基于轻量化三维时域卷积网络（3D-TCN）的气象雷达雷暴预测方法

doi:10.12305/j.issn.1001-506X.2025.12.10

系统工程与电子技术 ›› 2025, Vol. 47 ›› Issue (12): 4084-4092.doi: 10.12305/j.issn.1001-506X.2025.12.10

• 传感器与信号处理 • 上一篇

基于轻量化三维时域卷积网络（3D-TCN）的气象雷达雷暴预测方法

陆恩绩¹^,², 汪玲¹^,²^,*, 朱岱寅¹^,², 周晔³

1. 南京航空航天大学电子信息工程学院，江苏南京 211106
2. 南京航空航天大学雷达成像与微波光子技术教育部重点实验室，江苏南京 210016
3. 中国航空工业集团雷华电子技术研究所，江苏无锡 214063

收稿日期:2024-09-27 修回日期:2025-01-16 出版日期:2025-04-17 发布日期:2025-04-17
通讯作者: 汪玲
作者简介:陆恩绩（2000—），男，硕士研究生，主要研究方向为气象雷达雷暴识别与跟踪
朱岱寅（1974—），男，教授，博士，主要研究方向为雷达、雷达成像
周　晔（1982—），女，高级工程师，硕士，主要研究方向为机载气象雷达系统和信号处理技术
基金资助:
工信部民机项目资助课题

Weather radar thunderstorm prediction method based on lightweight three-dimensional temporal convolutional network

Enji LU¹^,², Ling WANG¹^,²^,*, Daiyin ZHU¹^,², Ye ZHOU³

1. College of Electronic and Information Engineering，Nanjing University of Aeronautics and Astronautics，Nanjing 211106，China
2. Key Laboratory of Radar Imaging and Microwave Photonics，Ministry of Education，Nanjing University of Aeronautics and Astronautics，Nanjing 210016，China
3. Leihua Electronic Technologv Research Institute，Aviation Industry Corporation of China，Wuxi 214063，China

Received:2024-09-27 Revised:2025-01-16 Online:2025-04-17 Published:2025-04-17
Contact: Ling WANG

摘要/Abstract

摘要：

针对现有大多数基于循环神经网络的雷暴预测模型普遍存在梯度爆炸、长时依赖关系捕捉能力薄弱以及因无法并行计算导致的效率低下等问题，提出三维时域卷积神经网络（three-dimensional time-domain convolutional neural network，3D-TCN）用于气象雷达雷暴预测。在TCN核心的残差块中引入三维卷积操作，从雷达图像中提取时空特征，实现最长1 h的精准预测。基于数据集对3D-TCN模型的雷暴演变预测性能进行验证，并与时空记忆解耦的空时长短时记忆（spationtemporal long short-term memory，ST-LSTM）网络、记忆增强网络及传统交叉相关算法展开对比。实验结果表明，3D-TCN 模型在临界成功指数等评价指标上表现更优，且凭借轻量化结构大幅缩短训练时间。与时空记忆解耦的 ST-LSTM 网络相比，其临界成功指数平均提升0.037，最高提升幅度接近0.13，训练时间则从4~6 min缩短至1 min以内，验证了所提方法的有效性。

关键词: 时序卷积网络, 气象雷达, 雷暴预测, 深度学习

Abstract:

To address the problem of gradient explosion, weak ability to capture long-term dependencies, and low efficiency caused by the inability to perform parallel computing in most existing thunderstorm prediction models based on recurrent neural networks (RNN), a three-dimensional time-domain convolutional neural network (3D-TCN) is proposed for meteorological radar thunderstorm prediction. A 3D convolution operation is introduced into the residual block of TCN core to extract spatiotemporal features from radar images, achieving accurate prediction for up to 1 h. The thunderstorm evolution prediction performance of the 3D-TCN model is validated based on the Hong Kong Observatory 7 (HKO-7) dataset, and is compared with the spatiotemporal memory decoupled spatiotemperal long short-term memory (ST-LSTM) network, memory enhancement network, and traditional cross-correlation algorithms. The experimental results show that the 3D-TCN model performs better in evaluation metrics such as critical success index, and significantly reduces training time with its lightweight structure. Compared with the ST-LSTM network decoupled from spatiotemporal memory, its critical success index has increased by an average of 0.037, with the highest increase approaching 0.13. The training time has been shortened from 4 min to 6 min to less than 1 min, which verifies the effectiveness of the proposed method.

Key words: temporal convolutional networks (TCN), weather radar, thunderstorm prediction, deep learning

中图分类号:

TN 959.4

陆恩绩, 汪玲, 朱岱寅, 周晔. 基于轻量化三维时域卷积网络（3D-TCN）的气象雷达雷暴预测方法[J]. 系统工程与电子技术, 2025, 47(12): 4084-4092.

Enji LU, Ling WANG, Daiyin ZHU, Ye ZHOU. Weather radar thunderstorm prediction method based on lightweight three-dimensional temporal convolutional network[J]. Systems Engineering and Electronics, 2025, 47(12): 4084-4092.

图/表 15

图1

图2

图3

图4

图5

表1

表2

表3

表4

图6

图7

图8

图9

表5

表6

参考文献 30

4	RINEHART R E, GARVEY E T. Three-dimensional storm motion detection by conventional weather radar[J]. Nature, 1978, 273 (5660): 287- 289. doi: 10.1038/273287a0
5	SHI X J, CHEN Z R, WANG H, et al. Convolutional LSTM network: a machine learning approach for precipitation nowcasting[C]// Proc. of the 29th International Conference on Neural Information Processing Systems, 2015: 802−810.
6	JING J R, LI Q, PENG X, et al. HPRNN: a hierarchical sequence prediction model for long-term weather radar echo extrapolation[C]// Proc. of the IEEE International Conference on Acoustics, Speech and Signal Processing, 2020: 4142−4146.
7	WANG Y B, LONG M S, WANG J M, et al. PredRNN: recurrent neural networks for predictive learning using spatiotemporal LSTMs[C]// Proc. of the 31st International Conference on Neural Information Processing Systems, 2017: 879−888.
8	WANG Y B, GAO Z F, LONG M S, et al. PredRNN++: towards a resolution of the deep-in-time dilemma in spatiotemporal predictive learning[C]// Proc. of the 35th International Conference on Machine Learning, 2018, 80: 5123−5132.
9	WANG Y B, WU H X, ZHANG J J, et al. PredRNN: a recurrent neural network for spatiotemporal predictive learning [J]. IEEE Trans. on Pattern Analysis and Aachine Intelligence, 2023, 45(2): 2208−2225.
10	何诗扬, 汪玲, 朱岱寅, 等. 基于时空记忆解耦RNN的雷暴预测方法[J]. 系统工程与电子技术, 2023, 45 (11): 3474- 3480.
	HE S Y, WANG L, ZHU D Y, et al. Thunderstorm prediction method based on spatio temporal memory decoupling RNN[J]. Systems Engineering and Electronics, 2023, 45 (11): 3474- 3480.
11	CHEN Q, LIU Y B, GE M F, et al. A novel bayesian-optimization-based adversarial TCN for RUL prediction of bearings[J]. IEEE Sensors Journal, 2022, 22 (21): 20968- 20977. doi: 10.1109/JSEN.2022.3209894
12	TONG C, ZHANG L H, LI H, et al. Temporal inception convolutional network based on multi-head attention for ultra-short-term load forecasting [J]. IET Generation, Transmission & Distribution, 2022, 16(8): 1680−1696.
13	ALTAHERI H, MUHAMMAD G, ALSULAIMAN M. Physics-informed attention temporal convolutional network for EEG-based motor imagery classification[J]. IEEE Trans. on Industrial Informatics, 2022, 19 (2): 2249- 2258.
14	李云峰, 高云鹏, 张蓬鹤, 等. 多目标优化时域卷积神经网络的窃电行为高准确检测算法 [J]. 电网技术, 2024, 48(8): 3449−3458.
	LI Y F, GAO Y P, ZHANG P H, et al. High-accuracy detection algorithm of electricity theft behavior based on multi-objective optimization time-domain convolutional neural network [J]. Power System Technology, 2024, 48(8): 3449−3458.
15	ZHANG B, WANG S, DENG L, et al. Ship motion attitude prediction model based on IWOA-TCN-Attention[J]. Ocean Engineering, 2023, 272, 113911. doi: 10.1016/j.oceaneng.2023.113911
16	SHU J, LIAO Y C, LI J H. Spatial-temporal attention TCN-based link prediction for opportunistic network[J]. Electronics, 2024, 13 (5): 957- 970. doi: 10.3390/electronics13050957
17	张梦凡, 丁兵兵, 贾国栋, 等. 基于TCN-BiLSTM与LSTM模型对比预测北洛河径流[J]. 北京林业大学学报, 2024, 46 (4): 141- 148.
	ZHANG M F, DING B B, JIA G D, et al. Comparative prediction of runoff in the Beiluo River, Shaanxi Province of Northwestern China based on TCN-BiLSTM and LSTM models[J]. Journal of Beijing Forestry University, 2024, 46 (4): 141- 148.
