LS + MAR taking into account the axial component of AAM are more accurate than the predictions of LOD and UT1–UTC based on LS extrapolation or on LS + AR. In particular, the UT1–UTC predictions based on LS + MAR during El Niño/La Niña events exhibit considerably smaller prediction errors than those calculated by means of LS or LS + AR. The AAM χ 3 time-series is predicted using LS + MAR with higher accuracy than applying LS extrapolation itself in the case of medium-term predictions (up to 100 days in the future). However, the predictions of AAM χ 3 reveal the best accuracy for LS + AR. 相似文献
12.
This paper aims at the prediction of both global mean sea level anomalies (SLAs) and gridded SLA data in the east equatorial Pacific obtained from TOPEX/Poseidon and Jason-1 altimetric measurements. The first prediction technique (denoted as LS) is based on the extrapolation of a polynomial-harmonic deterministic least-squares model describing a linear trend, annual and semi-annual oscillations. The second prediction method (denoted as LS + AR) is a combination of the extrapolation of a polynomial-harmonic model with the autoregressive forecast of LS residuals. In the case of forecasting global mean SLA data, both techniques allow one to compute the predictions of comparable accuracy (root mean square error for 1-month in the future is of 0.5 cm). In the case of predicting gridded SLA data, the LS + AR prediction method gains significantly better prediction accuracy than the accuracy obtained by the LS technique during El Niño 1997/1998, La Niña 1998/1999 and during normal conditions. 相似文献
13.
根据LS模型参数中钱德勒项的时变性质,提出了基于钱德勒参数时变修正的CLS模型,并用CLS+AR对极移进行预测。实验结果表明,CLS+AR模型在极移预测精度上较LS+AR模型有较大改善。 相似文献
14.
Possible improvement of Earth orientation forecast using autocovariance prediction procedures 总被引：3，自引：2，他引：1
Autocovariance prediction has been applied to attempt to improve polar motion and UT1-UTC predictions. The predicted polar
motion is the sum of the least-squares extrapolation model based on the Chandler circle, annual and semiannual ellipses, and
a bias fit to the past 3 years of observations and the autocovariance prediction of these extrapolation residuals computed
after subtraction of this model from pole coordinate data. This prediction method has been applied also to the UT1-UTC data,
from which all known predictable effects were removed, but the prediction error has not been reduced with respect to the error
of the current prediction model. However, the results show the possibility of decreasing polar motion prediction errors by
about 50 for different prediction lengths from 50 to 200 days with respect to the errors of the current prediction model.
Because of irregular variations in polar motion and UT1-UTC, the accuracy of the autocovariance prediction does depend on
the epoch of the prediction. To explain irregular variations in x, y pole coordinate data, time-variable spectra of the equatorial components of the effective atmospheric angular momentum, determined
by the National Center for Environmental Prediction, were computed. These time-variable spectra maxima for oscillations with
periods of 100–140 days, which occurred in 1985, 1988, and 1990 could be responsible for excitation of the irregular short-period
variations in pole coordinate data. Additionally, time-variable coherence between geodetic and atmospheric excitation function
was computed, and the coherence maxima coincide also with the greatest irregular variations in polar motion extrapolation
residuals.
Received: 22 October 1996 / Accepted: 16 September 1997 相似文献
15.
This paper presents normal time–frequency transform (NTFT) application in harmonic/quasi-harmonic signal prediction. Particularly, we use the normal wavelet transform (a special NTFT) to make long-term polar motion prediction. Instantaneous frequency, phase and amplitude of Chandler wobble, prograde and retrograde annual wobbles of Earth’s polar motion are analyzed via the NTFT. Results show that the three main wobbles can be treated as quasi-harmonic processes. Current instantaneous harmonic information of the three wobbles can be acquired by the NTFT that has a kernel function constructed with a normal half-window function. Based on this information, we make the polar motion predictions with lead times of 1 year and 5 years. Results show that our prediction skills are very good with long lead time. An abnormality in the predictions occurs during the second half of 2005 and first half of 2006. Finally, we provide the future (starting from 2013) polar motion predictions with 1- and 5-year leads. These predictions will be used to verify the effectiveness of the method proposed in this paper. 相似文献
16.
