基于EEMD-LSTM模型的天山北坡经济带年降水量预测

doi:10.13866/j.azr.2021.05.05

摘要/Abstract

摘要：

降水量预测是现代气候预测业务的核心和难点。耦合模型在新疆降水量预测的研究应用屈指可数,因此,通过尝试建立集合经验模态分解（Ensemble Empirical Mode Decomposition,EEMD）和长短期记忆神经网络（Long Short-Term Memory Network,LSTM）的耦合模型对天山北坡经济带降水量进行预测研究。将1965—2019年天山北坡经济带共55 a的年降水量数据进行EEMD分解,转换成4个平稳分量和趋势项,通过谱分析得出各个分量的准周期,为后续训练LSTM模型提供基础。根据EEMD分解后的各分量训练得出LSTM网络模型并利用该模型进行研究区降水量预测。结果表明：EEMD-LSTM耦合模型预测2010—2019年天山北坡经济带降水量的平均相对误差为13.38%,均方根误差为38.03 mm,认为EEMD-LSTM耦合模型对天山北坡经济带降水量预测精度较好。利用EEMD-LSTM耦合模型预测2020—2029年天山北坡经济带年降水量,其中有6 a降水偏多,4 a降水偏少,2025年可能为极端湿润年,降水偏多超过20%;而2021年为极端干旱年,降水量预计低于200 mm。本文探索了干旱区降水量预测的新方法,并为气象防灾减灾工作提供参考依据。

关键词: 集合经验模态分解, 长短期记忆网络, 年降水量, 预测

Abstract:

Precipitation prediction is both an essential and challenging component of modern climate prediction. In the precipitation prediction in Xinjiang, the research and application of coupling models are very limited. Therefore, this paper attempts to establish the ensemble empirical mode decomposition (EEMD) and long short-term memory (LSTM) neural network coupling models to predict precipitation in the Northern Slope Economic Belt of Tianshan Mountains. Firstly, the precipitation data recorded in the study area during 55 years, from 1965 to 2019, were decomposed into four stationary components and trend terms using the EEMD, and the quasi period of each component was obtained by spectral analysis, which provided the basis for the subsequent training of the LSTM model. Then, each EEMD component was trained into the LSTM network model, and the models were used for predictions. The reconstruction and comparison of the results revealed that the average relative error and root mean square error of the 2010-2019 model were 13.38% and 38.03 mm, respectively. Therefore, it is inferred that the EEMD-LSTM coupling model can achieve a better precipitation prediction accuracy in the study area. The model was used to predict annual precipitation in the Northern Slope Economic Belt of Tianshan Mountains from 2020 to 2029; within this period, six years presented more and four years presented less precipitation. Year 2025 is predicted to be an extremely humid year with more than 20% of precipitation in excess; while 2021 is anticipated as an extremely dry year, with expected precipitation amounting to less than 200 mm. This study explored a new method of precipitation prediction in an arid area, and provided a reference for meteorological disaster prevention and mitigation.

Key words: ensemble empirical mode decomposition, long short-term memory network, annual precipitation, prediction

杨倩,秦莉,高培,张瑞波. 基于EEMD-LSTM模型的天山北坡经济带年降水量预测[J]. 干旱区研究, 2021, 38(5): 1235-1243.

YANG Qian,QIN Li,GAO Pei,ZHANG Ruibo. Prediction of annual precipitation in the Northern Slope Economic Belt of Tianshan Mountains based on a EEMD-LSTM model[J]. Arid Zone Research, 2021, 38(5): 1235-1243.

