Study On Modeling Of The Auroral Oval Boundary And Intensity And Prediction Of Auroral Substrom

Auroral oval is closely related to solar wind,magnetosphere,ionosphere and their coupling process.The boundary and intensity of auroral oval are important geophysical parameters.Therefore,it is very important to establish an accurate auroral oval prediction model for the study of space physics.When the earth's magnetosphere is active strongly,the nightside area of the auroral oval appear to light up and expand,and then return to the normal level,that auroral activity events is known as auroral substorm.Auroral substorm have a great influence on human activities.It can affect the communication in high latitudes.Based on this,this paper conducts a detailed analysis and research on the modeling method of auroral oval and the prediction method of auroral substorm.To address the problem of low accuracy of existing models in predicting auroral oval boundary,this paper proposes an auroral oval boundary model based on neural network.In this paper,two neural networks,error Back Propagation(BP)neural network and Generalized Regression Neural Network(GRNN),are used to model the boundary of auroral oval with the spacial parameters as input.Experimental results show that the auroral oval boundary model based on GRNN improves the accuracy of auroral oval boundary prediction.In view of the low prediction accuracy in the current auroral oval intensity model,this paper constructs the auroral oval intensity model based on the GRNN with spatial parameters as input.By extracting appropriate auroral oval intensity features and using GRNN to model the auroral oval intensity.Experimental results show that the model based on GRNN and grid features can be used to predict the intensity distribution of auroral oval.In order to avoid extracting the auroral oval intensity features and improve prediction accuracy of aurora oval intensity,this paper constructs the aurora oval intensity model based on Generative Adversarial Networks(GAN).To make the predict result more consistent with the real data,L1 regular term and structure similarity loss function term are added into the loss function of the network.The auroral oval intensity prediction model is obtained through the training of GAN.Experiment results show that aurora oval intensity model based on GAN improve the accuracy of auroral oval intensity prediction.The current predict method based on auroral images is not suitable for the data of low-orbit satellites with low temporal resolution and high spatial resolution.This paper presents a new method to predict the substorm using the auroral images with low temporal resolution and high spatial resolution.This paper determines whether the auroral image is in the stage of substorm by detecting the WTS structure.Experiment results show that this method can be used to predict the substorm with low temporal resolution and high spatial resolution auroal oval images.