Research On Spacecraft Orbit Fitting And Prediction Algorithm Based On Limited Observation Data

High-precision and fast spacecraft orbit fitting and prediction can improve the in-orbit service capability of satellites,reduce the energy consumption of orbit maintenance,and extend the service life of satellites.The orbit fitting and prediction method based on traditional dynamics model needs to model the perturbation force of satellite accurately.Due to the complexity of the space mechanics environment where the satellite is located,it is difficult to construct all the perturbation models accurately,which leads to some errors in satellite orbit fitting and prediction based on traditional dynamics models.In order to solve this problem,this paper studies the orbit fitting and prediction algorithm based on limited data,using the data fitting method,and neural network based on limited observation data to establish fitting and prediction model of the spacecraft orbit around in the traditional method of disturbing force of precise modeling difficulty,effectively improves the track fitting and prediction accuracy and speed.Firstly,a high precision numerical prediction model and an analytical prediction model based on Brouwer’s mean elements theory are established based on the dynamics model of spacecraft orbit in this paper.Combined with the almanac data of PRN03 satellite,the simulation analysis of the high precision numerical prediction model and the analytical prediction model based on Brouwer’s mean elements theory is carried out.The effectiveness of subsequent algorithms is compared.Then,in views of the problem that the traditional analytic method has a little of calculation and poor fitting precision,while the numerical method has a high fitting precision and a lot of calculation,this paper puts forward a kind of limited observation data,numerical fitting combined with analytical prediction of orbit fitting algorithm,using Brouwer’s mean elements theory analytical computation of orbit,introduced based on Fourier series numerical fitting,The error between the analytical results and the measured data is eliminated,which can effectively improve the accuracy of the track fitting and keep the faster calculation speed of the analytical method.Finally,to solve the problem of poor accuracy of Fourier series fitting smooth model for orbit prediction with limited observation data,two kinds of orbit prediction models based on neural network are proposed,which are based on nonlinear autoregressive neural network(NARX)with external input and short and long time memory neural network(LSTM)respectively.Combined with Brouwer’s mean elements theory,orbit prediction is carried out.The simulation results show that the error correction rate of the two models is about 70%,the prediction accuracy is between the analytical prediction and the high precision numerical prediction,and the calculation time is much lower than the high precision numerical prediction method.