基于模型数据双驱动的短波MUF短期预测网络

doi:10.11959/j.issn.1000-436x.2023234

通信学报 ›› 2023, Vol. 44 ›› Issue (12): 99-111.doi: 10.11959/j.issn.1000-436x.2023234

• 学术论文 • 上一篇

基于模型数据双驱动的短波MUF短期预测网络

李俊兵¹, 曾囿钧², 曾孝平¹, 李国军³, 白晨曦⁴

¹ 重庆大学微电子与通信工程学院，重庆 400040
² 中国工程物理研究院电子工程研究所，四川绵阳 621900
³ 重庆邮电大学超视距可信信息传输研究所，重庆 400065
⁴ 陆军工程大学通信士官学校，重庆 400036

修回日期:2023-11-13 出版日期:2023-12-01 发布日期:2023-12-01
作者简介:李俊兵（1994- ），男，四川资阳人，重庆大学博士生，主要研究方向为短波组网、宽带短波传输技术
曾囿钧（1993- ），男，四川资阳人，中国工程物理研究院电子工程研究所博士生，主要研究方向为智能辅助短波无线通信、强化学习
曾孝平（1965- ），男，四川广安人，博士，重庆大学教授、博士生导师，主要研究方向为下一代移动通信、无线通信、空间信息网
李国军（1978- ），男，四川资阳人，博士，重庆邮电大学教授、博士生导师，主要研究方向为复杂恶劣环境超视距无线通信与网络
白晨曦（1987- ），男，内蒙古乌兰察布人，陆军工程大学通信士官学校讲师，主要研究方向为信息通信和物联网技术
基金资助:
国家自然科学基金资助项目(U21A20448);国家自然科学基金资助项目(U20A20157);国家自然科学基金资助项目(U22A2006);重庆市基础研究与前沿探索基金资助项目(cstc2021ycjh-bgzxm0072)

Short-term prediction network for short-wave MUF based on model-data dual-driven

Junbing LI¹, Youjun ZENG², Xiaoping ZENG¹, Guojun LI³, Chenxi BAI⁴

¹ School of Microelectronics and Communication Engineering, Chongqing University, Chongqing 400040, China
² Institute of Electronic Engineering, China Academy of Engineering Physics, Mianyang 621900, China
³ Lab of BLOS Reliable Information Transmission of Chongqing University of Posts and Telecommunications, Chongqing 400065, China
⁴ Communication NCO Academy, Army Engineering University of PLA, Chongqing 400036, China

Revised:2023-11-13 Online:2023-12-01 Published:2023-12-01
Supported by:
The National Natural Science Foundation of China(U21A20448);The National Natural Science Foundation of China(U20A20157);The National Natural Science Foundation of China(U22A2006);The Chongqing Basic Research and Frontier Exploration Project(cstc2021ycjh-bgzxm0072)

摘要/Abstract

摘要：

针对短波最大可用频率（MUF）经典模型方法预测精度低及机器学习方法训练集数据获取难度大的问题，提出一种模型数据双驱动的双向门控递归单元（BiGRU）网络用于MUF短期预测。模型驱动方面，利用经典MUF预测模型生成的大规模数据集作为模型驱动训练集，经过2D CNN和BiGRU网络联合学习后，获得一个初步网络。数据驱动方面，使用小规模的实测数据集对初步网络进行二次训练，得到最终网络 CNN-BiGRU-NN。仿真结果表明，所提网络与GRU网络、LSTM网络以及VOACAP模型相比，在日期尺度和时刻尺度上的平均均方根误差（RMSE）均有降低。

关键词: 短波通信, 最大可用频率, 短期预测, 模型数据双驱动, CNN-BiGRU-NN

Abstract:

Predicting the maximum available frequency of short-wave communication presents the challenges of low prediction accuracy of classical prediction model methods and difficulty in obtaining training set data for machine learning prediction methods.To address this issue, a model-data dual-driven bidirectional gated recurrent unit (BiGRU) network for short-term prediction of MUF was proposed.On the model-driven, a large-scale dataset generated by the classical MUF prediction model was used as the model-driven training set, and a preliminary network was obtained after joint learning of the 2D CNN and the BiGRU network.On the data-driven, the preliminary network was trained twice using a small-scale measured dataset to obtain the final network CNN-BiGRU-NN.The simulation results show that the proposed network has reduced average root mean squared error (RMSE) at both daily and momentary scales compared with the GRU network, LSTM network and VOACAP model.

