基于数据驱动方法的在轨卫星智能温度预测

doi:10.12305/j.issn.1001-506X.2024.05.16

系统工程与电子技术 ›› 2024, Vol. 46 ›› Issue (5): 1619-1627.doi: 10.12305/j.issn.1001-506X.2024.05.16

• 系统工程 • 上一篇

基于数据驱动方法的在轨卫星智能温度预测

夏青¹, 邱实¹^,*, 刘新颖¹^,², 刘明¹, 郭金生¹, 林晓辉¹

1. 哈尔滨工业大学航天学院, 黑龙江哈尔滨 150001
2. 北京卫星环境工程研究所, 北京 100094

收稿日期:2023-01-28 出版日期:2024-04-30 发布日期:2024-04-30
通讯作者: 邱实
作者简介:夏青(1998—), 女, 硕士研究生, 主要研究方向为基于人工智能的大规模星座自主智能管控
邱实(1990—), 男, 教授, 博士, 主要研究方向为大规模星座数字孪生、智能飞行器
刘新颖(1987—), 女, 工程师, 硕士研究生, 主要研究方向为航天器空间环境模拟技术
刘明(1981—), 男, 教授, 博士, 主要研究方向为卫星姿轨控制与健康管理
郭金生(1987—), 男, 教授, 博士, 主要研究方向为卫星热控系统设计
林晓辉(1966—), 男, 教授, 博士, 主要研究方向为卫星总体设计与系统仿真

Data-driven-based approach for intelligent temperature forecasting of in-orbit satellites

Qing XIA¹, Shi QIU¹^,*, Xinying LIU¹^,², Ming LIU¹, Jinsheng GUO¹, Xiaohui LIN¹

1. School of Astronautics, Harbin Institute of Technology, Harbin 150001, China
2. Beijing Institute of Satellite Environmental Engineering, Beijing 100094, China

Received:2023-01-28 Online:2024-04-30 Published:2024-04-30
Contact: Shi QIU

摘要/Abstract

摘要：

针对传统卫星温度预测方法在预测精度和鲁棒性方面表现不佳，难以满足高维度耦合数据预测需求的问题，提出一种针对卫星温度遥测数据的多元时序数据预测模型——改进的时间序列处理模块(advanced time series processing module, ATSPM)-Net。首先，构建了包含一维卷积和门控循环单元(gated recurrent unit, GRU)的ATSPM，以对高度耦合的遥测数据中的时间依赖关系进行多尺度提取。接着，设计了多元时序数据预测模型ATSPM-Net。通过堆叠ATSPM，ATSPM-Net确保模型的灵活感受野，从而实现高准确率和鲁棒性的遥测数据预测。最后，在5个数据集上进行的数值实验结果表明，相较于其他类型的时序数据预测模型，ATSPM-Net在参数量较少的情况下能展现出更优异的温度预测性能。

关键词: 温度预测, 遥测数据, 时序数据预测

Abstract:

Aiming at the problem of poor prediction accuracy and robustness of traditional satellite temperature forecasting methods, which are difficult to meet the demand for high-dimensional coupled data forecasting, a multivariate time series data forecasting model for satellite temperature telemetry data—advanced time series processing module (ATSPM)-Net is proposed. Firstly, an ATSPM consisting of one-dimensional convolution and gated recurrent unit(GRU) is constructed to extract temporal dependencies from highly coupled telemetry data at multiple scales. Next, a multivariate temporal data forecasting model ATSPM-Net is designed. By stacking ATSPM, ATSPM-Net ensures the flexible receptive field of the model, thereby achieving high accuracy and robustness in telemetry data forecasting. Finally, numerical experiments conducted on five datasets showed that compared to other types of time series data forecasting models, ATSPM-Net can demonstrate better temperature forecasting performance with fewer parameters.

Key words: temperature forecasting, telemetry data, time series data forecasting

中图分类号:

V416.4

夏青, 邱实, 刘新颖, 刘明, 郭金生, 林晓辉. 基于数据驱动方法的在轨卫星智能温度预测[J]. 系统工程与电子技术, 2024, 46(5): 1619-1627.

Qing XIA, Shi QIU, Xinying LIU, Ming LIU, Jinsheng GUO, Xiaohui LIN. Data-driven-based approach for intelligent temperature forecasting of in-orbit satellites[J]. Systems Engineering and Electronics, 2024, 46(5): 1619-1627.

