测绘学报 ›› 2020, Vol. 49 ›› Issue (5): 580-588.doi: 10.11947/j.AGCS.2020.20190156

• 大地测量学与导航 • 上一篇    下一篇

卫星钟差预报的T-S模糊神经网络法

王旭1,2, 柴洪洲1, 王昶3   

  1. 1. 信息工程大学地理空间信息学院, 河南 郑州 450001;
    2. 辽宁生态工程职业学院测绘工程学院, 辽宁 沈阳 110101;
    3. 辽宁科技大学土木工程学院, 辽宁 鞍山 114051
  • 收稿日期:2019-04-27 修回日期:2019-12-27 发布日期:2020-05-23
  • 作者简介:王旭(1983-),男,博士生,讲师,研究方向为测量数据处理理论与方法。E-mail:wangxu19830411@126.com
  • 基金资助:
    国家自然科学基金(41574010;41604013;41904039)

T-S fuzzy neural network to predict satellite clock bias

WANG Xu1,2, CHAI Hongzhou1, WANG Chang3   

  1. 1. Institute of Surveying and Mapping, Information Engineering University, Zhengzhou 450001, China;
    2. Institute of Surveying and Mapping Engineering, Liaoning Vocational College of Ecological Engineering, Shenyang 110101, China;
    3. School of Civil Engineering, University of Science and Technology Liaoning, Anshan 114051, China
  • Received:2019-04-27 Revised:2019-12-27 Published:2020-05-23
  • Supported by:
    The National Natural Science Foundation of China (Nos. 41574010;41604013;41904039)

摘要: 结合钟差数据的特点,提出了一种基于变化率的T-S模糊神经网络(TSFNN)钟差预报模型。首先计算相邻历元间钟差的变化率值并对其进行建模;然后利用TSFNN模型预报钟差变化率值,再将预报的变化率值还原,得到钟差预报值;最后,通过算例将本文所建模型与IGU-P产品、二次多项式模型(QP)及灰色模型(GM(1,1))进行试验对比。结果表明:在使用变化率方法后,TSFNN模型预报的精度和稳定性分别提高了69.8%和76.3%,而且与IGU-P钟差产品相比,预报的精度高出约10倍,同时模型预报的效果优于两种常用模型。因此,该模型可以实现卫星钟差较高精度的预报。

关键词: 卫星钟差, T-S模糊神经网络, 变化率, 预报

Abstract: To find a high accuracy method for SCB(satellite clock bias) prediction based on characteristics of SCB data, this study proposes an T-S fuzzy neural network model based on the change rate method to predict SCB. the change rate of two SCB values of adjacent epoch is first calculated to obtain the corresponding change rate sequences. Then, modeling is performed based on the change rate sequence to predict the change rate value using the T-S fuzzy neural network model (TSFNN). Finally, the predicted sequences are recovered to the corresponding predicted SCB. The new model is compared with IGU-P products, quadratic polynomial (QP) model, and gray model (GM (1,1)) through the predicted sequences. The results show that the prediction precisions and stability of TSFNN model have been improved about 69.8% and 76.3% respectively after using the change rate method, and the accuracy is about 10 times higher than IGU-P products. the prediction effect of the proposed model is better than two common models. Therefore, the proposed model can achieve high accuracy prediction of SCB.

Key words: satellite clock bias, T-S fuzzy neural network, change rate, prediction

中图分类号: