基于时空记忆解耦RNN的雷暴预测方法

doi:10.12305/j.issn.1001-506X.2023.11.13

摘要/Abstract

摘要：

使用循环神经网络进行雷暴的外推预测, 利用气象雷达历史反射率因子资料给出未来一小时的雷暴预测结果。网络的核心是时空长短时记忆(spatiotemporal long short-term memory, ST-LSTM)单元, 加入了记忆解耦结构以分离时间记忆和空间记忆状态。在香港天文台(Hong Kong Observatorg, HKO)的HKO-7数据集的基础上筛选雷暴数据, 构建训练及测试数据集。将有记忆解耦结构、无记忆解耦结构的ST-LSTM网络和MIM(memory in memory)网络以及传统的单体质心法进行比较。预报评分因子数值比较和个例分析检验结果表明，预测神经网络在探测成功概率、临界成功指数上均高于单体质心法, 虚警率低于单体质心法。加入记忆解耦结构的网络预报因子评分高于ST-LSTM网络和MIM网络, 雷暴回波外推的预测效果更好, 尤其是强回波的预测效果更好。

关键词: 循环神经网络, 雷暴预测, 气象雷达, 深度学习

Abstract:

The recurrent neural network (RNN) is applied to the thunderstorm prediction, and the prediction results in the next hour are given by using the historical reflectivity factor data of weather radar. The core of the network is spatiotemporal long short-term memory (ST-LSTM) unit, and a memory decoupling structure is added to separate temporal memory and spatial memory states. The training and testing datasets are constructed from thunderstorm data, which are sifted through the HKO-7 dataset of the Hong Kong Observatory. The results of ST-LSTM network with memory decoupling structure and without memory decoupling structure, are compared with the MIM (memory in memory) network and the traditional centroid tracking method. Numerical comparison of prediction score factor and individual case analysis results show that the prediction neural network is higher than the centroid tracking method in probability of detection success and critical success index, and the false alarm ratio is lower than the centroid tracking method. The prediction factor score of the network with memory decoupling structure is higher than that of the conventional ST-LSTM network and the MIM network, and the prediction effect of strong echo is better.

Key words: recurrent neural network (RNN), thunderstorm prediction, weather radar, deep learning

中图分类号:

TN959.4

何诗扬, 汪玲, 朱岱寅, 钱君. 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.

Shiyang HE, Ling WANG, Daiyin ZHU, Jun QIAN. Thunderstorm prediction method based on spatiotemporal memory decoupling RNN[J]. Systems Engineering and Electronics, 2023, 45(11): 3474-3480.

图/表 13

图1

图2

表1

图3

表2

表3

表4

表5

表6

图4

图5

图6

图7

参考文献 22

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	黄兴友, 马玉蓉, 胡苏蔓. 基于深度学习的天气雷达回波序列外推及效果分析[J]. 气象学报, 2021, 79 (5): 817- 827.
	HUANG X Y , MA Y R , HU S M . Extrapolation and effect analysis of weather radar echo sequence based on deep learning[J]. Journal of Meteorology, 2021, 79 (5): 817- 827.
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
4	ZHONG S X, ZENG X X, LING Q, et al. Spatiotemporal con-volutional LSTM for radar echo extrapolation[C]//Proc. of the 54th Asilomar Conference on Signals, Systems, and Computers, 2020: 58-62.
5	YU Y , SI X S , HU C H , et al. A review of recurrent neural networks: LSTM cells and network architectures[J]. Neural Computation, 2019, 31 (7): 1235- 1270. doi: 10.1162/neco_a_01199
6	MIKOLOV T, KARAFIÁT M, BURGET L, et al. Recurrent neural network based language model[C]//Proc. of the 11th Annual Conference of the International Speech Communication Association, 2010.
7	张心宇, 刘源, 宋佳凝. 基于LSTM神经网络的短期轨道预报[J]. 系统工程与电子技术, 2022, 44 (3): 939- 947.
	ZHANG X Y , LIU Y , SONG J N . Short-term orbit prediction based on LSTM neural network[J]. Systems Engineering and Electronics, 2022, 44 (3): 939- 947.
8	何春蓉, 朱江. 基于注意力机制的GRU神经网络安全态势预测方法[J]. 系统工程与电子技术, 2021, 43 (1): 258- 266.
	HE C R , ZHU J . Security situation prediction method of GRU neural network based on attention mechanism[J]. Systems Engineering and Electronics, 2021, 43 (1): 258- 266.
9	胡玉可, 夏维, 胡笑旋, 等. 基于循环神经网络的船舶航迹预测[J]. 系统工程与电子技术, 2020, 42 (4): 871- 877.
	HU Y K , XIA W , HU X X , et al. Vessel trajectory prediction based on recurrent neural network[J]. Systems Engineering and Electronics, 2020, 42 (4): 871- 877.
10	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 Conference on Neural Information Processing Systems, 2015: 802-810.
11	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 Conference on Neural Information Processing Systems, 2017: 5617-5627.
12	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.
13	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.
14	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.
15	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 Machine Intelligence, 2021, 2013- 2029.
16	韩丰, 龙明盛, 李月安, 等. 循环神经网络在雷达临近预报中的应用[J]. 应用气象学报, 2019, 30 (1): 61- 69.
	HAN F , LONG M S , LI Y A , et al. Application of recurrent neural networks in radar proximity forecasting[J]. Journal of Applied Meteorological Science, 2019, 30 (1): 61- 69.
17	WANG Y B, JIANG L, YANG M H, et al. Eidetic 3D LSTM: a model for video prediction and beyond[C]//Proc. of the International Conference on Learning Representations, 2019.
18	WANG Y B, ZHANG J J, ZHU H Y, et al. Memory in me-mory: a predictive neural network for learning higher-order non-stationarity from spatiotemporal dynamics[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9154-9162.
19	YUUKI T , JUN I . Long short-term memory recurrent-neural-network-based bandwidth extension for automatic speech recognition[J]. Acoustical Science and Technology, 2016, 37 (6): 319- 321.
20	HOCHREITER S , SCHMIDHUBER J . Long short-term me-mory[J]. Neural Computation, 1997, 9 (8): 1735- 1780.
21	WANG Z , BOVIK A , SHEIKH H , et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Trans.on Image Processing, 2004, 13 (4): 600- 612.
22	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.

