基于SSA去噪的级联LSTM网络地球极移短期预报方法

测绘学报 ›› 2026, Vol. 55 ›› Issue (1): 46-58.

• 大地测量学与导航 • 上一篇

基于SSA去噪的级联LSTM网络地球极移短期预报方法

张文渊¹^,²(), 彭劲松¹(), 韦纳都³, 高雨¹, 张书毕¹^,²

^1.中国矿业大学环境与测绘学院，江苏　徐州　221116
^2.中国矿业大学自然资源部国土环境与灾害监测重点实验室，江苏　徐州　221116
^3.科学技术部国际科技合作中心，北京　100862

收稿日期:2025-03-07 修回日期:2025-09-15 发布日期:2026-02-13
通讯作者: 彭劲松 E-mail:zhangwy@cumt.edu.cn;pjs@cumt.edu.cn
作者简介:张文渊（1996—），男，博士，副教授，研究方向为GNSS大气探测及气候变化应用、GNSS数据处理及ERP预报。E-mail：zhangwy@cumt.edu.cn
基金资助:
国家自然科学基金(42404016);江苏省自然科学基金(BK20241669);中央高校基本科研业务费专项(2024QN11077);中国矿业大学研究生创新计划项目(2025WLJCRCZL230)

A short-term prediction method for Earth's polar motion using cascaded LSTM networks based on SSA denoising

Wenyuan ZHANG¹^,²(), Jinsong PENG¹(), Nadu WEI³, Yu GAO¹, Shubi ZHANG¹^,²

^1.School of Environment and Spatial Informatics, China University of Mining and Technology, Xuzhou 221116, China
^2.MNR Key Laboratory of Land Environment and Disaster Monitoring, China University of Mining and Technology, Xuzhou 221116, China
^3.International Science and Technology Cooperation Center, Ministry of Science and Technology of the People's Republic of China, Beijing 100862, China

Received:2025-03-07 Revised:2025-09-15 Published:2026-02-13
Contact: Jinsong PENG E-mail:zhangwy@cumt.edu.cn;pjs@cumt.edu.cn
About author:ZHANG Wenyuan (1996—), male, PhD, associate professor, majors in GNSS atmospheric monitoring and climate change applications, GNSS data processing, and ERP forecasting. E-mail: zhangwy@cumt.edu.cn
Supported by:
The National Natural Science Foundation of China(42404016);Natural Science Foundation of Jiangsu Province(BK20241669);Fundamental Research Funds for the Central Universities(2024QN11077);Graduate Innovation Program of China University of Mining and Technology(2025WLJCRCZL230)

摘要/Abstract

摘要：

地球极移是深空探测和卫星精密定轨的关键参数，其高精度预报模型是空间大地测量领域的研究热点。针对长短期记忆（LSTM）神经网络在短期预测中由于训练场景与应用场景不一致而导致的预测误差累积以及忽略信号噪声影响的问题，本文提出了一种基于奇异谱分析（SSA）去噪的级联LSTM网络地球极移短期预报方法。该方法首先利用SSA算法剔除极移时序信号的高频噪声项，随后充分考虑未来不同预测天数的场景特征变化，通过级联架构实现前序子模型输出与后续子模型输入的误差抵偿传导，构建了多个子模型相互连接、逐级传递的级联式LSTM框架。利用1984—2024年的EOP 20 C04序列数据进行了试验验证，结果表明：对于1~10天的短期预报，本文方法在极移X和Y方向的预测结果的平均绝对误差（MAE）分别为1.70和0.93 mas，相较于递归LSTM模型的MAE分别降低了42.8%和48.1%，同时相较于SSA-递归LSTM模型的预报精度分别提升了11.1%和28.8%。此外，本文模型在未来6~10天的极移预报中具有显著优势，论证了本文方法可有效抑制预报误差积累，提高中后期预报精度，将模型预报结果应用于卫星轨道的天球坐标系与地球坐标系转换，显著提升了坐标转换精度。

关键词: 地球极移, 短期预报, SSA, 级联LSTM, 去噪优化

Abstract:

Earth polar motion (PM), a pivotal geodynamic parameter governing deep space exploration and satellite precise orbit determination, necessitates high-precision prediction models that persist as a research focus in space geodesy. To address the issues of accumulated prediction errors caused by inconsistencies between training and application scenarios, as well as the effect of signal noise in long short-term memory (LSTM) neural networks, we propose a short-term PM prediction method with a cascaded LSTM architecture based on singular spectrum analysis (SSA) denoising. The proposed method first employs SSA to eliminate high-frequency noise components from polar motion time series signals. Subsequently, it fully considers the evolving scenario characteristics across different future prediction horizons, and constructs an interconnected cascaded LSTM framework where multiple sub-models are sequentially connected for progressive information transfer. The experimental results based on the EOP 20 C04 dataset spanning 1984 to 2024 demonstrate significant improvements: For 1~10 days short-term predictions, the proposed method achieves mean absolute errors (MAE) of 1.70 mas and 0.93 mas in the X and Y polar motion components, respectively. Compared to recursive LSTM baselines, the proposed model achieves 42.8% and 48.0% improvements, respectively. Furthermore, it outperforms existing SSA-recursive LSTM hybrid benchmarks by 11.0%and 28.5%in MAE reductions. Significantly, the cascaded architecture demonstrates superior predictive capability in 6~10 days forecasts, validating its effectiveness in mitigating error propagation while enhancing mid-to-long-term forecast stability. The prediction results are applied to the transformation between celestial and Earth coordinate systems for satellite orbits, significantly improving the accuracy of coordinate conversion.

Key words: Earth's polar motion, short-term prediction, singular spectrum analysis, cascaded LSTM, denoising optimization

中图分类号:

P228

张文渊, 彭劲松, 韦纳都, 高雨, 张书毕. 基于SSA去噪的级联LSTM网络地球极移短期预报方法[J]. 测绘学报, 2026, 55(1): 46-58.

Wenyuan ZHANG, Jinsong PENG, Nadu WEI, Yu GAO, Shubi ZHANG. A short-term prediction method for Earth's polar motion using cascaded LSTM networks based on SSA denoising[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(1): 46-58.