18	焦卫东, 杨蓓. MDAT-Net: 一种融合MSC和时空注意力的TCN航迹预测方法 [EB/OL]. [2024-09-27]. https://doi.org/10.13700/j.bh.1001-5965.2023.0717.
	JIAO W D, YANG B. MDAT-Net: a TCN trajectory prediction method fusing MSC and spatio-temporal attention [EB/OL]. [2024-09-27]. https://doi.org/10.13700/j.bh.1001-5965.2023.0717.
19	杨达, 刘家威, 郑斌, 等. 基于时域卷积网络与注意力机制的车辆换道轨迹预测模型[J]. 交通运输系统工程与信息, 2024, 24 (2): 114- 126.
	YANG D, LIU J W, ZHENG B, et al. A vehicle lane-changing trajectory prediction model based on temporal convolutional networks and attention mechanism[J]. Journal of Transportation Systems Engineering and Information Technology, 2024, 24 (2): 114- 126.
20	YE W, KUANG H, DENG K, et al. LGTCN: a spatial–temporal traffic flow prediction model based on local-global feature fusion temporal convolutional network[J]. Applied Sciences, 2024, 14 (19): 8847- 8862. doi: 10.3390/app14198847
21	SHI X J, GAO Z H, LAUSEN L, et al. Deep learning for precipitation nowcasting: a benchmark and a new model[C]//Proc. of the 31st International Conference on Neural Information Processing Systems, 2017: 5622−5632.
22	AYZEL G, HEISTERMANN M, WINTERRATH T. Optical flow models as an open benchmark for radar-based precipitation nowcasting （rainymotion v0.1）[J]. Geoscientific Model Development, 2019, 12 (4): 1387- 1402. doi: 10.5194/gmd-12-1387-2019
23	QIN Y J, GAN F F, XIA B Z, et al. Remaining useful life estimation of bearing via temporal convolutional networks enhanced by a gated convolutional unit[J]. Engineering Applications of Artificial Intelligence, 2024 (133): 108308.
1	MULLER R, BARLEBEN A, HAUSSLER S, et al. A novel approach for the global detection and nowcasting of deep convection and thunderstorms[J]. Remote Sensing, 2022, 14 (14): 3372- 3384. doi: 10.3390/rs14143372
2	D’AMATO G, VITALE C, D’AMATO M, et al. Thunderstorm-related asthma: what happens and why[J]. Clinical and Experimental Allergy, 2016, 46 (3): 390- 396. doi: 10.1111/cea.12709
3	ZHANG F G, LAI C, CHEN W J. Weather radar echo extrapolation method based on deep learning[J]. Atmosphere, 2022, 13 (5): 815- 834. doi: 10.3390/atmos13050815
24	RAZA A, UDDIN J, ALMUHAIMEED A, et al. AIPs-SnTCN: predicting anti-inflammatory peptides using fastText and transformer encoder-based hybrid word embedding with self-normalized temporal convolutional networks[J]. Chemical Information and Modeling, 2023, 63 (21): 6537- 6554. doi: 10.1021/acs.jcim.3c01563
25	OORD D V A, KALCHBRENNER N, KAVUKCUOGLU K. Pixel recurrent neural networks [C]//Proc. of the International Conference on Machine Learning, 2016: 1747−1756.
26	CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Deeplab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFS[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2017, 40 (4): 834- 848.
27	OORD D V A, DIELEMAN S, ZEN H, et al. WaveNet: a generative model for raw audio [EB/OL]. [2024-09-27]. https://www.arxiv.org.pdf.1609.03499.
28	GILLELAND E, AHIJEVYCH D, BROWN B G, et al. Intercomparison of spatial forecast verification methods[J]. Weather and Forecasting, 2009, 24 (5): 1416- 1430. doi: 10.1175/2009WAF2222269.1
29	MARZBAN C. Scalar measures of performance in rare-event situations[J]. Weather and Forecasting, 1998, 13 (3): 753- 763. doi: 10.1175/1520-0434(1998)013<0753:SMOPIR>2.0.CO;2
30	龚勋, 胡嘉骏, 徐年平, 等. 基于深度学习的多普勒气象雷达回波外推短临预报对比研究[J]. 中国军转民, 2022 (13): 76- 80.
	GONG X, HU J J, XU N P, et al. A comparative study of Doppler weather radar echo extrapolation short prognosis based on deep learning[J]. Defense Industry Conversion in China, 2022 (13): 76- 80.

