Improved MEEMD-LSTM Ionospheric Modeling And Its Outlier Detection Research

As an important part of the near-Earth space environment,ionospheric research has become a hot issue in advancing discussions related to global modeling of the ionosphere.Exploring the generation of ionospheric delay errors to provide theoretical and data basis for GNSS signal processing and short-term pre-seismic detection is an important method to improve GNSS positioning and navigation accuracy.This paper constructs a model based on the classical empirical modal decomposition,alignment entropy,vector machine and the related theory of long and short time neural network for the annual variation of 2013 in China region and the short time anomaly in the seismic region,carries out a comprehensive analysis and measurement,and analyzes the regularity and heterogeneous process of ionospheric variation from the spatial and temporal dimensions,and mainly summarizes the following contents.(1)Detailed introduction of the contents of the ionospheric TEC data files detected by the IGS International Data Analysis Center,and analysis of the correlation changes in ionospheric TEC values arising from the occurrence of solar activity,geomagnetic activity,and temporal changes using the ionospheric grid products provided by the IGS International Data Analysis Center.The results of the correlation analysis show that the changes in ionospheric TEC show a strong correlation with solar activity,producing a corresponding activity with the intensity of solar activity.(2)The set of empirical mode decomposition methods is decomposed into EMD decomposition algorithm(empirical mode decomposition),EEMD decomposition algorithm(ensemble empirical mode decomposition)and CEEMD decomposition algorithm(complete ensemble empirical mode decomposition),and the MEEMD decomposition algorithm and the improved MEEMD decomposition algorithm(modified ensemble empirical mode decomposition with SVM)are introduced.We compare the simulations of the real signal with the various methods,and find that the improved MEEMD algorithm can effectively decompose the ionospheric TEC data,improve the modal mixing phenomenon,and enhance the ionospheric measurement efficiency.(3)An improved MEEMD-LSTM ionospheric TEC prediction model is proposed for the spatial and temporal prediction of global ionospheric TEC,and its principle is described in detail.The ionospheric grid products of 2013 are extracted and the ionospheric TEC data of the first nine months are used for modeling,and the test sets made from the remaining three months of modeling data are predicted by the ARMA model and the improved MEEMD-LSTM prediction model,respectively,and the prediction results are analyzed in detail,and it is found that the root-mean-square error(0.02)of the new model is smaller than that of the ARMA model The root mean square error(0.04)of the new model was found to be smaller than the root mean square error of the ARMA model(0.02),which improved the prediction efficiency value on the temporal and spatial resolution of ionospheric data.(4)The improved MEEMD-LSTM ionospheric TEC model is used to detect the changes of ionospheric TEC anomalies before the earthquake on April 20,2013.The measured and analyzed results show that the new method used in this paper effectively reduces the systematic errors in the traditional prediction process,eliminates the fluctuation factors unrelated to the earthquake,and the model can effectively detect the anomalous values in the case of non-seismic characterized ionospheric disturbances near the epicenter area on April 20.Therefore,the pre-earthquake prediction model constructed based on the MEEMD-LSTM prediction model can effectively detect pre-earthquake TEC anomalies,effectively reducing false warnings due to TEC changes caused by non-seismic activities,and thus highlighting TEC changes caused by seismic activities,taking into account the effects and changes of various factors in time and space.