Research On Grid Point Prediction Model Of Total Electron Content In Ionosphere

The ionosphere has caused many inconveniences to human activities by affecting the propagation of radio waves,so related researches are richful.Total Electron Content(TEC),as an important parameter for studying the morphology and structure of the ionosphere,is an important research topic in the field of solar-terrestrial space physics.The TEC data acquisition cannot fully meet human needs in terms of timeliness.The TEC prediction model can predict the future TEC value,which can be used in various fields such as navigation and communication,therefore,it has important research significance.For the prediction of TEC grid points,the existing empirical models based on historical data are mainly based on traditional methods such as Autoregressive Moving Average Model(ARMA)or neural network methods.The prediction accuracy of the model is still lacking,and donnt have enough analysis of the model adaptability.In response to the above problems,this paper introduces a deep learning method that has developed rapidly in recent years,and uses the historical TEC,solar activity indexes and geomagnetic activity indexes to design parameter combination experiments to obtain the optimal parameter combination as the model input,then proposes a TEC grid point prediction model based on Gate Recurrent Unit(GRU).The TEC prediction experiment results in the next 24 hours at 60 grid points around the world show that the adaptability of the TEC grid point prediction model in this paper is better in the northern hemisphere than in the southern hemisphere,and the adaptability of the mid and low latitude grid points is better than that of high latitude grid points;the average relative accuracy of the prediction model during the magnetic disturbance period is slightly higher than that during the magnetic quiet period;compared to the recursive neural network(RNN),Long Short Time Memory(LSTM)and Bi-LSTM,the root mean square error(RMSE)of the prediction model in this paper is reduced by 19.2%.In this paper,the proposed prediction model is further applied to the detection of TEC anomalies before the earthquake.We use the prediction model to construct the background values of TEC before the jiuzhaigou earthquake on August 8,2017,and then extraxt the TEC anomalies.The results show that the TEC background value predicted by the model can effectively detect the TEC anomalies at grid points near the epicenter.Compared with the anomaly detection results of the sliding quartile method,it is proved that this method can detect the negative TEC anomalies before the earthquake earlier,and can detect the positive anomalies more positively.The work of this paper provides technical accumulation for the use of deep learning technology to build a TEC grid point prediction model with higher accuracy and better adaptability,and provides useful attempts in applying the prediction model to pre-earthquake TEC anomaly detection.