图/表 11

图1

图2

图3

图4

图5

图6

图7

表1

表2

图8

表3

参考文献 32

[1]	加松, 徐天河, 杨红雷. LS+AR极移预报模型的两种修正算法[J]. 测绘学报, 2018, 47(S1): 71-77. DOI: . doi: 10.11947/j.AGCS.2018.20180296
	JIA Song, XU Tianhe, YANG Honglei. Two improved algorithms for LS+AR prediction model of the polar motion[J]. Acta Geodaetica et Cartographica Sinica, 2018, 47(S1): 71-77. DOI: . doi: 10.11947/j.AGCS.2018.20180296
[2]	吴飞. 几种改进的地球自转参数预报方法[J]. 测绘学报, 2023, 52(7): 1229. DOI: . doi: 10.11947/j.AGCS.2023.20210716
	WU Fei. Several improved methods for prediction of earth rotation parameters[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(7): 1229. DOI: . doi: 10.11947/j.AGCS.2023.20210716
[3]	魏娜, 周雨欣, 许雪晴, 等. 顾及有效角动量与IGS超快解数据的极移预报方法[J]. 地球物理学报, 2024, 67(4): 1356-1367.
	WEI Na, ZHOU Yuxin, XU Xueqing, et al. A hybrid polar motion prediction method combining effective angular momentum and IGS ultra rapid data[J]. Chinese Journal of Geophysics, 2024, 67(4): 1356-1367.
[4]	WAN Lihua, WEI Erhu, JIN Shuanggen. Earth rotation parameter estimation from GNSS and its impact on aircraft orbit determination[C]//Proceedings of 2014 China Satellite Navigation Conference. Berlin: Springer, 2014: 105-114.
[5]	WANG Qianxin, HU Chao, XU Tianhe, et al. Impacts of Earth rotation parameters on GNSS ultra-rapid orbit prediction: derivation and real-time correction[J]. Advances in Space Research, 2017, 60(12): 2855-2870.
[6]	KOSEK W, KALARUS M, NIEDZIELSKI T, et al. Forecasting of the Earth orientation parameters: comparison of different algorithms[C]//Proceedings of 2007 Journées Systèmes de Référence Spatio-temporels. Paris: Observatoire de Paris, 2007: 155-158.
[7]	KALARUS M, SCHUH H, KOSEK W, et al. Achievements of the Earth orientation parameters prediction comparison campaign[J]. Journal of Geodesy, 2010, 84(10): 587-596.
[8]	严凤, 姚宜斌. 地球自转参数短期预报方法及其实现[J]. 大地测量与地球动力学, 2012, 32(4): 71-75.
	YAN Feng, YAO Yibin. Short-term prediction methods and realization of earth rotation parameters[J]. Journal of Geodesy and Geodynamics, 2012, 32(4): 71-75.
[9]	XU X Q, ZHOU Y H, LIAO X H. Short-term earth orientation parameters predictions by combination of the least-squares, AR model and Kalman filter[J]. Journal of Geodynamics, 2012, 62: 83-86.
[10]	王志文, 王潜心, 何义磊, 等. 一种考虑PMX与PMY之间相关性的极移预报新方法[J]. 大地测量与地球动力学, 2017, 37(11): 1178-1182.
	WANG Zhiwen, WANG Qianxin, HE Yilei, et al. A new method to predict pole shift based on the correlation between PMX and PMY[J]. Journal of Geodesy and Geodynamics, 2017, 37(11): 1178-1182.
[11]	YE Fei, YUAN Yunbin. A modified stochastic model for LS+AR hybrid method and its application in polar motion short-term prediction[J]. Geodesy and Geodynamics, 2024, 15(1): 100-105.
[12]	赵丹宁, 高蕊, 雷雨. 利用小波分解改进极移预报模型[J]. 武汉大学学报(信息科学版), 2019, 44(12): 1797-1801.
	ZHAO Danning, GAO Rui, LEI Yu. Improvement of the polar motion prediction model using wavelet decomposition[J]. Geomatics and Information Science of Wuhan University, 2019, 44(12): 1797-1801.
[13]	王小辉, 王琪洁, 刘建. 经验模式分解在极移超短期预报中的应用[J]. 天文学报, 2012, 53(6): 519-526.
	WANG Xiaohui, WANG Qijie, LIU Jian. Application of empirical mode decomposition in the ultra short-term prediction of polar motion[J]. Acta Astronomica Sinica, 2012, 53(6): 519-526.
[14]	KONG Qiaoli, HAN Jingwei, JIN Xin, et al. Polar motion prediction using the combination of SSA and ARMA[J]. Geodesy and Geo-dynamics, 2023, 14(4): 368-376.
[15]	GUESSOUM S, BELDA S, MODIRI S, et al. Joint short-term prediction of polar motion and length of day with multi-task deep learning methods[J]. Earth, Planets and Space, 2025, 77(1): 25.
[16]	MODIRI S, BELDA S, HEINKELMANN R, et al. Polar motion prediction using the combination of SSA and Copula-based analysis[J]. Earth, Planets and Space, 2018, 70(1): 115.
[17]	赵丹宁, 雷雨, 乔海花. 利用奇异谱分析提高极移中长期预报精度[J]. 天文学报, 2024, 65(3): 30-41.
	ZHAO Danning, LEI Yu, QIAO Haihua. Improvement in the medium-and long-term prediction accuracy of polar motion using singular spectrum analysis[J]. Acta Astronomica Sinica, 2024, 65(3): 30-41.
[18]	赵丹宁, 雷雨. 基于极限学习机的极移中长期预报[J]. 天文研究与技术, 2020, 17(2): 178-184.
	ZHAO Danning, LEI Yu. Medium-and long-term prediction of polar motion using extreme learning machine[J]. Astronomical Research & Technology, 2020, 17(2): 178-184.
[19]	王宗建. 基于BP神经网络的地球自转参数短期预测[D]. 西安: 长安大学, 2015.
	WANG Zongjian. The short-term prediction of Earth rotation parameters based on BP neural network[D]. Xi'an: Changan University, 2015.
[20]	王帅民, 赵亿奇, 王振华, 等. 不同组合模型的地球自转参数预报对比[J]. 大地测量与地球动力学, 2024, 44(4): 377-381.
	WANG Shuaimin, ZHAO Yiqi, WANG Zhenhua, et al. Comparison of Earth rotation parameters prediction based on different combination models[J]. Journal of Geodesy and Geodynamics, 2024, 44(4): 377-381.
[21]	WANG Guocheng, LIU Lintao, TU Yi, et al. Application of the radial basis function neural network to the short term prediction of the Earth's polar motion[J]. Studia Geophysica et Geodaetica, 2018, 62(2): 243-254.
[22]	GOU Junyang, KIANI SHAHVANDI M, HOHENSINN R, et al. Ultra-short-term prediction of LOD using LSTM neural networks[J]. Journal of Geodesy, 2023, 97(5): 52.
[23]	GOU Junyang, RÖSCH C, SHEHAJ E, et al. Modeling the differences between ultra-rapid and final orbit products of GPS satellites using machine-learning approaches[J]. Remote Sensing, 2023, 15(23): 5585.
[24]	WANG Chenxiang, ZHANG Pengfei. Improving the accuracy of polar motion prediction using a hybrid least squares and long short-term memory model[J]. Earth, Planets and Space, 2023, 75(1): 153.
[25]	YU Kehao, YANG Kai, SHEN Tonghui, et al. Estimation of earth rotation parameters and prediction of polar motion using hybrid CNN-LSTM model[J]. Remote Sensing, 2023, 15(2): 427.
[26]	WANG Xuqiao, DU Lan, ZHANG Zhongkai, et al. Leveraging the empirical wavelet transform in combination with convolutional LSTM neural networks to enhance the accuracy of polar motion prediction[J]. Research in Astronomy and Astrophysics, 2024, 24(9): 095019.
[27]	WANG Leyang, QUE Haibo, WU Fei. The CNN-LSTM-attention model for short term prediction of the polar motion[J]. Measurement Science and Technology, 2025, 36(1): 016323.
[28]	AKYILMAZ O, KUTTERER H, SHUM C K, et al. Fuzzy-wavelet based prediction of Earth rotation parameters[J]. Applied Soft Computing, 2011, 11(1): 837-841.
[29]	雷雨. 地球自转参数高精度预报方法研究[D]. 西安: 中国科学院研究生院(国家授时中心), 2016.
	LEI Yu. Study on methods for high accuracy prediction of Earth rotation parameters[D]. Xi'an: National Time Service Center, Chinese Academy of Sciences, 2016.
[30]	魏晓蓓. 地球自转参数极移的预报方法研究[D]. 青岛: 山东科技大学, 2017.
	WEI Xiaobei. Researches on the prediction methods of polar motion of the Earth rotation parameters[D]. Qingdao: Shandong University of Science and Technology, 2017.
[31]	王辰晨, 徐工, 苏晓庆, 等. 融合FBPBPF和NTFT的极移弱周期分量提取及其激发原因分析[J]. 地球物理学进展, 2025, 40(3): 905-912.
	WANG Chenchen, XU Gong, SU Xiaoqing, et al. FBPBPF and NTFT are used to extract the weak periodic component of polar motion and analyze its excitation reasons[J]. Progress in Geophysics, 2025, 40(3): 905-912.
[32]	TAN Zhirong, YE Fei, HUA Liangchun. Polar motion ultra-short-term prediction of least-squares+multivariate autoregressive hybrid method by using the Kalman filter[J]. Sensors, 2024, 24(19): 6260.

