Research On The Short-Term Forecast Of Ionospheric TEC And Its Anomaly Detection Based On Combined Model

Total Electron Content(TEC)is one of the indicators describing the density of free electrons in the ionosphere,which is a region of the Earth’s atmosphere with a high degree of ionization,and the magnitude of its value is a measure of the dramatic spatial and temporal variability of the ionosphere.Therefore,its prediction and analysis are a hot topic in ionospheric research at present.In this paper,we use a Q-learning algorithm based on reinforcement learning to optimize the combination of genetic algorithm-optimized BP neural network model and long and short-term memory network model and then establish a combined deep learning ionospheric TEC prediction model and its application in seismic anomaly detection.The main contents of the paper and related conclusions are as follows(1)The principles of two forecast models,the GA-BP neural network and LSTM model are described,and the applicability analysis is carried out using ionospheric TEC data.The experimental results show that good forecasting results can be achieved when using a sample sequence length and forecast duration of about 5:1(i.e.,25 d is selected as the training sample and 5 d after the forecast).Secondly,the ionospheric data of strong magnetic storm period,medium magnetic storm period,weak magnetic storm period,and no magnetic storm period are selected and tested with the GA-BP model and LSTM model to analyze the applicability of the neural network in the application of ionospheric TEC short-term forecasting.Ionospheric variability.(2)To address the complex stochastic and fluctuating factors affecting the ionosphere and the inherent limitations of a single forecast model that cannot forecast the best value under multiple factors,a combined model based on a reinforcement learning Q-learning algorithm that combines a genetic algorithm optimized BP neural network model and a long and short term memory network model is proposed and its principles are explained.Finally,the combined model is used to forecast the ionospheric TEC data for different solar activities,different geomagnetic environments,and different seasons,and the results are compared by using the GA-BP model and the LSTM model to forecast the total electron content of the ionosphere.The results show that the combined model can effectively integrate the advantages of a single model and obtain better forecasting results and higher stability in different environments.(3)The combined model is applied to detect the ionospheric anomalies during the earthquake period.The ionospheric data during the Balkan earthquake are analyzed,and the combined model is used to detect the anomalies and compare them with the traditional sliding quadrature distance method for anomaly detection.The experimental results show that the combined model can effectively detect TEC anomalies before the earthquakes with the maximum exclusion of geomagnetic and solar activities.