基于变分模态分解的GNSS位置时间序列去噪分析

Denoising Analysis on GNSS Position Time Series Based on VMD

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的情况,尝试用变分模态分解(VMD)对山西地区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,得到若干的模态分量;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V去噪效果。结果表明,变分模态分解(VMD)能有效地分离序列中噪声信号,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、高精度地壳运动信息,更好地为地震预报服务。

The time series of GNSS reference station superimposed complex noise and it is difficult to effectively denoise,this paper uses VMD to denoise the time series of GNSS reference stations in Shanxi province.First,we search for the optimal solution of variational modes and obtain a series of modal components;second,we eliminate the high-frequency signal in modal components,carry out the superposition and reconstruction of modal functions and obtain the denoised time series of station positions;finally,RMSE index is used to evaluate the denoising effect.The result show that the VMD method can effectively separate noise signals from series,reduce the impact of non-structural information and improve the accuracy of GNSS reference station time series.It provides reliable and high-accuracy crustal movement related information for analysis and prediction personnel,and better service for earthquake prediction.