基于改进双阶段注意力机制的降水智能预报
作者:
中图分类号:

TP399

基金项目:

国家自然科学基金(61773220);国家重点研发计划(2016YFC0203301);江苏省自然科学基金(BK20150523);国家自然科学基金联合重点项目(U20B2061)


Intelligent precipitation forecast based on improved dual-stage attention mechanism
Author:
  • 摘要
  • | |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • | |
  • 文章评论
    摘要:

    为提高现有时间序列算法降水预报的准确率, 本文提出一种基于改进双阶段注意力机制的时间序列降水预报模型(DeepAMogLSTM).该算法分为两部分, 在输入注意力机制中, 使用三层注意力机制对输入序列进行多重关注, 选择稳定的输入特征; 在时间注意力中, 通过选择与目标值最相关的隐层状态, 捕获时间序列的长期相关性.算法同时引入形变长短时记忆网络(Mogrifier LSTM), 增强模型特征表示能力.模型使用2016—2019年预处理的自动站点特征数据和欧洲中期天气预报中心(ECMWF)气象场模式资料进行集成预报, 并利用同期实况观测资料进行模式预报订正.实验结果表明: 该模型在时效为2 h的降水预报中, 各项数值评价指标均有改善, 其中均方根误差为1.877 mm, 平均绝对误差为0.727 mm, 拟合优度(R2)为0.783;同时与其他模型预报订正效果相比, 该模型较好地拟合了实际降水空间分布.

    Abstract:

    In order to improve the existing time series algorithms for precipitation forecasting, this paper proposes a time series precipitation forecasting model (DeepAMogLSTM) based on improved dual-stage attention mechanism.The algorithm can be divided into two parts.In the input attention stage, a three-layer attention mechanism is designed to pay multiple attention to the input sequence; while in the time attention stage, the hidden state most relevant to the target value is selected to calculate the long-term correlation of the time sequence.In this manner, input features can be stably selected and input into the prediction structure.The algorithm also introduces Mogrifier LSTM (Long Short-Term Memory) to enhance the feature representation ability.The model uses preprocessed automatic station data from 2016 to 2019 and ECMWF weather field model data for integrated forecast, and corrects the model forecasts using observation data of the same period.The experimental results show that the evaluation indexes of the model are improved in the 2-hour precipitation nowcasting, in which the maximum square root error is 1.877 mm, the maximum average absolute error is 0.727 mm, and the goodness of fit (R2) is 0.783.At the same time, the modeled precipitation fits actual precipitation in spatial distribution, which is better than the correction effect of other models.

