Abstract：In view of the nonlinear and high noise characteristics of ionospheric total electron content(TEC), we establish a short-term ionospheric combination prediction model based on empirical wavelet transform(EWT) and Elman neural network. We use the model to forecast the ionospheric TEC time series in different geomagnetic environments. The results show that EWT-Elman combination model can reflect the variation characteristics of ionospheric TEC. The average relative accuracy of the combination model during geomagnetic quiescence is 93%, and the root mean square error is 1.04 TECu. During geomagnetic disturbance, the average relative accuracy is 92.4%, and the root mean square error is 2.18 TECu. The highest average relative accuracy of the single Elman model, EMD-Elman model and EWT-BP model is 90.7%, and the minimum root mean square error is 1.33 TECu during geomagnetic quiescence. The highest average relative accuracy is 90.7%, and the minimum root mean square error is 2.57 TECu during geomagnetic disturbance. Compared with other models, the method in this paper has the best prediction effect.