图/表 11

图1

图2

图3

图4

表1

表2

图5

表3

图6

表4

表5

参考文献 29

[1]	刘颖, 任宏利, 张培群, 等. 中国夏季降水的组合统计降尺度模型预测研究[J]. 气候与环境研究, 2020, 25(2): 163-171.
	[ Liu Ying, Ren Hongli, Zhang Peiqun, et al. Application of the hybrid statistical downscaling model in summer precipitation prediction in China[J]. Climatic and Environmental Research, 2020, 25(2): 163-171. ]
[2]	任冉. 基于集合经验模态分解(EMD-EEMD)的中国夏季降水预报方法的研究[D]. 南京: 南京信息工程大学, 2014.
	[ Ren Ran. The Research about the Summer Precipitation Prediction in China Based on EEMD Method[D]. Nanjing: Nanjing University of Information Science & Technology, 2014. ]
[3]	迟道才, 张特男, 吴秀明, 等. ARMA模型在太子河流域年降水量预测中的应用[J]. 沈阳农业大学学报, 2012, 43(3): 607-610.
	[ Chi Daocai, Zhang Tenan, Wu Xiuming, et al. Application of ARMA model in the annual precipitation forecast of Taizi River[J]. Journal of Shenyang Agriculture University, 2012, 43(3): 607-610. ]
[4]	陈沪生, 周玉良, 周平, 等. 基于小波和ARIMA的黄山市年降水量分析及预测[J]. 南水北调与水利科技, 2019, 17(5): 50-55.
	[ Chen Husheng, Zhou Yuliang, Zhou Ping, et al. Analysis and prediction of annual precipitation in Huangshan City based on wavelet and ARIMA[J]. South-to-North Water Transfers and Water Science & Technology, 2019, 17(5): 50-55. ]
[5]	李佳秀, 杜春丽, 杜世飞, 等. 新疆极端降水事件的时空变化及趋势预测[J]. 干旱区研究, 2015, 32(6): 60-69.
	[ Li Jiaxiu, Du Chunli, Du Shifei, et al. Temporal spatial variation and trend prediction of extreme precipitation events in Xinjiang[J]. Arid Zone Research, 2015, 32(6): 60-69. ]
[6]	薛春芳, 侯威, 赵俊虎, 等. 集合经验模态分解在区域降水变化多尺度分析及气候变化响应研究中的应用[J]. 物理学报, 2013, 62(10): 1-7.
	[ Xue Chunfang, Hou Wei, Zhao Junhu, et al. The application of ensemble empirical mode decomposition method in multiscale analysis of region precipitation and its response to the climate change[J]. Acta Physica Sinica, 2013, 62(10): 1-7. ]
[7]	刘天虎, 刘天龙. 集合经验模态分解下中国新疆降水变化趋势的区域特征[J]. 沙漠与绿洲气象, 2015, 9(4): 17-24.
	[ Liu Tianhu, Liu Tianlong. Regional features of precipitation variation trends over Xinjiang in China by the Ensemble Empirical Mode Decomposition method[J]. Desert and Oasis Meteorology, 2015, 9(4): 17-24. ]
[8]	罗那那, 巴特尔·巴克, 吴燕锋. 基于集合经验模态分解北疆降水多尺度变化特征[J]. 水土保持研究, 2017, 24(4): 362-367.
	[ Luo Nana, Bake Batur, Wu Yanfeng. Precipitation Multi-Scale characteristics by Ensemble Empirical Mode Decomposition in Northern Xinjiang[J]. Research of Soil and Water Conservation, 2017, 24(4): 362-367. ]
[9]	党池恒, 张洪波, 陈克宇, 等. 长短期记忆神经网络在季节性融雪流域降水-径流模拟中的应用[J]. 华北水利水电大学学报(自然科学版), 2020, 41(5): 10-18, 33.
	[ Dang Chiheng, Zhang Hongbo, Chen Keyu, et al. Application of the long short-term memory neural network for rainfall-runoff simulation in seasonal snowmelt basin[J]. Journal of North China University of Water Resources and Electric Power (Natural Science Edition), 2020, 41(5): 10-18, 33. ]
[10]	吴麟, 冯利华. 基于BP神经网络的义乌市降水量预测[J]. 浙江水利科技, 2014, 42(2) : 52-54.
	[ Wu Lin, Feng Lihua. Prediction of precipitation in Yiwu based on BP neural networks[J]. Zhejiang Hydrotechnics, 2014, 42(2): 52-54. ]
[11]	沈艳, 杨春雷, 张庆国, 等. 基于RBF神经网络的池州市降水序列预测[J]. 安徽农业大学学报, 2012, 39(3): 451-455.
	[ Shen Yan, Yang Chunlei, Zhang Qingguo, et al. Prediction of precipitation series based on RBF neural network in Chizhou City[J]. Journal of Anhui Agricultural University, 2012, 39(3): 451-455. ]
[12]	张飞涟, 刘严萍. 经验模态分解与神经网络方法在降水预测领域的应用研究[C]// 中国系统工程学会. 中国系统工程学会第十八届学术年会论文集—A01系统工程. 合肥: 中国系统工程学会, 2014.
	[ Zhang Feilian, Liu Yanping. Application of Empirical Mode Decomposition and Neural Network in Precipitation Prediction[C]// Systems Engineering Society of China. Proceedings of the 18th Annual Meeting of Systems Engineering Society of China: A01 Systems Engineering. Hefei: Systems Engineering Society of China, 2014. ]
[13]	宁理科. 地形地貌对天山山区降水的影响研究[D]. 石河子: 石河子大学, 2013.
	[ Ning Like.Study on the Influence of Topography and Geomorphology on Precipitation over Tianshan Mountains, Central Asia[D]. Shihezi: Shihezi University, 2013. ]
