Application Of Empirical Mode Decomposition In The Multi Scale Analysis And Prediction Of Ionospheric TEC

Ionosphere is an important area to human in the field of space physics research and ionospheric Total Electron Content(TEC) is a vital parameter to perform the variation and distribution of ionosphere of different time. Therefore, the predicted analysis of ionospheric TEC is significant in the study of the ionosphere. In this paper, the forecast model is established through the analysis of ionospheric TEC and its influence factors to predict the short-term change of ionospheric TEC. The main contents of this paper are as follows:Firstly, the purpose and significance of analysis and prediction for the ionosphere and ionospheric TEC are discussed. Then, basic structure of ionosphere and commonly used detection methods are described, calculation method of ionospheric TEC and the basic situation of IGS center which is the TEC data source in this paper are explained. The extraction of ionospheric TEC data is realized through the interpretation of the data format of ionospheric data file IONEX provided by IGS. Based on the spectrum analysis function of empirical mode decomposition, short period variation of ionospheric TEC is analyzed by using the TEC data in 2009 and we can obtain that TEC has the short cycle characteristics of 162 days, 58 days and 27 days.Meanwhile, cycle characteristics of the main influence factors on Ionospheric TEC which obtains the sunspot number and the geomagnetic Dst index is analyzed by using the EMD method. Through the analysis, it is concluded that the SSN contains 22 years, 11 years, 1 years long cycle characteristics and 27 days short cycle characteristics, DST index contains 33 years, 11 years, 1 year, half year long cycle characteristics and 54 days, 27 days short cycle characteristics. In addition to the level of solar activity and geomagnetic index, from two global ionospheric TEC distribution maps of the same time in different cumulative annual day and the same cumulative annual day in different time, conclusions are drawn that ionospheric TEC is also affected by local time, cumulative annual day, geographical location and other factors. The correlation between ionospheric TEC and its influence factors is analyzed by using correlation coefficient method. Meanwhile, correlation between IMF and its influence factors is analyzed in the same way. It is conducive to the analysis of the correlation between the factors and TEC.In this paper, the EMD-wavelet denoising method is improved by introducing permutation entropy, and the denoising preprocessing of ionospheric TEC data is realized. Finally, based on ionospheric TEC analysis and its influencing factors analysis, the single factor model, multi factor model and multi factor-multi scale model based EMD are established on the basis of Elman neural network method. Compared with the time series ARMA model, it is concluded that the ARMA model can get better prediction results when the TEC sequence changes smoothly, but the result is poor in the case of large fluctuation of TEC variation. Compared with the ARMA model, the single factor model has an average increase of 2 percentage points in the accuracy of the prediction error of less than 3TECU, the multi factor model not only has a great improvement in the stability, but also it has an average increase of 4 percentage points in the accuracy of the prediction error of less than 3TECU, the multi factor-multi scale model is also improved in stability, and the accuracy of the error less than 3TECU is increased by 6 percentage points.