Construction of TEC prediction model based on LSTM

Abstract

This paper attempts to apply the LSTM model in deep learning to the prediction of Ionospheric TEC parameters and forecasts TEC parameters(static day and storm time) 48 hours in advance. According to the solar F10.7 index, sunspot number(SSN), geomagnetic Dst index, geomagnetic time accumulation index ap(τ), solar wind velocity Vz, and the interplanetary southward magnetic field component(IMF Bz) of the previous 6 days, an empirical model based on LSTM network is established to predict the Ionospheric TEC of Beijing in the next 1–48 hours. The results of the proposed prediction model are compared with those of the traditional neural network model. The LSTM model and BPNN model established in this paper present high-efficiency performance in the prediction of Ionospheric TEC, and the prediction error of the LSTM model is smaller than the BPNN model.