[14]	施雅风, 沈永平, 胡汝骥. 西北气候由暖干向暖湿转型的信号、影响和前景初步探讨[J]. 冰川冻土, 2002, 24(3): 219-226.
	[ Shi Yafeng, Shen Yongping, Hu Ruji. Preliminary study on signal, impact and foreground of climatic shift from warm-dry to warm-humid in Northwest China[J]. Journal of Glaciology and Geocryology, 2002, 24(3): 219-226. ]
[15]	方创琳. 天山北坡城市群可持续发展战略思路与空间布局[J]. 干旱区地理, 2019, 42(1): 1-11.
	[ Fang Chuanglin. Strategic thinking and spatial layout for the sustainable development of urban agglomeration in northern slope of Tianshan Mountains[J]. Arid Land Geography, 2019, 42(1): 1-11. ]
[16]	邸浩, 赵学军, 张自力. 基于EEMD-LSTM-Adaboost的商品价格预测[J]. 统计与决策, 2018, 34(13): 72-76.
	[ Di Hao, Zhao Xuejun, Zhang Zili. Commodity price forecasting based on EEMD-LSTM-Adaboost[J]. Statistics & Decision, 2018, 34(13): 72-76. ]
[17]	Wu Z, Huang N E. Ensemble empirical mode decomposition: Anoise-assisted data-analysis method[J]. Advances in Adaptive Data Analysis, 2009, 1(1): 1-41. doi: 10.1142/S1793536909000047
[18]	郑一, 孙晓峰, 陈健, 等. 基于集合经验模态的随钻脉冲信号优良降噪算法[J]. 石油勘探与开发, 2012, 39(6): 113-116. doi: 10.1016/S1876-3804(12)60022-6
	[ Zheng Yi, Sun Xiaofeng, Chen Jian, et al. Extracting pulse signals in measurement while drilling using optimum denoising methods based on the ensemble empirical mode decomposition[J]. Petroleum Exploration and Development, 2012, 39(6): 113-116. ] doi: 10.1016/S1876-3804(12)60022-6
[19]	Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. pmid: 9377276
[20]	赵驰. 深度循环神经网络在特定场景自动问答中的应用研究[D]. 武汉: 武汉理工大学, 2018.
	[ Zhao Chi. Application Research on Deep Recurrent Neural Networks in Automatic Question-Answer under Specific Occasions[D]. Wuhan: Wuhan University of Technology, 2018. ]
[21]	Graves A. Long Short-Term Memory: Supervised Sequence Labelling with Recurrent Neural Networks[M]. Berlin, Heidelberg: Springer, 2012: 1735-1780.
[22]	朱灵子. 基于统计模型的汛期降水预测研究——以义乌市为例[D]. 金华: 浙江师范大学, 2015.
	[ Zhu Lingzi. Study of Flood Season Precipitation Prediction Based on Statistical Model: Taking Yiwu City as an Example[D]. Jinhua: Zhejiang Normal University, 2015. ]
[23]	白慧, 高辉, 刘长征, 等. MODES系统对贵州月气温、降水预测初步评估[J]. 沙漠与绿洲气象, 2016, 10(5): 58-63.
	[ Bai Hui, Gao Hui, Liu Changzheng, et al. Assessment of multi-model downscaling ensemble prediction system for monthly temperature and precipitation prediction in Guizhou[J]. Desert and Oasis Meteorology, 2016, 10(5): 58-63. ]
[24]	杨莲梅, 刘晶. 新疆水汽研究若干进展[J]. 自然灾害学报, 2018, 27(2): 1-13.
	[ Yang Lianmei, Liu Jing. Some advances of water vapor research in Xinjiang[J]. Journal of Natural Disasters, 2018, 27(2): 1-13. ]
[25]	戴新刚, 汪萍, 张凯静. 近60年新疆降水趋势与波动机制分析[J]. 物理学报, 2013, 62(12): 527-537.
	[ Dai Xingang, Wang Ping, Zhang Kaijing. A study on precipitation trend and fluctuation mechanism in northwestern China over the past 60 years[J]. Acta Physica Sinica, 2013, 62(12): 527-537. ]
[26]	胡跃文, 杨小怡. 北极涛动与北大西洋涛动的低频变化特征[J]. 气象科学, 2007, 99(3): 84-90.
	[ Hu Yuewen, Yang Xiaoyi. Low-frequency variability of AO and NAO[J]. Journal of the Meteorological Sciences, 2007, 99(3): 84-90. ]
[27]	关学锋, 孙卫国, 李敏姣, 等. 1965—2012年新疆北部地区气候变化及其对北极涛动的响应[J]. 干旱区研究, 2016, 33(4): 681-689.
	[ Guan Xuefeng, Sun Weiguo, Li Minjiao, et al. Climate change in north Xinjiang and its response to Arctic Oscillation during the period of 1965-2012[J]. Arid Zone Research, 2016, 33(4): 681-689. ]
[28]	刘德钊, 周海东, 荣莉, 等. GNNM(1, 1)模型在城市用水量预测中的应用[J]. 山东建筑大学学报, 2006, 21(4): 335-337.
	[ Liu Dezhao, Zhou Haidong, Rong Li, et al. Predicting water demands of cities by GNNM (1, 1)[J]. Journal of Shandong Jianzhu University, 2006, 21(4): 335-337. ]
[29]	王婷婷, 钱晓东. 时间序列的非线性趋势预测及应用综述[J]. 计算机工程与设计, 2010, 31(7): 1545-1549.
	[ Wang Tingting, Qian Xiaodong. Survey of non-linear trend forecast and application of time series[J]. Computer Engineering and Design, 2010, 31(7): 1545-1549. ]