预测	真实
预测	有雷暴	无雷暴
有雷暴	NA	NB
无雷暴	NC	ND

预测时间/min	CSI			POD			FAR
预测时间/min	30 dBz	40 dBz	50 dBz	30 dBz	40 dBz	50 dBz	30 dBz	40 dBz	50 dBz
6	0.833	0.640	0.377	0.875	0.675	0.433	0.055	0.156	0.168
30	0.640	0.389	0.137	0.737	0.492	0.190	0.126	0.180	0.232
60	0.462	0.243	0.057	0.565	0.274	0.059	0.171	0.278	0.295

预测时间/min	CSI			POD			FAR
预测时间/min	30 dBz	40 dBz	50 dBz	30 dBz	40 dBz	50 dBz	30 dBz	40 dBz	50 dBz
6	0.815	0.653	0.360	0.865	0.715	0.428	0.069	0.130	0.170
30	0.590	0.358	0.135	0.649	0.412	0.164	0.131	0.203	0.308
60	0.335	0.149	0.053	0.475	0.228	0.054	0.148	0.240	0.309

预测方法	CSI	POD	FAR
交叉相关法	0.35	0.30	0.46
MIM网络^[30]	0.51	0.58	0.22
ST-LSTM+记忆解耦网络	0.59	0.65	0.13
3D-TCN	0.64	0.74	0.13

模型	CSI	POD	FAR
无膨胀卷积版本	0.57	0.68	0.19
完整模型	0.64	0.74	0.13

基于轻量化三维时域卷积网络（3D-TCN）的气象雷达雷暴预测方法

Weather radar thunderstorm prediction method based on lightweight three-dimensional temporal convolutional network

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 15

编辑推荐

Metrics

本文评价

模型	CSI	POD	FAR	收敛迭代数
无残差连接版本	0.58	0.68	0.16	35000
完整模型	0.64	0.74	0.13	20000

[1]	李开明, 张袁鹏, 罗迎, 代肖楠. 弹道导弹雷达目标识别研究进展[J]. 系统工程与电子技术, 2025, 47(9): 2870-2889.
[2]	季策, 马相宇, 牟晓宇, 赵家毅. TS-GRU-VTA：基于深度学习的车辆信道估计方案[J]. 系统工程与电子技术, 2025, 47(9): 3093-3098.
[3]	闫小伟, 凌冲, 石胜斌. 地表未爆子弹药快速检测系统设计与实现[J]. 系统工程与电子技术, 2025, 47(8): 2639-2645.
[4]	张琬滢, 高由兵, 李泽一, 李鹏飞, 张伟. 基于双路径特征融合网络的背景信号抑制算法[J]. 系统工程与电子技术, 2025, 47(7): 2406-2413.
[5]	张新征, 闫梦可, 朱晓林. 噪声伪标签容忍的半监督SAR目标识别[J]. 系统工程与电子技术, 2025, 47(6): 1796-1805.
[6]	姜智杰, 宋恒, 胡楠, 段兰茜, 曹平. 隧道环境毫米波雷达目标识别与分类算法[J]. 系统工程与电子技术, 2025, 47(5): 1453-1460.
[7]	聂千祁, 沙明辉, 朱应申, 王崇宇, 崔念强. 基于盲源分离结合奇异谱分析的雷达多分量信号识别方法[J]. 系统工程与电子技术, 2025, 47(4): 1168-1175.
[8]	袁磊, 杨艳娟, 郭毅, 戴鹏. 相关噪声下基于深度学习的LDPC码码率半盲识别算法[J]. 系统工程与电子技术, 2025, 47(4): 1335-1345.
[9]	沈夏闰, 李若楠, 张昊田. 基于CVAE-LSTM的服务器KPI异常检测[J]. 系统工程与电子技术, 2025, 47(3): 1019-1027.
[10]	刘晓琳, 郭梦娇, 李卓. Dueling DQN优化下的航班延误自适应图卷积循环网络预测方法[J]. 系统工程与电子技术, 2025, 47(2): 568-579.
[11]	付卫红, 张鑫钰, 刘乃安. 基于多尺度融合神经网络的同频同调制单通道盲源分离算法[J]. 系统工程与电子技术, 2025, 47(2): 641-649.
[12]	张文中, 侯长波, 赵鹏旗, 刘思成. 基于GCN的无人机目标航迹关联[J]. 系统工程与电子技术, 2025, 47(12): 4005-4011.
[13]	姜杰, 闫文君, 凌青, 张立民. 基于STOD的无人机舰船图像微小目标检测方法[J]. 系统工程与电子技术, 2025, 47(11): 3559-3567.
[14]	高飞, 席睿达, 邢相薇, 邢妍, 张强, 党红杏. SAR有源干扰分类识别图像数据集仿真构建[J]. 系统工程与电子技术, 2025, 47(10): 3257-3269.
[15]	赵昱宇, 刘耀蓬, 王雨潇. 融合机理与数据的陀螺飞轮姿态测量方法[J]. 系统工程与电子技术, 2025, 47(10): 3455-3463.