To meet the need for real-time and high-accuracy predictions of polar motion (PM), the singular spectrum analysis (SSA) and the autoregressive moving average (ARMA) model are combined for short- and long-term PM prediction. According to the SSA results for PM and the SSA prediction algorithm, the principal components of PM were predicted by SSA, and the remaining components were predicted by the ARMA model. In applying this proposed method, multiple sets of PM predictions were made with lead times of two years, based on an IERS 08 C04 series. The observations and predictions of the principal components correlated well, and the SSA \(+\) ARMA model effectively predicted the PM. For 360-day lead time predictions, the root-mean-square errors (RMSEs) of PMx and PMy were 20.67 and 20.42 mas, respectively, which were less than the 24.46 and 24.78 mas predicted by IERS Bulletin A. The RMSEs of PMx and PMy in the 720-day lead time predictions were 28.61 and 27.95 mas, respectively. 相似文献
17.
针对GNSS站坐标时间序列信噪不易分离的问题,在传统EMD去噪方法的基础上,本文提出了一种联合LMD与EMD的坐标时间序列去噪方法。该方法首先采用LMD分解原始坐标时间序列,基于连续均方误差（CMSE）原则分离高频噪声与低频信号,保持低频分量不变;然后对高频分量进行EMD去噪;最后以2次分解所得低频信号之和作为去噪后时间序列。以仿真数据与8个GNSS基准站实测数据进行试验,通过多种评价指标进行精度评估。结果表明,与传统EMD方法相比,联合LMD与EMD的方法能够更加精确地去除坐标时间序列中的噪声。 相似文献
18.
针对GPS多路径提取过程中EMD算法存在模态混叠效应及小波（Wavelet）去噪局限性的问题,提出了一种基于CEEMD-Wavelet-SavGol模型的多路径提取算法。为了能够充分提取高频和低频中的有用信息,该算法利用完备经验模态分解（CEEMD）对信号进行分解,得到一系列从高频到低频的模态函数（IMF）,并根据模量标准化累计均值法对尺度进行区分,然后分别采用Wavelet和SavGol滤波对高频分量和低频分量进行降噪,将降噪后的IMF进行重构,得到降噪后的信号。最后将该模型用于GPS多路径误差提取的实例中,并与Wavelet、EMD、CEEMD模型进行对比,证明了新模型的有效性。 相似文献
19.
Angular momentum forecasts for up to 10 days into the future, modeled from predicted states of the atmosphere, ocean and continental hydrosphere, are combined with the operational IERS EOP prediction bulletin A to reduce the prediction error in the very first day and to improve the subsequent 90-day prediction by exploitation of the revised initial state and trend information. EAM functions derived from ECMWF short-range forecasts and corresponding LSDM and OMCT simulations can account for high-frequency mass variations within the geophysical fluids for up to 7 days into the future primarily limited by the accuracy of the forecasted atmospheric wind fields. Including these wide-band stochastic signals into the first days of the 90-day statistical IERS predictions reduces the mean absolute prediction error even for predictions beyond day 10, especially for polar motion, where the presently used prediction approach does not include geophysical fluids data directly. In a hindcast experiment using 1 year of daily predictions from May 2011 till July 2012, the mean prediction error in polar motion, compared to bulletin A, is reduced by 32, 12, and 3 % for prediction days 10, 30, and 90, respectively. In average, the prediction error for medium-range forecasts (30–90 days) is reduced by 1.3–1.7 mas. Even for UT1-UTC, where AAM forecasts are already included in IERS bulletin A, we obtain slight improvements of up to 5 % (up to 0.5 ms) after day 10 due to the additional consideration of oceanic angular momentum forecasts. The improved 90-day predictions can be generated operationally on a daily basis directly after the publication of the related IERS bulletin A product finals2000A.daily. 相似文献