[1]黄颖,陆虹,黄小燕,等.基于EOF和LSTM的广西月降水量预测模型研究[J].气象研究与应用,2023,44(02):20-26.[doi:10.19849/j.cnki.CN45-1356/P.2023.2.04]
　Huang Ying,Lu Hong,Huang Xiaoyan,et al.Study on monthly precipitation prediction model in Guangxi based on EOF and LSTM[J].Journal of Meteorological Research and Application,2023,44(02):20-26.[doi:10.19849/j.cnki.CN45-1356/P.2023.2.04]
点击复制

基于EOF和LSTM的广西月降水量预测模型研究()
分享到：

气象研究与应用[ISSN:1673-8411/CN:45-1356/P]

卷:
第44卷
期数:
2023年02期
页码:
20-26
栏目:
研究论文
出版日期:
2023-06-15

文章信息/Info

Title:
Study on monthly precipitation prediction model in Guangxi based on EOF and LSTM
作者:
黄颖1;  陆虹2;  黄小燕1;  赵华生1;  吴玉霜3
1. 广西壮族自治区气象科学研究所, 南宁 530022;
2. 广西壮族自治区气候中心, 南宁 530022;
3. 广西壮族自治区气象台, 南宁 530022
Author(s):
Huang Ying1;  Lu Hong2;  Huang Xiaoyan1;  Zhao Huasheng1;  Wu Yushang3
1. Guangxi Institute of Meteorological Sciences, Nanning 530022, China;
2. Guangxi Climate Center, Nanning 530022, China;
3. Guangxi Meteorological Observatory, Nanning 530022, China
关键词:
气候预测; 长短期记忆神经网络; 自然正交展开; 深度学习
Keywords:
Monthly precipitation prediction; long and short-term memory neural network; empirical orthogonal function; deep learning
分类号:
P457.6
DOI:
10.19849/j.cnki.CN45-1356/P.2023.2.04
摘要:
针对夏季降水天气过程具有时间相关性和非线性变化的特点,以及现有预报方法未能充分获取月降水量的本质特征而造成的建模因子处理和预报建模困难等问题,提出了一种以自然正交展开(EOF)与深度学习长短期记忆神经网络(LSTM)相结合的月降水量预测模型。以广西81个气象观测站7月降水量为预报研究对象,对81站7月降水量作EOF计算,选取累积方差贡献超过76%的前7个主分量作为预报分量,再利用LSTM模型建立月降水量的深度学习预测模型,以1960-2016年81站7月降水量为建模样本,2017-2022年为独立样本进行建模研究。结果表明,在相同的预报建模样本和相同的预报因子条件下,新建立的预测模型比线性逐步回归预报方法有更高的预报能力,显示了对非线性月降水量预测问题的适用性。由于LSTM模型隐层里引入了存储单元状态和门结构,使得网络能够保留长期的状态,更适合于处理和预测时间序列中间隔和延迟相对较长的重要问题。
Abstract:
In view of the characteristics of time dependence and nonlinear changes of precipitation weather process in summer, and the essential characteristics of monthly precipitation that are failed to fully obtain from the existing forecasting methods, it is difficult to deal with modeling factors and forecast modeling. To solve this problem, a monthly precipitation prediction model based on the Empirical Orthogonal Function (EOF) and short-term memory neural network(LSTM) of deep learning has been proposed in this paper. Taking the July precipitation of 81 stations in Guangxi as the forecast object, EOF is calculated for the July precipitation of 81 stations, and the first seven principal components with cumulative variance contribution of more than 76% are selected as the forecast components. Then a deep learning prediction model has been established using LSTM model for monthly precipitation, with the July precipitation at 81 stations from 1960 to 2016 as the modeling samples and 2017-2022 as independent samples. Results show that under the same modeling samples and factors, the newly established forecast model has a higher forecast ability than the linear stepwise regression forecast method, demonstrating its applicability to the nonlinear monthly precipitation forecast problem. Further analysis shows that the LSTM model introduces the storage cell state and gate structure in the hidden layer, which allows the network to retain long-term states, making it more suitable for processing and predicting the important problems with relatively long interval and delay in the time series.

参考文献/References:

[1] 丁一汇. 我国短期气候预测业务系统[J].气象,2004,30(12):11-16.
[2] 李维京,张培群,李清泉,等. 动力气候模式预测系统业务化及其应用[J]. 应用气象学报,2005(增刊1):1-11.
[3] SCHEPEN A,WANG Q J,ROBERTSON D E. Combining the strengths of statistical and dynamical modeling approaches for forecasting Australian seasonal rainfall[J]. Journal of Geophysical Research:Atmospheres,2012, 117(D20).
[4] WEI F Y,XIE Y,MANN M E.Probabilistic trend of anomalous summer rainfall in Beijing:role of interdecadal variability[J]. Journal of Geophysical Research:Atmospheres, 2008,113(D20).
[5] 杜良敏,柯宗建,刘长征,等.基于聚类分区的中国夏季降水预测模型[J].气象, 2016,42(1):89-96.
[6] 郑然, 刘嘉慧敏, 马振峰.年际增量方法在西南夏季降水预测中的应用[J].气象学报,2019,77(3):489-496.
[7] 覃志年, 陈丽娟, 唐红玉, 等.月尺度动力模式产品解释应用系统及预测技巧[J].应用气象学报,2010,21(5):614-620.
[8] 何慧, 金龙, 覃志年, 等.动力延伸预报产品在广西月降水预报中的应用[J].应用气象学报,2007(5):727-731.
[9] 何慧,陆虹,覃卫坚,等.人工神经网络在月降水量预测业务中的研究和应用综述[J].气象研究与应用,2021,42(1):1-6.
[10] KARIMA N,BOUDHAR A,OUATIKI H,et al.Deep learning approach with LSTM for daily streamflow prediction in a semi-arid area:a case study of Oum Er-Rbia River Basin, Morocco[J]. Water,2023,15(2):262.
[11] FAN G J, YU P Y, WANG Q, et al. Short-term motion prediction of a semi-submersible by combining LSTM neural network and different signal decomposition methods[J]. Ocean Engineering,2023.
[12] LIU H J, LEI D X, YUAN J, et al. Ionospheric TEC prediction in China based on the multiple-attention LSTM model[J]. Atmosphere,2022,13(11):1939.
[13] THANONGSAK X,Lee H,Lee G. Air pollution prediction using Long Short-Term Memory(LSTM) and Deep Autoencoder (DAE) Models[J]. Sustainability,2020,12(6):2570.
[14] 孙际钰. 基于LSTM网络的海表温度预报技术研究[D].大连:大连海洋大学, 2022.
[15] 沈皓俊, 罗勇, 赵宗慈, 等.基于LSTM网络的中国夏季降水预测研究[J].气候变化研究进展,2020,16(3):263-275.
[16] 程鹏宇,赵嘉,韩龙哲,等.双向多尺度LSTM的短时温度预测[J].江西师范大学学报(自然科学版),2022,46(2):134-139.
[17] 倪铮,梁萍.基于LSTM深度神经网络的精细化气温预报初探[J].计算机应用与软件,2018, 35(11):233-236.
[18] 徐楠楠. 基于LSTM的中国大陆地区ERA5日尺度降水预测方法研究[D].南京:南京邮电大学, 2021.
[19] 张飞鸿. 基于对抗型LSTM和时序残差网络的短临降水预测研究[D].南京:南京信息工程大学,2020.
[20] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[21] 陈剑飞,李勇,刘俊江,等.耦合陆面水文模型和机器学习方法的水库径流量预报及应用[J].气象研究与应用,2022,43(1):1-7.
[22] 黄启桥,麦雄发,李玲,等.基于ConvLSTM的广西短临降水预报[J].气象研究与应用,2021,42(4):44-49.
[23] 梁振清,陈生.基于深度学习和雷达观测的华南短临预报精度评估[J].气象研究与应用,2020,41(1):41-47.
[24] 张明,杜裕,洪国平.LSTM在宜昌短期电力负荷预测中的试验应用[J].气象研究与应用,2019,40(3):72-77.
[25] FELLX A G, JURGEN S, FRED C. Learning to forget:continual prediction with LSTM[C].International Conference on Artificial Neural Networks,Edinburgh,UK,1999(2):850-855.
[26] 张进, 牛志伟, 齐慧君,等. K-means++和LSTM模型在大坝监测资料分析中的应用[J].三峡大学学报(自然科学版), 2021,43(4):25-30.
[27] GERS F A,SCHMIDHUBER E. LSTM recurrent networks learn simple context-free and context-sensitive languages[J].IEEE Transactions on Neural Networks,2001,12(6):1333-1340.
[28] GRAVES A. Supervised sequence labelling with recurrent neural networks[C]. Springer-Verlag Berlin Heidelberg, 2012.
[29] 吴洪宝, 吴蕾. 气候变率诊断和预测方法[M]. 北京:气象出版社,2005.
[30] 覃卫坚,苗春生,陈丽,等.EOF在广西春季降水分析中的应用[J].气象研究与应用, 2007(2):42-45.

相似文献/References:

[1]陈思蓉,周秀华,陆虹,等.月动力延伸预测产品DERF2.0对广西气温和降水的预测评估[J].气象研究与应用,2016,37(03):16.
　Chen Si-rong,Zhou Xiu-hua,He Hui-gen.Temperature and precipitation evaluation of monthly Dynamic extended range forecast Operational system DERF2.0 in Guangxi[J].Journal of Meteorological Research and Application,2016,37(02):16.
[2]何慧,陆虹,覃卫坚,等.人工神经网络在月降水量预测业务中的研究和应用综述[J].气象研究与应用,2021,42(01):1.[doi:10.19849/j.cnki.CN45-1356/P.2021.1.01]
　He Hui,Lu Hong,Qin Weijian,et al.Research and application of artificial neural network in monthly precipitation forecast[J].Journal of Meteorological Research and Application,2021,42(02):1.[doi:10.19849/j.cnki.CN45-1356/P.2021.1.01]
[3]覃卫坚,何莉阳,蔡悦幸.基于两种机器学习方法的广西后汛期降水预测模型[J].气象研究与应用,2022,43(01):8.[doi:10.19849/j.cnki.CN45-1356/P.2022.1.02]
　Qin Weijian,He Liyang,Cai Yuexing.Prediction model of post-flood season precipitation in Guangxi based on two machine learning methods[J].Journal of Meteorological Research and Application,2022,43(02):8.[doi:10.19849/j.cnki.CN45-1356/P.2022.1.02]

备注/Memo

备注/Memo:
收稿日期:2023-3-1。
基金项目:广西自然科学基金项目(2023GXNSFAA026414)、国家自然科学基金项目(42065004)、广西重点研发计划项目(桂科AB21196041)、广西自然科学基金项目(2018GXNSFAA281229)
作者简介:黄颖(1983-),女,硕士,高级工程师,主要从事天气预报技术方法研究与业务应用工作。E-mail:yinger2001@126.com
更新日期/Last Update: 1900-01-01