Ionospheric TEC forecast model of based on CEEMD and GRNN

Aiming at the problem of non-linear and non-stationary electrons in the ionospheric total electron content (TEC), and high noise caused by multiple factors, a CEEMD-GRNN ionospheric TEC prediction model combining the complementing ensemble empirical mode decomposition (CEEMD) and generalized regression neural network (GRNN) to solve the problem of poor fitting and low prediction accuracy caused by direct use of raw data for prediction. The ionospheric data of 2019 from IGS center for high, middle and low latitude are used. Different day of year data with magnetic storms and without magnetic storms are tested. Results show that the RMSE of low latitude is 0.97 and the relative accuracy is 91.28, which verifies that the accuracy of the CEEMD-GRNN forecast model is higher than that of the EMD-GRNN and the single GRNN model.