Accuracy evaluation of ionospheric prediction based on BP neural network model

In view of the characteristics of the high noise, nonlinear and non-stationary dynamic sequence of the total electron content (TEC) time series, based on the BP neural network (BPNN) model, the TEC data of the global ionospheric map (GIM) products provided by the center for orbit determination in Europe (CODE) in the middle and low latitudes, middle latitudes and high latitudes and the corresponding time points, longitude and latitude, solar radio flux F_10.7 data, equatorial geomagnetic activity index D_st The global geomagnetic activity index K_p data were trained and ionospheric prediction was carried out. The results confirmed that the BPNN model based on BP neural network can better predict the low latitude, middle latitude and high latitude ionospheric TEC values, and the average relative accuracy reached 90.5%, 88.7% and 85.35% respectively, the adjustment residuals are 1.505 TECU, 1.595 TECU, and 1.885 TECU, with RMSE values of 1.94 TECU, 2.13 TECU, and 3.08 TECU, respectively.