Key words: short-wave communication, maximum usable frequency, short-term prediction, model-data dual-driven, CNN-BiGRU-NN

中图分类号:

TN92

李俊兵, 曾囿钧, 曾孝平, 李国军, 白晨曦. 基于模型数据双驱动的短波MUF短期预测网络[J]. 通信学报, 2023, 44(12): 99-111.

Junbing LI, Youjun ZENG, Xiaoping ZENG, Guojun LI, Chenxi BAI. Short-term prediction network for short-wave MUF based on model-data dual-driven[J]. Journal on Communications, 2023, 44(12): 99-111.

图/表 12

图1

图2

图3

图4

图5

图6

表1

各种方法在不同日期同一时刻的Δfhour"

时刻	$χ_{1}$ 和 $χ_{2}$ 作为训练集			$χ_{2}$ 作为训练集			VOACAP/MHz
时刻	CNN-BiGRU-NN/MHz	CNN-GRU/MHz	LSTM/MHz	CNN-BiGRU-NN/MHz	CNN-GRU/MHz	LSTM/MHz	VOACAP/MHz
0:30	1.487	1.082	1.213	1.621	1.539	1.517	6.630
1:30	1.344	1.029	1.222	1.470	1.447	1.380	8.032
2:30	2.159	1.633	2.114	2.629	2.644	2.315	6.983
3:30	4.687	3.902	4.776	5.692	5.752	5.233	4.564
4:30	3.932	3.475	3.924	4.563	4.819	4.380	4.382
5:30	2.774	2.578	2.681	3.053	2.939	2.773	5.011
6:30	3.270	3.057	3.209	3.952	3.635	3.368	4.409
7:30	3.205	2.964	3.046	4.497	3.772	3.602	4.427
8:30	2.240	2.131	2.117	2.724	2.490	2.259	6.553
9:30	2.919	3.229	3.329	2.658	2.793	2.825	9.487
10:30	3.497	3.963	4.519	3.071	3.328	3.476	12.256
11:30	3.886	4.706	5.191	3.322	3.523	3.907	15.284
12:30	3.164	4.613	4.333	2.758	3.009	3.270	16.475
13:30	2.377	4.301	3.824	2.353	2.727	2.949	16.534
14:30	0.939	2.523	1.150	0.929	1.015	1.057	16.756
15:30	0.985	1.424	0.918	0.645	0.552	0.486	16.940
16:30	0.952	1.343	1.419	0.746	0.684	0.685	16.599
17:30	0.831	0.931	1.192	0.772	0.730	0.703	15.306
18:30	0.806	0.766	1.191	0.919	0.880	0.834	13.300
19:30	0.989	0.714	1.140	1.040	1.013	0.958	9.967
20:30	0.646	0.946	1.140	0.561	0.469	0.571	7.765
21:30	0.980	1.321	1.149	0.711	0.784	0.907	6.757
22:30	1.332	1.161	1.556	1.545	1.481	1.416	4.537
23:30	1.502	1.464	0.952	1.605	1.606	1.433	4.385

表1

表2

各种方法在同一日期不同时刻的Δfday"