图/表 15

图1

表1

图2

表2

图3

图4

图5

图6

图7

图8

图9

图10

图11

表3

表4

参考文献 30

1	ABDELAL G F , ABUELFOUTOUH N , HAMDY A , et al. Thermal fatigue analysis of small-satellite structure[J]. International Journal of Mechanics and Materials in Design, 2006, 3 (2): 145- 159.
2	ABDELAL G F , ABUELFOUTOUH N , GAD A H . Finite element analysis for satellite structures: applications to their design, manufacture and testing[M]. Berlin: Springer Science & Business Media, 2012.
3	ATAR C , AKTAS M . Advances in thermal modeling and analysis of satellites[J]. Gazi University Journal of Science, 2022, 35 (1): 42- 58. doi: 10.35378/gujs.840191
4	CIHAN A , AKTAS M , SOZBIR N . Investigation of external heats for cubesats at various low earth orbits[J]. Pamukkale Universitesi Muhendislik Bilimleri Dergisi, 2023, 29 (4): 314- 321.
5	KOVACS R , JOZSA V . Thermal analysis of the SMOG-1 PocketQube satellite[J]. Applied Thermal Engineering, 2018, 139, 506- 513. doi: 10.1016/j.applthermaleng.2018.05.020
6	YANG L , LI Q , KONG L , et al. Quasi-all-passive thermal control system design and on-orbit validation of Luojia 1-01 satellite[J]. Sensors, 2019, 19 (4): 827. doi: 10.3390/s19040827
7	BONNICI M , MOLLICONE P , FENECH M , et al. Analytical and numerical models for thermal related design of a new pico-satellite[J]. Applied Thermal Engineering, 2019, 159, 113908. doi: 10.1016/j.applthermaleng.2019.113908
8	ZHANG H , ZHAO X M , MEI Q , et al. On-orbit thermal deformation prediction for a high-resolution satellite camera[J]. Applied Thermal Engineering, 2021, 195, 117152. doi: 10.1016/j.applthermaleng.2021.117152
9	徐高祥. 某卫星的热分析与温度控制研究[D]. 哈尔滨: 哈尔滨工业大学, 2021.
	XU G X. Research on thermal analysis and temperature control of a satellite[D]. Harbin: Harbin Institute of Technology, 2021.
10	HE Z , WEN T , LIU X , et al. Loss estimation and thermal analysis of a magnetic levitation reaction flywheel with PMB and AMB for satellite application[J]. Energies, 2022, 15 (4): 1584. doi: 10.3390/en15041584
11	BENGIO Y , GOODFELLOW I , COURVILLE A . Deep learning[M]. Cambridge: MIT Press, 2017.
12	OMEARA C, SCHLAG L, WICKLER M. Applications of deep learning neural networks to satellite telemetry monitoring[C]// Proc. of the Spaceops Conference, 2018.
13	JIANG B , GONG H , ZHU M , et al. Attention-LSTM architecture combined with Bayesian hyperparameter optimization for indoor temperature prediction[J]. Building and Environment, 2022, 224 (10): 109536.
14	RUßWURM M, ALI M, ZHU X X, et al. Model and data uncertainty for satellite time series forecasting with deep recurrent models[C]//Proc. of the IEEE International Geoscience and Remote Sensing Symposium, 2020: 7025-7028.
15	YUAN Q Q , SHEN H F , LI T W , et al. Deep learning in environmental remote sensing: achievements and challenges[J]. Remote Sensing of Environment, 2020, 241, 111716. doi: 10.1016/j.rse.2020.111716
16	HEWAMALAGE H , BERGMEIR C , BANDARA K . Recurrent neural networks for time series forecasting: current status and future directions[J]. International Journal of Forecasting, 2021, 37 (1): 388- 427. doi: 10.1016/j.ijforecast.2020.06.008
17	AL-NAJAR M , THOUMYRE G , BERGSMA E W J , et al. Satellite derived bathymetry using deep learning[J]. Machine Learning, 2023, 112, 1107- 1130. doi: 10.1007/s10994-021-05977-w
18	SHIH S Y , SUN F K , LEE H . Temporal pattern attention for multivariate time series forecasting[J]. Machine Learning, 2019, 108 (8): 1421- 1441.
19	CAO D F , WANG Y J , DUAN J Y , et al. Spectral temporal graph neural network for multivariate time-series forecasting[J]. Advances in Neural Information Processing Systems, 2020, 33, 17766- 17778.
20	MOSKOLAI W R , ABDOU W , DIPANDA A . Application of deep learning architectures for satellite image time series prediction: a review[J]. Remote Sensing, 2021, 13 (23): 4822.
21	QIN J , JIANG H , LU N , et al. Enhancing solar PV output forecast by integrating ground and satellite observations with deep learning[J]. Renewable and Sustainable Energy Reviews, 2022, 167, 112680.
22	石梦鑫. 基于注意力机制神经网络的遥测数据预测方法[D]. 北京: 中国科学院大学(中国科学院国家空间科学中心), 2020.
	SHI M X. Telemetry data prediction method based on attention mechanism neural network[D]. Beijing: University of Chinese Academy of Sciences(National Space Science Center, Chinese Academy of Sciences), 2020.
23	XIAO C J , CHEN N C , HU C L , et al. A spatiotemporal deep learning model for sea surface temperature field prediction using time-series satellite data[J]. Environmental Modelling & Software, 2019, 120, 104502.
24	LI C, ZHANG Y N, ZHAO G H. Deep learning with long short-term memory networks for air temperature predictions[C]// Proc. of the International Conference on Artificial Intelligence and Advanced Manufacturing, 2019: 243-249.
25	张军, 吴朋莉, 石陆魁, 等. 联合MOD11A1和地面气象站点数据的多站点温度预测深度学习模型[J]. 计算机应用, 2023, 43 (1): 321- 328.
	ZHANG J , WU P L , SHI L K , et al. Deep learning model for multi-station temperature prediction combined with MOD11A1 and meteorological station data[J]. Journal of Computer Applications, 2023, 43 (1): 321- 328.
26	YU F, KOLTUN V. Multi-scale context aggregation by dilated convolutions[EB/OL]. [2023-01-05]. https://arxiv.org/pdf/1511.07122.pdf.
27	IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]// Proc. of the International Conference on Machine Learning, 2015: 448-456.
28	BA J L, KIROS J R, HINTON G E. Layer normalization[EB/OL]. [2023-01-25]. https://arxiv.org/pdf/1607.06450.pdf.
29	HENDRYCKS D, GIMPEL K. Gaussian error linear units (GELUs)[EB/OL]. [2023-01-10]. https://arxiv.org/abs/1606.08415v4.
30	GAO X Y, WANG Y, GAO Y, et al. Short-term load forecasting model of gru network based on deep learning framework[C]// Proc. of the IEEE 2nd Conference on Energy Internet and Energy System Integration, 2018.

