Prediction of short-wave maximum usable frequency by long short-term memory neural networks based on deep learning techniques

Aiming at the problem of large error of long-term prediction model based on ITU-R P.533 model, this paper proposes a prediction algorithm of long-term and short-term memory neural network based on deep learning technology. This algorithm uses ionospheric data obtained by vertical sounding technology to construct ionospheric model and simulate short-wave propagation path, so as to predict the future medium and short-term maximum usable frequency (MUF) of short-wave communication. Simulation results show that the root mean square error (RMSE) and mean absolute error (MAE) of the predicted maximum usable frequency (MUF) are 1.3666 and 1.0891 for the next 24h.The RMSE and MAE of the MUF for the next 48h are 1.3723 and 1.0771 respectively, and the RMSE and MAE of the MUF for the next 72h are 1.5707 and 1.2519, respectively. LSTM shows good validity and accuracy in ionospheric MUF short-term prediction, and provides valuable reference for ionospheric research and short-wave communication.