    参考文献
    [1] Buisán S T, Smith C D, Ross A. The potential for uncertainty in numerical weather prediction model verification when using solid precipitation observations[J]. Atmospheric Science Letters, 2020, 21(7). DOI: 10.1002/asl.976
    [2] 智协飞, 赵忱. 基于集合成员订正的强降水多模式集成预报[J]. 应用气象学报, 2020, 31(3): 303-314 ZHI Xiefei, ZHAO Chen. Heavy precipitation forecasts based on multi-model ensemble members[J]. Journal of Applied Meteorological Science, 2020, 31(3): 303-314
    [3] 李启芬, 刘婷婷, 陈海山, 等. 基于土壤湿度和年际增量方法的中国夏季气温预测试验[J]. 气象科学, 2016, 36(5): 629-638 LI Qifen, LIU Tingting, CHEN Haishan, et al. Prediction of summer temperature in China based on soil moisture and interannual increment approach[J]. Journal of the Meteorological Sciences, 2016, 36(5): 629-638
    [4] 任海青. 基于灰色系统的年降雨量预测模型探析[J]. 水资源开发与管理, 2017(3): 61-63 REN Haiqing. Discussion of annual rainfall forecast model based on grey system[J]. Water Resources Development and Management, 2017(3): 61-63
    [5] 程敏, 张耀文, 姜纪沂, 等. 基于时间序列模型的降雨量预测分析[J]. 水科学与工程技术, 2019 (1): 1-5 CHENG Min, ZHANG Yaowen, JIANG Jiyi, et al. Rainfall prediction based on time series model[J]. Water Sciences and Engineering Technology, 2019(1): 1-5
    [6] 张帅, 魏正英, 张育斌. 递归神经网络在降雨量预测中的应用研究[J]. 节水灌溉, 2017 (5): 63-66, 71 ZHANG Shuai, WEI Zhengying, ZHANG Yubin. Forecasting rainfall with recurrent neural network[J]. Water Saving Irrigation, 2017 (5): 63-66, 71
    [7] Qin Y, Song D J, Chen H F, et al. A dual-stage attention-based recurrent neural network for time series prediction[C]//Proceedings of the 26th International Joint Conference on Artificial Intelligence, 2017: 2627-2633
    [8] Melis G, Ko čiský T, Blunsom P. Mogrifier LSTM[J]. arXiv e-print, 2019, arXiv: 1909.01792
    [9] Le K T, Chaux C, Richard F J P, et al. An adapted linear discriminant analysis with variable selection for the classification in high-dimension, and an application to medical data[J]. Computational Statistics and Data Analysis, 2020, 152(1): 107031
    [10] Cai L Q, Maiti T. Variable selection and estimation for high-dimensional spatial autoregressive models[J]. Scandinavian Journal of Statistics, 2020, 47(2): 587-607
    [11] Dai Y L, Lu Z Y, Zhang H D, et al. A correction method of environmental meteorological model based on long-short-term memory neural network[J]. Earth and Space Science, 2019, 6(4): 2214-2226
    [12] 尹珩, 张珂, 孙辰, 等. 武汉城区臭氧时空分布及其与气象因子相关性研究[J]. 环境与可持续发展, 2017, 42(6): 150-152 YIN Heng, ZHANG Ke, SUN Chen, et al. Study on the spatial-temporal distribution of ozone and its correlation with meteorological factors in urban district of Wuhan[J]. Environment and Sustainable Development, 2017, 42(6): 150-152
    [13] Tian B Q, Fan K. Climate prediction of summer extreme precipitation frequency in the Yangtze River valley based on sea surface temperature in the southern Indian Ocean and ice concentration in the Beaufort Sea[J]. International Journal of Climatology, 2020, 40(9): 4117-4130
    [14] Bahdanau D, Cho K, Bengio Y. Neural machine translation by jointly learning to align and translate[J]. Computer Science, 2014: 1-15
    [15] Vapnik V, Golowich S E, Smola A. Support vector method for function approximation, regression estimation, and signal processing[J]. Advances in Neural Information Processing Systems, 2008, 9: 281-287
    [16] Greff K, Srivastava R K, Koutnik J, et al. LSTM: a search space odyssey[J]. IEEE Transactions on Neural Networks & Learning Systems, 2016, 28(10): 2222-2232
    [17] Lai G K, Chang W C, Yang Y M, et al. Modeling long- and short-term temporal patterns with deep neural networks[C]//The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, 2018: 95-104
    引证文献
    网友评论
    网友评论
    分享到微博
    发 布
引用本文

戈苗苗,陆振宇,梁邵阳,夏英茹.基于改进双阶段注意力机制的降水智能预报[J].南京信息工程大学学报(自然科学版),2021,13(6):744-752
GE Miaomiao, LU Zhenyu, LIANG Shaoyang, XIA Yingru. Intelligent precipitation forecast based on improved dual-stage attention mechanism[J]. Journal of Nanjing University of Information Science & Technology, 2021,13(6):744-752

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-03-27
  • 在线发布日期: 2022-01-21

地址:江苏省南京市宁六路219号    邮编:210044

联系电话:025-58731025    E-mail:nxdxb@nuist.edu.cn

南京信息工程大学学报 ® 2025 版权所有  技术支持:北京勤云科技发展有限公司