基于变分模态分解的GNSS坐标时序降噪方法

GNSS Coordinate Time Series Denoising Method Based on Variational Mode Decomposition

摘要: 针对GNSS基准站位置时间序列中叠加复杂噪声，且很难将其有效剥离，本文尝试用变分模态分解(VMD)对某地区GNSS基准站时间序列做去噪处理。首先，通过搜寻变分模态的最优解，得到若干模态分量。其次，剔除模态分量中的高频信号，进行模态函数的叠加与重构，得到去噪后的站点时间序列。最后，用均方根误差(RMSE)指标评价去噪效果。结果表明，变分模态分解(VMD)能有效地分离序列中的噪声信号，削弱其对GNSS基准站时间序列的影响，降低非构造信息的影响。为分析预报人员提供可靠、高精度的地壳运动信息，更好地为地震预报服务。

Abstract: In view of superimposed complex noise in GNSS reference station position time series and it is diffi-cult to effectively peel it off, this paper attempts to use Variational Mode Decomposition (VMD) to denoise the time series of GNSS reference stations in a certain region. First, we search for the opti-mal solution of variational modes and get a series of mode components. Second, we eliminate the high-frequency signal in mode components, the superposition and reconstruction of mode functions are carried out, and get the time series after denoising. Finally, the Root Mean Squared Error (RMSE) index is used to evaluate the denoising effect. The results show that the Variational Mode Decompo-sition (VMD) could effectively separate noise signals from series, weaken its influence on the GNSS reference station time series and reduce the impact of nonstructural information. It can provide re-liable and high-precision crustal movement information for analysis and prediction personnel, and better serve earthquake prediction.