Research On GNSS Coordinate Time Series Modeling And Noise Reduction Method Based On Empirical Wavelet Analysis

The current research hotspot of GNSS coordinate time series data processing is the processing and analysis of non-smooth,non-linear,noise-containing coordinate time series signals using statistical mathematical theory and signal processing techniques.The complex GNSS coordinate time series not only contain inherited tectonic motions that can reflect the observatory under the plate tectonic stress field,but also have the influence of nonlinear variations generated by geophysical and other factors.The use of reasonable signal processing methods can further understand many unexplained geodynamic phenomena,which is of great significance to the study of Earth plate deformation,geophysics,geometeorology and other fields.Due to the different geographical environment of each GNSS station and the joint influence of many factors such as changes in satellite observation conditions,the noise percentage and noise type of GNSS coordinate time series differ to some extent,which have different effects on coordinate time series seasonal analysis and accurate prediction.Based on this,this paper takes the complex nonlinear GNSS coordinate time series as the research object,and carries out the research work of time series signal seasonality analysis and noise reduction.In addition,building a reasonable prediction model to accurately predict the future trend and seasonal oscillation changes of GNSS coordinate time series is also the main research direction of coordinate time series analysis.This paper establishes a GNSS elevation time series prediction model with strong adaptive capability,which can provide theoretical and technical reference significance for geohazard early warning and plate motion analysis,and further expand the scope of GNSS coordinate time series application.The specific research work of this paper is as follows:(1)A signal extraction method based on the combination of empirical wavelet transform and multiscale permutation entropy is studied for the difficult extraction of annual and semiannual seasonal signal features in GNSS coordinate time series.The simulated experimental results show that the extracted components of the method are in good agreement with the constructed components;the measured experimental results show that the extracted results have higher energy proportion and stronger signal volatility,which can better reflect the annual and semi-annual signal characteristics in the original signal and have better applicability.(2)A new adaptive noise reduction method combining empirical wavelet transform and non-local mean filtering with sample entropy optimization is proposed to address the problems that EWT cannot determine the appropriate high-frequency noise component boundaries in the noise reduction process and some effective signals are lost in the mixed components.The experimental results show that this method has better noise reduction effect and better stability and applicability than the traditional algorithm,and can better present the local motion trend changes and the smaller amplitude of the periodic oscillation changes in the original signal.(3)To address the problems of poor Prophet model decomposition and poor nonlinear signal prediction,an improved Prophet prediction method using EWT is proposed.The method uses the multi-resolution analysis capability of EWT to improve the weaknesses of Prophet decomposition effect and nonlinear signal prediction,so as to improve the prediction accuracy.Firstly,the simulation data and some real measurement data are used to determine the flexibility parameters of the Prophet model using retrospective prediction,and the feasibility of the method is verified.The experimental results show that the improved method has higher prediction accuracy and more stable prediction effect than the traditional method and single Prophet model,and the improvement of prediction accuracy is more obvious in the short-term prediction.