日期	$χ_{1}$ 和 $χ_{2}$ 作为训练集			$χ_{2}$ 作为训练集			VOACAP/MHz
日期	CNN-BiGRU-NN/MHz	CNN-GRU/MHz	LSTM/MHz	CNN-BiGRU-NN/MHz	CNN-GRU/MHz	LSTM/MHz	VOACAP/MHz
10月9日	1.910	1.825	2.282	2.277	2.188	2.033	8.338
10月10日	2.251	2.474	2.640	2.501	2.480	2.401	9.188
10月11日	1.760	1.934	1.960	1.831	1.845	1.807	9.979
10月12日	3.044	3.155	3.263	3.224	3.211	3.128	8.815
10月13日	2.115	2.301	2.288	2.236	2.228	2.164	9.102
10月14日	1.722	2.334	1.974	1.837	1.857	1.856	10.506
10月15日	2.112	2.524	2.359	2.278	2.267	2.228	9.819
10月16日	2.434	2.539	2.437	2.693	2.617	2.514	8.832
10月17日	1.850	1.920	2.281	1.895	1.893	1.818	10.140
10月18日	1.601	1.744	1.681	1.695	1.678	1.595	9.972
10月19日	1.945	1.799	1.830	2.059	1.981	1.870	9.526
10月20日	1.636	2.161	2.595	1.649	1.676	1.680	10.580
10月21日	1.552	2.259	1.823	1.438	1.503	1.601	11.385
10月22日	1.847	2.108	2.058	2.008	2.000	1.947	10.545
10月23日	1.662	2.099	2.080	1.457	1.587	1.654	11.217
10月24日	2.073	2.267	2.294	2.038	2.086	2.081	11.022
10月25日	1.821	1.928	2.073	2.318	2.217	2.071	8.825
10月26日	2.778	2.542	2.981	3.352	3.179	2.980	7.647
10月27日	2.426	2.447	2.550	2.546	2.528	2.466	9.087
10月28日	2.736	2.785	2.976	2.727	2.722	2.659	9.827
10月29日	2.736	2.845	3.401	3.064	3.056	2.984	8.695
10月30日	2.649	2.659	2.704	2.225	2.362	2.406	10.845

表2

表3

重庆—海口链路各种方法的平均RMSE"

参数	$χ_{1}$ 和 $χ_{2}$ 作为训练集			$χ_{2}$ 作为训练集			VOACAP
参数	CNN-BiGRU-NN	CNN-GRU	LSTM	CNN-BiGRU-NN	CNN-GRU	LSTM	VOACAP
RMSE_day/MHz	2.287 9	2.924 9	3.238 4	3.022 6	2.978 4	3.055 6	10.964 4
η_day	—	3.32%	12.68%	6.44%	5.05%	7.45%	74.21%
RMSE_hour/MHz	2.545 2	2.583 4	2.911 6	2.668 9	2.627 6	3.332 5	9.962 1
η_hour	—	1.48%	12.58%	4.63%	3.14%	23.62%	74.45%

表3

表4

重庆—昆明链路各种方法的平均RMSE"

参数	$χ_{1}$ 和 $χ_{2}$ 作为训练集			$χ_{2}$ 作为训练集			VOACAP
参数	CNN-BiGRU-NN	CNN-GRU	LSTM	CNN-BiGRU-NN	CNN-GRU	LSTM	VOACAP
RMSE_day/MHz	1.182 4	1.194 3	1.359 6	1.215 7	1.481 8	2.105 7	7.548 2
η_day	—	0.99%	13.03%	2.74%	20.21%	43.85%	84.34%
RMSE_hour/MHz	1.060 1	1.068 1	1.176 6	1.092 6	1.298 0	1.916 7	7.355 8
η_hour	—	0.75%	9.90%	2.97%	18.33%	44.69%	85.89%

