顾及噪声项的连续运行参考站高程建模方法

A modeling method for the height of reference station in continuous operation considering the noise term

针对未考虑噪声情况下利用非线性周期拟合模型得到的高程预测值与实际值存在偏差较大的问题,该文提出一种顾及噪声项的CORS站高程建模预测方法。对CORS站高程原始数据进行CEEMD分解,利用自相关函数绝对值均值变点理论判定噪声与信号分界点;对噪声序列建立ARMA模型、信号序列建立非线性周期拟合模型;将非线性周期拟合模型和ARMA模型组合进行CORS站高程数据的预测。实验结果表明:顾及噪声项的非线性速度场预测方法高程预测精度优于未考虑噪声项的非线性预测方法。该方法为CORS站高程数据的建模预测提供了一种新的途径。

Aiming at the problem of big deviation between the predicted height value and the actual height value obtained by nonlinear periodic fitting model without considering noise,this paper presents a method of height modeling and prediction of CORS stations with considering noise term.This method decomposes the original height data of CORS stations by CEEMD,and uses the theory of absolute value mean change of autocorrelation function to determine the demarcation point between noise and signal.The ARMA model is established for noise sequences and the nonlinear periodic fitting model is established for signal sequences.The nonlinear periodic fitting model and ARMA model are combined to predict the height data of CORS stations.The experimental results show that the height prediction accuracy of the nonlinear velocity field prediction method considering the noise term is better than that without considering the noise term.The proposed method provides a new way to model and predict the height data of CORS stations.