The Research of Low Earth Orbit Prediction of Satellite Based on Deep Neural Network

Hai-tao YANG, Jun-peng ZHU, Jian ZHANG


Satellite orbit prediction is a basic requirement in satellite applications. The current orbit prediction mainly depends on the dynamic model. Because of the limitations of the detection equipment and the satellite orbit data cannot be updated in time, which cause the dynamical model long-term orbit divergence to be serious. Using deep neural network as a method of orbit prediction which can predict the future data by training the satellite orbit data and grasp the implicit relationship between the data. The neural network model is optimized and the prediction data is compared with the actual data. The error of 20 days forecast is reduced to 2km, which improves the accuracy of neural network forecasting satellite orbit.


Satellite orbit prediction, Deep neural network


Full Text:



  • There are currently no refbacks.