GPS Ionospheric Prediction Research Based On Chaotic Theory

The earth atmosphere is an important component of the close-earth spatial environment. Among it, the ionosphere is greatly influencing the electromagnetic wave signal transmission related engineerig applications, such as navigation, geodetic and telecommunications, etc. With the fast development of the GPS satellite navigation system and the continuous improvement of its positioning and navigation accuracy, the impact of the ionospheric time delay on electromagnetic wave signal has increasingly become a hot topic concerned by human being. The requirement on the ionospheric real-time broadcast and forecast has also become more and more forceful, with the help of GPS.The time series prediction theory and methods are widely applied to every field of the natural science and social economy. For them, many scholars have put forwards various linear and nonlinear methods. Afterwards, some scholars found that there is perhaps a fixed dynamic mechanism for the seemingly random system, and there is perhaps a regular track profile in the seemingly irregular time series data. At the end of the 20s century, the finding of chaotic phenomena and its theory provided a fully innovative thinking way for the analysis and prediction on the nonlinear time series.This thesis mainly studies the chaotic characteristics of ionospheric TEC time series. Its chaotic characteristics are analyzed from both qualitative and quantitative angles. On this basis, the systematically dynamic characteristic of ionospheric TEC is renewed by means of phase space reconstruction. Moreover, the systematic and deep research is performed on the denoising and prediction of the time series.Firstly, some methods and workflows are described on the ionospheric TEC to be calculated and extracted by GPS technology. Regarding the Wenchuan Earthquake, the abnormal variation on TEC data ahead of the disaster is calculated and detected by means of Chongqing GPS-CORS stations. The ionospheric abnormal variation before solar eclipse is analyzed by means of the IGS station observations in China and the past conclusion is proved.Secondly, based on the theory and method of chaotic analysis, we re-construct the phase space for the ionospheric TEC. Its chaotic characteristic can be judged through the re-constructed phase diagram and trajectory Laypunov index and so on. By analyzing the variation rule of the maximum Laypunov index at different latitudes, we preliminarily get the time scale for prediction. The comparison of phase space after TEC data reconstrucation with different time sampling rates shows that their included systematic dynamics information is not wholly the same.Thirdly, we do the time-frequency analysis and multi-scale analysis on the ionospheric TEC data. Combining with some major influencing factors on TEC, we also do the relevant ionospheric analysis. These provide some fundamental information for the next time series analysis and prediction. By utilizing the phase space reconstruction on the ionospheric TEC time series as well as combing the wavelet decompostion in denoising methods research, the denoising result is improved through the full applicaion of the chaotic characteristic of the TEC data itself.Fourthly, according to the local similarity feature theory of the chaotic time series prediction, we use the one-rank local-region method to predict. Comparing with the common methods, we think there is not obvious advantage on time series prediction by chaotic local-region method. Moreover, by means of the neural networks method, we fully use the chaotic characteristic of the time series and set up a chaotic artificial nerual networks for TEC prediction and analysis. The issue related to determining the neural networks structure is solved. And the learning algorithm for training is improved.Finally, based on the multi-variable time series analysis theory, we put the ionospheric TEC and its related variables together for combined prediction. Therefore, the prediction result is improved to some extent.This thesis is a full and deep practice of the chaos theory to the ionospheric TEC analysis. The parameters of the chaotic characeristic being acquired in the time series have also been applied to the data analysis and prediction methods. As far as the ionospheric TEC time series are concerned, some faily good results have been found through three respective aspects, namely data denoising, neural networks prediction model construction and nerual networks learning method. Its research results have an important reference meaning and a practical value for time series data processsing and prediction related work.