基于CEEMDAN-IPSO-KELM模型的BDS卫星钟差预报  

BDS Satellite Clock Bias Prediction Based on CEEMDAN-IPSO-KELM Model

作　　者：边奇海 张莎薇 雷荣智 刘子巍 刘敏 BIAN Qihai;ZHANG Shawei;LEI Rongzhi;LIU Ziwei;LIU Min(Zhejiang Provincial Land Survey and Planning Co.,Ltd.,Hangzhou 310030,China;Zhejiang University ofWater Resources and Hydropower,Hangzhou 310018,China;Zhejiang Academy of Surveying and Mapping,Hangzhou 310030,China)

机构地区：[1]浙江省国土勘测规划有限公司,浙江杭州310030 [2]浙江水利水电学院,浙江杭州310018 [3]浙江省测绘科学技术研究院,浙江杭州310030

出　　处：《地理空间信息》2025年第3期13-17,共5页Geospatial Information

基　　金：浙江省测绘科学技术研究院科研基金资助项目(DFP2021D0605)。

摘　　要：针对BDS卫星钟差数据呈非线性特征、难以准确预报的问题,结合自适应噪声完备集合经验模态分解(CEEMDAN)算法和核极限学习机(KELM)算法的优势,构建了组合钟差预报模型。首先充分利用CEEMDAN算法的信号分解能力自适应分解非平稳性钟差序列,并重构分解结果得到新的钟差序列;再利用改进粒子群优化(IPSO)算法优化KELM的核参数与正则化参数;最后重构不同钟差序列的预报结果,得到最终钟差预报结果。利用iGMAS提供的BDS钟差数据进行短期预报实验,结果表明该组合预报模型的单天和多天钟差预报精度均明显优于对比模型,丰富了现有BDS卫星钟差预报模型。Aiming at the nonlinear characteristics of BDS satellite clock bias data and the difficulty of accurate prediction,combining the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)algorithm and kernel extreme learning machine(KELM)algorithm,we constructed a new combined clock bias prediction model.Firstly,we used the signal decomposition ability of CEEMDAN algorithm to adaptively decompose the nonstationary clock bias sequence,and reconstructed the decomposition results to obtain new clock bias sequences.Then,we used the improved particle swarm optimization(IPSO)algorithm to optimize the kernel parameters and regularization parameters of KELM.Finally,we reconstructed the prediction results of different clock bias sequences to obtain the final clock bias prediction results.We used BDS clock bias data provided by the iGMAS to carry out short-term prediction experiments.The experimental results show that the prediction accuracy of model proposed in this paper is significantly better than that of the contrast models,whether it is single-day or multi-day clock bias prediction,which can enrich the existing BDS satellite clock bias prediction model.

关 键 词：BDS 钟差预报 CEEMDAN IPSO算法 KELM 

分 类 号：P228.4[天文地球—大地测量学与测量工程]