表4

图7

图8

参考文献 47

[1]	WANG J L , DING G R , WANG H C . HF communications:past,present,and future[J]. China Communications, 2018,15(9): 1-9.
[2]	王金龙 . 短波数字通信研究与实践[M]. 北京: 科学出版社, 2013.
	WANG J L . Short-wave digital communication research and practice[M]. Beijing: Science Press, 2013.
[3]	张家帅, 骆志刚, 张翔 ,等. 超短波通信系统的物理层仿真精确建模方法研究[J]. 电信科学, 2012,28(1): 86-90.
	ZHANG J S , LUO Z G , ZHA NG X ,et al. Research on exact simulation modeling approach for physical layer of ultra-short communication system[J]. Telecommunications Science, 2012,28(1): 86-90.
[4]	Recommendation ITU-RP. 533-14.Method for the prediction of the performance of HF circuits[S]. 2019.
[5]	Recommendation ITU-RP. 1240-7.ITU-R methods of basic MUF,operational MUF and ray-path prediction[S]. 2019.
[6]	孙宪儒 . 亚大地区 F_2 电离层预测方法[J]. 通信学报, 1987,8(6): 153-156.
	SUN X R . A method of predicting the ionospheric F2 layer in the Asia Oceania region[J]. Journal on Communications, 1987,8(6): 153-156.
[7]	李忠勤, 孙宪儒 . 亚大预测与 CCIR 推荐方法(340 报告)的比较[J]. 电波科学学报, 1992(2): 17-29.
	LI Z Q , SUN X R . Comparison between sub-large prediction and CCIR recommendation method (report 340)[J]. China Industrial Economics, 1992(2): 17-29.
[8]	曹红艳, 孙宪儒 . 新版亚大地区 F2 电离层频率预测方法[J]. 空间科学学报, 2009,29(5): 502-507.
	CAO H Y , SUN X R . A new method of predicting the ionospheric F₂ layer in the Asia Oceania region[J]. Chinese Journal of Space Science, 2009,29(5): 502-507.
[9]	鲁转侠, 曹红艳, 冯静 . “新版亚大地区 F2 电离层预测” 方法数据验证[J]. 中国电子科学研究院学报, 2011,6(1): 59-63.
	LU Z X , CAO H Y , FENG J . The new method data verification of predicting the ionospheric F₂layer in the Asia Oceania region[J]. Journal of China Acadewy of Electronics and Information Technology, 2011,6(1): 59-63.
[10]	FANG D , SOICHER H . The IONCAP program enhancement[C]// Proceedings of MILCOM Conference Record. Piscataway:IEEE Press, 2002: 288-290.
[11]	PERKI?M?KI J . VOACAP quick guide[R]. 2020.
[12]	TANYER S G , EROL C B . Broadcast analysis and prediction in the HF band[J]. IEEE Transactions on Broadcasting, 1998,44(2): 226-232.
[13]	高清 . 短波通信中的准实时频率预报技术[J]. 通信学报, 1987,8(3): 79-82.
	GAO Q . A quasi real-time prediction technique for HF communications[J]. Journal on Communications, 1987,8(3): 79-82.
[14]	SECAN J A , WILKINSON P J . Statistical studies of an effective sunspot number[J]. Radio Science, 1997,32(4): 1717-1724.
[15]	MARIN D , MIRO G , MIKHAILOV A V . A method for foF2 short-term prediction[J]. Physics and Chemistry of the Earth,Part C:Solar,Terrestrial ＆ Planetary Science, 2000,25(4): 327-332.
[16]	李国军, 郑广发, 叶昌荣 ,等. 基于稳健卡尔曼滤波的倾斜探测电离层MUF短期预报方法[J]. 通信学报, 2021,42(1): 79-86.
	LI G J , ZHENG G F , YE C R ,et al. Short-term prediction method of oblique sounding ionosphere MUF based on robust Kalman filter[J]. Journal on Communications, 2021,42(1): 79-86.
[17]	SEZEN U , SAHIN O , ARIKAN F ,et al. Estimation of hmF2 and foF2 communication parameters of ionosphere F2-layer using GPS data and IRI-plas model[J]. IEEE Transactions on Antennas and Propagation, 2013,61(10): 5264-5273.
