Researches on the Theories and Algorithms of Earth Orientation Paramters Prediction
|School||PLA Information Engineering University|
|Course||Geodesy and Survey Engineering|
|Keywords||Earth Orientation Parameters Earth's rotation Pole shift Length of day Robust Estimation Time Series Neural Networks Gray model|
Affected by many of the excitation source , the instantaneous change of the orientation parameters of the earth is extremely complicated , detailed study of its variation for GEODYNAMICS has important significance . Meanwhile , satellite navigation , deep space exploration and other short-term changes in Earth Orientation Parameters practical applications require high precision value , and the value of the change is difficult to obtain real-time , hence the need for short - term forecasting of the Earth Orientation Parameters . The use of existing Earth Orientation Parameters time-series data in-depth study of the applicable short-term forecasting of the Earth Orientation Parameters , the main work is as follows: the brief polar motion and long cycle entry , excitation source , and the Earth Orientation Determination of parameters. Based on least squares fitting and forecasting analysis confirmed the robust estimation fit and prediction of Earth orientation parameters is more appropriate . 2, further study of the of ARMA model family in the Earth Orientation Parameters analysis found that the least squares prediction residuals , day length of the relevant characteristics of the AR model to describe the shift characteristics are closer to those ARIMA model . Thus, ARIMA models to forecast the pole shift parameters , and put forward a new intercept correction method to compensate for its prediction error . Explore the feasibility of direct neural network for short-term forecasts ; neural network function model , time series stochastic model , a combination prediction model , and compare with the neural network prediction model based on least squares do . Proposed gray model , time series , neural networks combined to predict the Earth Orientation Parameters . The results show that the gray model is simple and convenient , and to be able to take full advantage of the latest data modeling , and thus better dynamic real-time ; better than the commonly used prediction model based on a combination of gray model prediction accuracy combination forecasting model based on least squares .