Application of General Regression Neural Network to the Prediction of LOD Change

Cai

2016

ABSTRACT. It was shown in the previous study that the increase of pole coordinates prediction error for about 100 days in the future is mostly caused by irregular short period oscillations. In this paper, the ultra short-term prediction of pole coordinates is studied for 10 days in the future by means of combination of empirical mode decomposition (EMD) and neural networks (NN), denoted EMD-NN. In the algorithm, EMD is employed as a low pass filter for eliminating high frequency signals from observed pole coordinates data. Then the annual and Chandler wobbles are removed a priori from pole coordinates data with high frequency signals eliminated. Finally, the radial basis function (RBF) networks are used to model and predict the residuals. The prediction performance of the EMD-NN approach is compared with that of the NN-only solution and the prediction methods and techniques involved in the Earth orientation parameters prediction comparison campaign (EOP PCC). The results show that the prediction accuracy of the EMD-NN algorithm is better than that of the NN-only solution and is also comparable with that of the other existing prediction method and techniques.

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Ultra Short-term Prediction of Pole Coordinates via Combination of Empirical Mode Decomposition and Neural Networks

Cai

2016

“…Various prediction strategies and techniques were employed in the past for improvement of the UT1-UTC or LOD predictions, e.g., autocovariance (AC) techniques (Kosek et al, 1998), artificial neural networks (ANN) (Schuh et al, 2002;Liao et al, 2012;Zhang et al, 2012;Lei et al, 2015(a)), fuzzy inference systems (FIS) (Akyilmaz and Kutterer, 2004), Gaussian process regression (GPR) (Lei et al, 2015(b)), autoregressive (AR) and multivariate autoregressive (MAR) techniques (Niedzielski and Kosek, 2008). In October 2005 the Earth Orientation Parameters Prediction Comparison Campaign (EOP PCC) was started in an effort to evaluate different methods available for forecasting EOP data under the same rules and conditions.…”

Section: Introductionmentioning

confidence: 99%

Application of Grey Model GM(1, 1) to Ultra Short-Term Predictions of Universal Time

Guo

et al. 2016

ABSTRACT.A mathematical model known as one-order one-variable grey differential equation model GM(1, 1) has been herein employed successfully for the ultra short-term (<10days) predictions of universal time (UT1-UTC). The results of predictions are analyzed and compared with those obtained by other methods. It is shown that the accuracy of the predictions is comparable with that obtained by other prediction methods. The proposed method is able to yield an exact prediction even though only a few observations are provided. Hence it is very valuable in the case of a small size dataset since traditional methods, e.g., least-squares (LS) extrapolation, require longer data span to make a good forecast. In addition, these results can be obtained without making any assumption about an original dataset, and thus is of high reliability. Another advantage is that the developed method is easy to use. All these reveal a great potential of the GM(1, 1) model for UT1-UTC predictions.

“…Up to now, various prediction methods and techniques have been developed to improve the prediction accuracy of LOD time-series, e.g., least-squares (LS) extrapolation of the harmonic model [Niedzielski T., Kosek W. 2008], autocovariance (AC) prediction [Koesk W. et al 1998], autoregressive (AR) prediction [Niedzielski T., Kosek W. 2008], artificial neural networks (ANN) [Schuh H. et al 2002], [Zhang XH. et al 2012], fuzzy inference systems (FIS) [Akyilmaz O., Kutterer H. 2004], [Akyilmaz O., Kutterer H. 2005], fuzzy-wavelet [Akyilmaz O. et al 2011] and combined solutions [Xu XQ.…”

Section: Introductionmentioning

confidence: 99%

Extreme Learning Machine for the Predictions of Length of Day

Cai

2015

ABSTRACT. This work presents short-and medium-term predictions of length of day (LOD) up to 500 days by means of extreme learning machine (ELM). The EOP C04 time-series with daily values from the International Earth Rotation and Reference Systems Service (IERS) serve as the data basis. The influences of the solid Earth and ocean tides and seasonal atmospheric variations are removed from the C04 series. The residuals are used for training of the ELM. The results of the prediction are compared with those from other prediction methods. The accuracy of the prediction is equal to or even better than that by other approaches. The most striking advantages of employing ELM instead of other algorithms are its noticeably reduced complexity and high computational efficiency.