数据集	L_in	L_out	T	训练集数量	测试集数量
Ⅰ	1	1	-1	39 999	10 000
Ⅱ	1	1	0	39 998	10 000
Ⅲ	10	1	0	39 414	9 854
Ⅳ	10	1	100	39 334	9 834
Ⅴ	10	10	100	38 753	9 689

参数	取值
批样本大小	32
训练轮次	100
初始学习率	0.000 1
优化器	Adam优化器

对比模型	参数	取值
TCN	Conv1D层(c)	3
	膨胀系数(d)	1, 2, 4
	激活函数(a)	RELU
	全连接层(f)	1
	连接顺序	c→c→c→f
2D-CNN	Conv2D层(c)	3
	BN层(b)	3
	激活函数(a)	RELU
	全连接层(f)	f
	连接顺序	c→b→a→c→b→a→c→b→a→f

数据集	模型	MSE	MAE	MSLE	SMAPE	R²	参数量
Ⅰ	TCN	0.263 8	0.441 6	0.062 1	1.4736	0.079 1	32 952
	2D-CNN	0.021 4	0.106 6	0.005 6	0.422 2	0.924 6	20 388
	ATSPM-Net	0.003 1	0.027 0	0.000 9	0.140 6	0.989 1	16 420
Ⅱ	TCN	0.264 2	0.441 8	0.062 4	1.473 6	0.079 5	32 952
	2D-CNN	0.023 0	0.108 0	0.005 5	0.404 4	0.919 7	20 388
	ATSPM-Net	0.003 4	0.026 2	0.000 8	0.136 9	0.988 1	16 420
Ⅲ	TCN	0.006 5	0.046 5	0.001 7	0.213 8	0.977 4	21 972
	2D-CNN	0.017 2	0.092 8	0.004 8	0.395 4	0.940 0	44 616
	ATSPM-Net	0.002 5	0.024 1	0.000 6	0.126 7	0.991 3	20 388
Ⅳ	TCN	0.020 5	0.090 7	0.005 1	0.357 0	0.928 4	21 972
	2D-CNN	0.036 3	0.136 4	0.008 9	0.476 8	0.873 7	44 616
	ATSPM-Net	0.005 6	0.037 3	0.001 3	0.180 1	0.980 5	20 388
Ⅴ	TCN	0.020 3	0.092 1	0.005 0	0.367 5	0.929 1	177 816
	2D-CNN	0.042 1	0.161 1	0.009 7	0.544 9	0.852 7	131 080
	ATSPM-Net	0.003 3	0.031 0	0.000 9	0.157 5	0.988 6	161 580

基于数据驱动方法的在轨卫星智能温度预测

Data-driven-based approach for intelligent temperature forecasting of in-orbit satellites

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 30

相关文章 0

编辑推荐

Metrics

本文评价