Short-term Prediction Of Electron Flux In The Outer Earth Radiation Belt Based On Ensemble Learning

Earth’s Outer Radiation Belt refer to an accumulation of large amounts of high-energy charged particles in the distant regions around the Earth.The electron of Earth’s outer radiation belt will pose a great threat to satellite flight and extravehicular activities of astronauts.Therefore,the study of electron flux in Earth’s Outer Radiation Belt has always been a very important part of space physics.And as artificial intelligence technology has become increasingly popular in recent years,people have begun to apply artificial intelligence technology to the study of electronic flux in Earth’s Outer Radiation Belt.The main content of this paper is the fitting of the electron flux in the Earth radiation belt,and then extends to the prediction of the electron flux in the outer earth radiation belt,and finally applies the same prediction model to the Saturn radiation belt.In order to study the variation of electron flux in the Earth radiation belt over time,artificial intelligence technology is used to predict the electron flux in the Earth radiation belt.The main research in this paper focuses on the short-term prediction of electron flux,which aims to provide a reference for the changes of electron flux in the next ten minutes for the satellite,so that the satellite can make corresponding adjustments in time to avoid the harm caused by high-energy electrons.The predicted energy level range in this paper includes all 20 energy levels provided by RBSP-A satellite(from 31.5ke V to 4216 ke V),which makes up for the small energy level range of previous studies.The overall prediction period was from 2013 to 2017,and more models were used than those used in previous studies,involving deep neural network(DNN),convolutional neural network(CNN),linear regression(LR)and integrated model(Light GBM).After model prediction,the experiment found that the RESULT of DNN model was the worst.However,LR model is better in short-term prediction,but its effect begins to decline with time extension.The Light GBM integrated model has been highly stable among the five models and has been superior to other independent models in terms of prediction accuracy.The above results can be seen from the calculated root mean square error and correlation coefficient.After obtaining the above results and ensuring their reliability in the short term,the effects of adding geomagnetic indices and the duration of adding geomagnetic indices on the prediction results are also studied,and the prediction results of the integrated model are mainly analyzed.It is found that the model is effective in a short time,but with the extension of time,the results will gradually deteriorate until failure.This is the key point of this paper.In order to verify the universality of the prediction model,five electron flux models from earth radiation belts are used to predict the electron flux of Saturn.We used Cassini data from 2004 to 2017 to apply five prediction models to Saturn’s electron fluxes.Following the experiment,it was discovered that the combined CNN and DNN model did not reveal the findings of the experiment on Earth,and that the results were the same as when only CNN was used.Among the five models,DNN is still the poorest.The findings of the LR and THE integrated models are comparable.However,in terms of accuracy,the integrated model’s prediction impact is still superior than that of LR.Unfortunately,we were unable to conduct earth-like geomagnetism follow-up studies due to a lack of data on Saturn’s solar wind.