Research On The Monitoring And Forecasting Model For Jiangsu Regional Ionosphere TEC

Ionosphere is one of the main error sources which affects the positioning accuracy of GPS. It is a focus for people to research on the monitoring and forecasting model for ionosphere TEC. The long-term monitoring and forecasting models which are suitable to China and surrounding areas have been established in China. However, the ionosphere short-term and regional forecast is still at the initial stage. In the paper, the partial continuous observation data of JSCORS were used to calculate TEC and do the research about ionosphere monitoring and short-term forecasting in Jiangsu region. The main content include the following aspects:1) The inversion principle and calculation processes of ionosphere TEC using the partial continuous observation data of JSCORS. The TEC was obtained by carrier smoothed pseudo-range method and carrier method, the receiver and satellite hardware delays and appropriate projection function were taken into account. Finally, we get TEC from puncture points.2) Construction methods of common experience and fitting ionosphere models were introduced. The regional polynomial model was selected as TEC of ionosphere monitoring model. According to the monitoring time length and regional usability, we can determine the polynomial order number and analysis the precision of polynomial model. The results show that the inner precision lies between 1.28-1.82 TECU and relative precision was higher than 90%. The results also show that the outside precision was1.44 TECU and relative precision was 92.4%.3) Some prediction models of ionosphere TEC were introduced. Time series model of AR(p) and BP neural network model were used in the paper. The paper gives a systematic introduction process of order selection of time series model of AR(p) and evaluating precision. The results show that average overall prediction accuracy of time series model of AR(p) was 92% and mean square error was ±1.60 TECU. Work flow of BP neural network model was described. Before the moment of ionosphere TEC, cosine function and linear function combination algorithm were used as input layer. Pieces of hidden layer were determined by mean square error. The results show that average overall prediction accuracy was 92.9% and mean square error was ±1.4TECU.4) By analysis on principle and method of combination, time series model of AR(p) and BP neural network model were combined to forecast ionosphere TEC. The paper analyses three models’ precision. Average overall prediction accuracy of combined model was 93.4% and mean square error was ±1.36 TECU. Prediction accuracy of combined model was higher than time series model of AR(p) and BP neural network model.5) I got ionosphere TEC graphs in Jiangsu province by combined model.