预报天数	方案1/mas	方案2/mas	MAE降低率1/（%）	方案3/mas	方案4/mas	MAE降低率2/（%）	MAE降低率3/（%）	MAE降低率4/（%）
1	0.816	0.816	0	0.816	0.816	0	0	0
2	1.310	1.142	12.8	0.923	0.915	30.2	19.9	0.9
3	1.816	1.540	15.2	1.109	1.051	42.1	31.8	5.2
4	2.300	1.930	16.1	1.356	1.266	45.0	34.4	6.6
5	2.762	2.213	19.9	1.640	1.534	44.5	30.7	6.5
6	3.212	2.587	19.5	1.945	1.732	46.1	33.0	11.0
7	3.665	2.914	20.5	2.274	1.995	45.6	31.5	12.3
8	4.101	3.226	21.3	2.604	2.258	44.9	30.0	13.3
9	4.581	3.542	22.7	3.004	2.491	45.6	29.7	17.1
10	5.088	3.919	23.0	3.411	2.914	42.7	25.6	14.6
均值	2.965	2.383	19.6	1.908	1.697	42.8	28.8	11.1

预报天数	方案1/mas	方案2/mas	MAE降低率1/（%）	方案3/mas	方案4/mas	MAE降低率2/（%）	MAE降低率3/（%）	MAE降低率4/（%）
1	0.676	0.676	0	0.554	0.554	18.0	18.0	0
2	0.956	0.910	4.8	0.643	0.607	36.5	33.3	5.6
3	1.216	1.150	5.4	0.766	0.651	46.5	43.4	15.0
4	1.455	1.377	5.4	0.895	0.714	50.9	48.1	20.2
5	1.672	1.579	5.6	1.034	0.780	53.3	50.6	24.6
6	1.880	1.744	7.2	1.253	0.890	52.7	49.0	29.0
7	2.115	1.904	10.0	1.530	1.019	51.8	46.5	33.4
8	2.374	2.090	12.0	1.813	1.193	49.7	42.9	34.2
9	2.645	2.311	12.6	2.117	1.363	48.5	41.0	35.6
10	2.901	2.473	14.8	2.425	1.506	48.1	39.1	37.9
均值	1.789	1.621	9.4	1.303	0.928	48.1	42.8	28.8

DOY	公报A/mm	LSTM递归/mm	LSTM级联/mm	SSA-递归LSTM/mm	SSA-级联LSTM/mm
223	33.3	60.1	54.1	101.5	96.4
224	65.5	102.0	104.9	92.0	72.8
225	99.7	140.4	136.4	94.7	46.1
226	134.8	162.5	146.3	103.0	33.1
227	173.3	180.4	165.9	110.8	20.9
228	217.8	201.2	257.6	116.8	54.7
229	248.0	221.9	292.7	141.6	82.3
230	252.1	235.0	237.2	196.3	82.8
均值	153.1	162.9	174.4	119.6	61.1

基于SSA去噪的级联LSTM网络地球极移短期预报方法

A short-term prediction method for Earth's polar motion using cascaded LSTM networks based on SSA denoising

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 32

相关文章 1

编辑推荐

Metrics

本文评价