Study On Time Series Ionosphere Prediction Models And Methods

In the process of GNSS navigation and positioning,the ionospheric delay error is one of the inevitable main sources of error.How to effectively correct the ionospheric delay to improve the navigation and positioning accuracy of GNSS is an important research direction.Total ionospheric electron content(TEC)is one of the most important parameters to characterize the physical characteristics of the ionosphere.The research and prediction of TEC not only have important significance for improving the precision of GNSS measurement and real-time navigation and positioning,but also provide relevant research data for geophysics,space environment and other disciplines.Based on this,this paper studies the changes of the ionospheric TEC,short-term forecasting and improvement of the forecasting model.The main contents and related conclusions are as follows:1.Using GIM grid data provided by IGS,the diurnal variations of ionospheric TEC during the quiet(DOY: 185,2009)and active(DOY: 75,2011)periods of solar activity are analyzed by 5112 grid points around the world.The decomposition and reconstruction algorithm of wavelet is studied.TEC values at different latitudes in the world for 11 years(2007~2017)were extracted by wavelet decomposition.The annual variation of the ionosphere and its influencing factors were studied based on solar activity.The conclusions are as follows:(1)The diurnal variation of ionospheric TEC in active period is larger than that in quiet period,and the diurnal peak value is much larger than that in quiet period.(2)The influence of the solar activity on the ionosphere is related to the latitude,and the degree of influence generally decreases with the increase of latitude.(3)The ionospheric TEC is higher in summer than in winter in the southern and northern hemispheres,and the change trend of TEC is consistent with the change of sunspot relative number.2.The principles and modeling methods of Holt-Winters additive model,multiplicative model and ARIMA model are studied.Based on the 15-day GIM grid data in quiet period,the TEC values above different latitudes and longitudes in the Northern Hemisphere are forecasted for 5 days by three models respectively,and the accuracy of the forecasting values is analyzed systematically.The experimental results show that: The experimental results show that:(1)Overall,the ARIMA model has the highest relative accuracy,the root mean square error is the smallest,and the forecasting effect is the best,followed by the Holt-Winters addition model,and the multiplication model has the lowest precision.(2)The latitude difference between the relative accuracy of the forecast value and the root mean square error is more significant than the longitude difference,and the error distribution of the three models is consistent.(3)On different longitudes,the peak of root mean square error of the three models occurs in the vicinity of 20 degrees north latitude for many times,and the overall performance is that the low latitude region is larger than the high latitude region.3.Several algorithms for calculating ionospheric delay and inversion of ionospheric vertical total electron content using dual-frequency observations are studied.The vertical total electron content(VTEC)of the station is obtained by interpolation of the projection function and polynomial interpolation.The W-ARIMA model is constructed by improving ARIMA model with wavelet decomposition.The VTEC values over 16 GNSS stations in China in 13 days before and after the occurrence of magnetic storms in 2011 were calculated by using phase smoothing pseudorange method.After interpolating missing epoch by time series,the ARIMA and W-ARIMA models are used to forecast the next three days based on the VTEC data of the first 10 days,and the prediction accuracy before and after the model improvement is compared and analyzed.The experimental results show that:(1)The improved model can effectively reduce the prediction residual near the daily maximum of VTEC,and the prediction accuracy is significantly improved compared with the original model.(2)The improvement effect of W-ARIMA on low latitude stations is higher than that of high latitude stations.(3)The residual error and root mean square error of the improved model are generally smaller than those of the original model,and the peak value of root mean square error in China can be greatly weakened.