分量	准周期/a	方差贡献率/%
IMF₁	2.9	54.54
IMF₂	7.5	15.70
IMF₃	16.0	7.75
IMF₄	51.2	0.45
RES	无	20.15

分量	隐藏层单元数/个	训练次数/次	Dropout
IMF₁	160	30000	0.5
IMF₂	100	600	0.5
IMF₃	16	5000	0.1
IMF₄	16	1000	0.2
RES	128	10000	0.5

	2010年	2011年	2012年	2013年	2014年	2015年	2016年	2017年	2018年	2019年
绝对误差/mm	16.44	17.48	37.22	9.07	9.38	59.48	52.35	20.17	39.3	64.04
相对误差/%	5.98	7.41	17.75	3.89	4.37	20.08	17.18	9.27	17.04	30.79
距平符号	同号	同号	异号	同号	同号	同号	同号	异号	异号	异号

	配对差值					T	自由度	显著性（双尾）
	平均值	标准偏差	标准误差	差值的95%置信区间
	平均值	标准偏差	标准误差	下限	上限
预测值与观测值	3.24	27.12	4.39	-5.68	12.15	0.74	37	0.47

	2020年	2021年	2022年	2023年	2024年	2025年	2026年	2027年	2028年	2029年
降水量/mm	218.8	192.81	263.12	237.89	208.8	277.6	213.95	236.3	242.2	239.59
距平百分率/%	-5.38	-16.62	13.78	2.87	-9.71	20.04	7.48	2.18	4.74	3.61
丰枯情况	偏少	偏少	偏多	偏多	偏少	偏多	偏少	偏多	偏多	偏多