数据集	图片数	序列数
训练集	222 514	9 453
测试集	56 804	2 429
总计	279 318	11 882

预测时间/min	CSI/dBz			POD/dBz			FAR/dBz
预测时间/min	30	40	50	30	40	50	30	40	50
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

预测时间/min	CSI/dBz			POD/dBz			FAR/dBz
预测时间/min	30	40	50	30	40	50	30	40	50
6	0.766	0.618	0.251	0.842	0.694	0.339	0.112	0.178	0.294
30	0.560	0.329	0.059	0.612	0.384	0.069	0.229	0.273	0.356
60	0.225	0.048	0.013	0.362	0.191	0.015	0.247	0.329	0.396

预测方法	CSI/dBz	POD/dBz	FAR/dBz
单体质心法	0.32	0.28	0.45
MIM网络^[22]	0.51	0.58	0.22
ST-LSTM网络	0.56	0.61	0.23
ST-LSTM+记忆解耦网络	0.59	0.65	0.13

[1]	汪萌, 诸兵. 不确定性建模在2D和3D目标检测中的应用[J]. 系统工程与电子技术, 2023, 45(8): 2370-2376.
[2]	邵凯, 杜自群, 王光宇. 基于模型剪枝动态调整压缩率的CSI反馈方法[J]. 系统工程与电子技术, 2023, 45(8): 2615-2622.
[3]	崔天舒, 王栋, 黄振. 面向空间认知通信的轻量化网络自动调制分类方法[J]. 系统工程与电子技术, 2023, 45(7): 2220-2226.
[4]	姜雨, 袁琪, 胡志韬, 吴薇薇, 顾欣. 基于气象因素的机场进离港延误预测[J]. 系统工程与电子技术, 2023, 45(6): 1722-1731.
[5]	陈洋, 廖灿辉, 张锟, 刘建, 王鹏举. 基于自监督对比学习的信号调制识别算法[J]. 系统工程与电子技术, 2023, 45(4): 1200-1206.
[6]	张晔, 侯毅, 欧阳克威, 周石琳. 单变量序列数据分类方法综述[J]. 系统工程与电子技术, 2023, 45(2): 313-335.
[7]	邵正途, 许登荣, 徐文利, 王晗中. 基于LSTM和残差网络的雷达有源干扰识别[J]. 系统工程与电子技术, 2023, 45(2): 416-423.
[8]	陈任飞, 彭勇, 李忠文. 基于持续无监督域适应策略的水面漂浮物目标检测方法[J]. 系统工程与电子技术, 2023, 45(11): 3391-3401.
[9]	陶臣嵩, 陈思伟, 肖顺平. 基于深度学习模型的SAR图像间歇采样转发干扰检测[J]. 系统工程与电子技术, 2023, 45(11): 3465-3473.
[10]	李浩然, 熊伟, 崔亚奇. 基于深度特征融合的SAR图像与AIS信息关联方法[J]. 系统工程与电子技术, 2023, 45(11): 3491-3497.
[11]	刘子超, 王江, 何绍溟. 基于深度学习的时间角度控制制导律[J]. 系统工程与电子技术, 2023, 45(11): 3579-3587.
[12]	季策, 宋博翰, 耿蓉, 梁敏骏. 快时变信道下基于深度学习的OFDM系统信道估计[J]. 系统工程与电子技术, 2023, 45(11): 3649-3655.
[13]	丁建立, 张琪琪, 王静, 霍纬纲. 基于Transformer-VAE的ADS-B异常检测方法[J]. 系统工程与电子技术, 2023, 45(11): 3680-3689.
[14]	宋婷, 兀泽朝, 高艾, 袁建平. 基于CycleGAN的月表图像数据增强方法[J]. 系统工程与电子技术, 2023, 45(10): 3041-3048.
[15]	李瑞泽, 张双辉, 刘永祥. 基于卷积ADMM网络的高效结构化稀疏ISAR成像方法[J]. 系统工程与电子技术, 2023, 45(1): 56-70.