[18]	ZENG W , ZHANG X J . Predictions of HF communication MUF in the region of the South China Sea[J]. IEEE Antennas and Propagation Magazine, 1999,41(4): 35-38.
[19]	任淑婷, 郭黎利 . 基于模糊小波神经网络的短波频率预测[J]. 通信技术, 2011,44(4): 37-39.
	REN S T , GUO L L . A prediction method for HF radio communication frequency based on FWNN[J]. Communications Technology, 2011,44(4): 37-39.
[20]	YU L , CHEN J , ZHANG Y M ,et al. Deep spectrum prediction in high frequency communication based on temporal-spectral residual network[J]. China Communications, 2018,15(9): 25-34.
[21]	WANG J , YANG C , AN W X . Regional refined long-term predictions method of usable frequency for HF communication based on machine learning over Asia[J]. IEEE Transactions on Antennas and Propagation, 2022,70(6): 4040-4055.
[22]	LI C , LIU Y J , WANG Y Q ,et al. Short-term prediction method of HF frequency based on deep learning network[C]// Proceedings of International Conference on Artificial Intelligence and Advanced Manufacturing (AIAM). Piscataway:IEEE Press, 2020: 360-363.
[23]	兰朝凤, 王顺博, 郭小霞 ,等. 基于DCNN和BiLSTM的单通道视听融合语音分离方法研究[J]. 电子学报, 2023,51(4): 914-921.
	LAN C F , WANG S B , GUO X X ,et al. A single channel audio-visual fusion speech separation method based on DCNN and BiLSTM[J]. Acta Electronica Sinica, 2023,51(4): 914-921.
[24]	RAO T V , SRIDHAR M , RATNAM D V ,et al. A bidirectional long short-term memory-based ionospheric foF2 and hmF2 models for a single station in the low latitude region[J]. IEEE Geoscience and Remote Sensing Letters, 2022,19: 1-5.
[25]	夏吉业, 张海勇, 徐池 ,等. 基于CNN-BiLSTM的短波通信频率预测研究[J]. 通信技术, 2020,53(6): 1311-1318.
	XIA J Y , ZHANG H Y , XU C ,et al. Frequency prediction of HF communication based on CNN-BiLSTM[J]. Communications Technology, 2020,53(6): 1311-1318.
[26]	沈琪琪, 朱德生 . 短波通信[M]. 西安: 西安电子科技大学出版社, 1989.
	SHEN Q Q , ZHU D S . Short wave communication[M]. Xi’an: Xidian University Press, 1989.
[27]	XU Z B , SUN J . Model-driven deep-learning[J]. National Science Review, 2018,5(1): 22-24.
[28]	王金龙, 陈瑾, 徐煜华 . 短波通信技术研究进展与发展需求[J]. 陆军工程大学学报, 2022(1): 1-7.
	WANG J L , CHEN J , XU Y H . On research advances and development requirements of high frequency communication technologies[J]. Journal of Army Engineering University of PLA, 2022(1): 1-7.
[29]	YAN Y S , LIU J M , XU S ,et al. MD3Net:integrating model-driven and data-driven approaches for pansharpening[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022,60: 1-16.
[30]	ZHANG H W , LI Q , SUN Z N ,et al. Combining data-driven and model-driven methods for robust facial landmark detection[J]. IEEE Transactions on Information Forensics and Security, 2018,13(10): 2409-2422.
[31]	YIN Z Q , XU W , XIE R J ,et al. Deep CSI compression for massive MIMO:a self-information model-driven neural network[J]. IEEE Transactions on Wireless Communications, 2022,21(10): 8872-8886.
[32]	HE H T , WEN C K , JIN S ,et al. Model-driven deep learning for joint MIMO channel estimation and signal detection[J]. IEEE Transactions on Signal Processing, 2020,68: 1702-1715.
