Several Improved Methods For Prediction Of Earth Rotation Parameters

The earth's rotation motion can be described by polar motion,length of day,precession and nutation and the parameters are called Earth Orientation Parameter(EOP).The EOPs contain a wealth of geodynamic information and play a significant role in several applications,including the determination of high-precision and highresolution satellite orbits,spacecraft tracking,astrogeodynamics,deep space exploration.Due to the complexity of the data processing involved,EOP products cannot be calculated in real-time.Thus,the EOPs are usually provided after a delay of hours to days and cannot satisfy the real-time demands of applications.Therefore,research on the prediction of EOPs has significant theoretical and practical value.In this paper,from the aspects of the determination of model order,data selection,sparse modeling,etc.,the prediction method of EOPs is improved.The main work and results are as follows:(1)In previous studies,there have been incorrectly used of the order determination criterion for the determining the order of autoregressive model(AR)model.In this paper,we analyze whether the order determination criterion is correctly used effects on the order of the AR model and the LS+AR(least squares extrapolation of harmonic and autoregressive)model predictive performance.The theory and experimental results show that the wrong use of the order determination criterion will reduce the final model order value,and greatly decrease for the mid-long term polar motion parameters prediction performance of the LS+AR model.(2)Aiming at the problem that redundant and stale data will cause the LS+AR model over learning,an experimental study is conducted to check the effect of different data volume on the final prediction performance of LS+AR model.The data selection range suitable for medium and long term polar motion parameters prediction is given.The experimental results showed that the method that the AR model parameters calculated by appropriate data volume can effectively improve the accuracy of longterm prediction of polar motion.(3)A weight least-squares extrapolation and Autoregressive model(WLS+AR)was proposed which uses the calculated precision of the corresponding parameters from EOP products as weighting factors to determine the weight matrices of the vectors of observation in the LS model.The experimental results show that the proposed method effectively improves the prediction accuracy of 1-270 day polar motion parameters and the 1-500 day ?LOD parameters.(4)In the calculating step of LS+AR model,the LS and AR model parameters are usually calculated in two steps and independently,and only the step-by-step optimal results can be achieved.A one-step method based on LS+AR model was proposed that the LS and AR models are simultaneously calculated to obtain the global optimal solution.The experimental results show that the method improves the medium and long term prediction accuracy of ?LOD parameter.(5)Traditional model selection methods can only determine the highest order of the AR model,but cannot further filter the items below the highest order.The sparse modeling method is introduced into the polar motion parameter prediction by using the LASSO and Elastic Net methods.The L1 norm and the L2 norm are added to the constraint,which plays the role of smooth and screening the model parameters and can enhance the stability of the model parameters and improve the accuracy of the forecast.The experimental results show that the proposed method with better stability and higher prediction performance,reduces the model order while maintaining effective information.(6)Conventional polar motion parameter prediction method models the Cartesian coordinate components of its.It is noted that the polar coordinate components of the polar motion parameter vector(ie,the polar angle and the polar diameter)show better regularity.In view of this,this paper proposes a novel idea of modeling and forecasting the polar motion parameter vector,and determines the optimal forecast combination model of polar angle and polar diameter.The experimental results show that compared with the current LS+AR model,one of the best polar motion parameter prediction models,the proposed forecasting scheme reaches a better prediction accuracy.