Study Of Ionosphere TEC Prediction Method

Study on the Prediction Method of Ionosphere TEC and anomaly of time-frequency domain analysis. Based on wavelet analysis, Ionosphere TEC time series are decomposed on wavelet channels every day. Analysis of series which corresponding to the channel by traditional time series models, grey model GM(1,1) and time series model are used to prediction, reconstruct the results and Reduction to Original time series for prediction. The results showed that time series analysis based on wavelet can keep the overall trend of practical data and the position where extreme value are appeared, average relative error was about to9%,feasibility of the method is validated. Depend on the chaos series, verification of chaos characteristics of Ionosphere TEC, calculation of the time delay and embedding dimension on Ionosphere TEC time series which we research on. According to nonlinear and nonstationarity on Ionosphere TEC time series, constructing prediction model by BP neural network, then we optimized neural network by DE. The prediction results are greatly improved.In fact, we were hardly to confirm the factors that caused of the abnormality because of the chaos characteristics. This article put forward on method of anomaly judgment based on Hilbert-Huang transform, Twenty days data before earthquake were transformed by, and analysis of the common characteristics on spectrum, combined with traditional results, Shows that abnormal was occurred. Compared to prediction results of three methods based on this abnormal, none of that got effective prediction results, so we think of that caused of abnormal was instantaneous, shows that small correlation to TEC data before abnormal happened.