[33]	SHEN H F , JIANG M H , LI J ,et al. Coupling model- and data-driven methods for remote sensing image restoration and fusion:improving physical interpretability[J]. IEEE Geoscience and Remote Sensing Magazine, 2022,10(2): 231-249.
[34]	HE H T , JIN S , WEN C K ,et al. Model-driven deep learning for physical layer communications[J]. IEEE Wireless Communications, 2019,26(5): 77-83.
[35]	任国春 . 短波通信原理与技术[M]. 北京: 机械工业出版社, 2020.
	REN G C . Shortwave communication principles and technology[M]. Beijing: Mechanical Industry Press, 2020.
[36]	Recommendation ITU-RP. 1239-3.ITU-R reference ionospheric characteristics[S]. 2012.
[37]	KOPKA H , M?LLER H G . MUF calculations including the effect of the earth’s magnetic field[J]. Radio Science, 1968,3(1): 53-56.
[38]	廖柳添 . 短波电台特征动态识别算法研究[D]. 广州:华南理工大学, 2016.
	LIAO L T . Research on dynamic identification algorithm of short-wave radio features[D]. Guangzhou:South China University of Technology, 2016.
[39]	BUCKLEY R , FURMAN W N . Application of machine learning techniques to HF propagation prediction[C]// Proceedings of IEEE Military Communications Conference (MILCOM). Piscataway:IEEE Press, 2021: 623-628.
[40]	LECUN Y , BENGIO Y , HINTON G . Deep learning[J]. Nature, 2015,521(7553): 436-444.
[41]	HOCHREITER S , SCHMIDHUBER J . Long short-term memory[J]. Neural Computation, 1997,9(8): 1735-1780.
[42]	DENG Y P , WANG L , JIA H ,et al. A sequence-to-sequence deep learning architecture based on bidirectional GRU for type recognition and time location of combined power quality disturbance[J]. IEEE Transactions on Industrial Informatics, 2019,15(8): 4481-4493.
[43]	TOLOSANA R , VERA-RODRIGUEZ R , FIERREZ J ,et al. Exploring recurrent neural networks for on-line handwritten signature biometrics[J]. IEEE Access, 2018,6: 5128-5138.
[44]	WANG Q , LI F , TANG Y ,et al. Integrating model-driven and data-driven methods for power system frequency stability assessment and control[J]. IEEE Transactions on Power Systems, 2019,34(6): 4557-4568.
[45]	谢博, 申国伟, 郭春 ,等. 基于残差空洞卷积神经网络的网络安全实体识别方法[J]. 网络与信息安全学报, 2020,6(5): 126-138.
	XIE B , SHEN G W , GUO C ,et al. Cyber security entity recognition method based on residual dilation convolution neural network[J]. Chinese Journal of Network and Information Security, 2020,6(5): 126-138.
[46]	XUAN Y , SI W G , ZHU J ,et al. Multi-model fusion short-term load forecasting based on random forest feature selection and hybrid neural network[J]. IEEE Access, 2021,9: 69002-69009.
[47]	WANG J Y , GRANT T , VELIPASALAR S ,et al. Taking a deeper look at the brain:predicting visual perceptual and working memory load from high-density fNIRS data[J]. IEEE Journal of Biomedical and Health Informatics, 2022,26(5): 2308-2319.

基于模型数据双驱动的短波MUF短期预测网络

Short-term prediction network for short-wave MUF based on model-data dual-driven

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 47

相关文章 3

Metrics

推荐阅读 0

[1]	李国军, 侯旭, 叶昌荣, 罗一平. 短波通信接入网广域协作资源分配算法[J]. 通信学报, 2023, 44(2): 112-121.
[2]	李国军, 向翠玲, 叶昌荣, 王尊立. 短波数字信道化通信探测一体化快速建链方法[J]. 通信学报, 2023, 44(1): 89-102.
[3]	李国军, 郑广发, 叶昌荣, 周银萍. 基于稳健卡尔曼滤波的倾斜探测电离层MUF短期预报方法[J]. 通信学报, 2021, 42(1): 79-86.