[1]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,(02):134-139.[doi:10.16357/j.cnki.issn1000-5862.2022.02.04]
　CHENG Pengyu,ZHAO Jia*,HAN Longzhe,et al.The Short-Term Temperature Prediction Based on Bidirectional Multi-Scale LSTM[J].Journal of Jiangxi Normal University:Natural Science Edition,2022,(02):134-139.[doi:10.16357/j.cnki.issn1000-5862.2022.02.04]
点击复制

双向多尺度LSTM的短时温度预测()
分享到：

《江西师范大学学报》（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2022年02期
页码:
134-139
栏目:
信息科学与技术
出版日期:
2022-03-25

文章信息/Info

Title:
The Short-Term Temperature Prediction Based on Bidirectional Multi-Scale LSTM
文章编号:
1000-5862(2022)02-0134-06
作者:
程鹏宇1; 赵 嘉1*; 韩龙哲1; 张翼英2; 武延年3
1.南昌工程学院信息工程学院,江西 南昌 330099; 2.天津科技大学人工智能学院,天津 300457; 3.深圳市国电科技通信有限公司,广东 深圳 518000
Author(s):
CHENG Pengyu1; ZHAO Jia1*; HAN Longzhe1; ZHANG Yiying2; WU Yannian3
1.School of Information Engineering,Nanchang Institute of Technology,Nanchang Jiangxi 330099,China; 2.College of Artificial Intelligence,Tianjin University of Science & Technology,Tianjin 300457,China; 3.Shenzhen Guodian Technology Communication Corporation Limited,Shenzhen Guangdong 518000,China
关键词:
长短期记忆网络;  温度预测;  多尺度;  反向特征;  跳跃连接;  特征融合
Keywords:
Long Short-Term Memory(LSTM);  temperature prediction;  multi-scale;  reverse feature;  skip connection;  feature fusion
分类号:
TP 183
DOI:
10.16357/j.cnki.issn1000-5862.2022.02.04
文献标志码:
A
摘要:
针对长短期记忆网络(long short-term memory,LSTM)无法有效提取温度数据的多尺度特征和反向特征的问题,该文提出了一种双向多尺度跳跃LSTM(bidirectional multi-scale skip long short-term memory,BMS-LSTM)的短时温度预测模型.该模型以LSTM为核心单元,采用双向深层网络结构提取反向特征; 根据温度数据日的周期性设置跳跃连接数提取多尺度特征,解决了指数增长的跳跃连接数后期跳跃尺度过大的问题; 最后使用全连接层进行特征融合预测.实验结果表明:BMS-LSTM成功提取了温度数据的多尺度特征和反向特征,预测均值误差仅为3.890,优于对比模型,是一种有效的短时温度预测模型.
Abstract:
Because at the LSTM(Long Short-Term Memory)cannot effectively extract the multi-scale features and inverse features of temperature data,the BMS-LSTM(Bidirectional Multi-scale Skip Long Short-Term Memory)short-term temperature prediction model is proposed.The model uses LSTM as the core unit,and uses a two-way deep network structure to extract reverse features.The multi-scale feature is extracted according to the daily cycle setting of the number of jump connections in the temperature data,which solves the problem of the jump scale being too large in the later stage of the exponentially increasing number of jump connections.Finally,the fully connected layer is used for feature fusion prediction.The experiment's results show that BMS-LSTM has successfully extracted the multi-scale features and reverse features of temperature data,and the average prediction error is only 3.890,which is better than the comparison model and is an effective short-term temperature prediction model.

参考文献/References:

[1] 陈广仁.气候变化关乎人类未来[J].科技导报,2011,29(34):8.
[2] 王灏晨,路凤,武继磊,等.中国气候变化对人口健康影响研究评述[J].科技导报,2014,32(28/29):109-116.
[3] 夏亮.气温的统计分析及短期预测[D].武汉:华中科技大学,2010.
[4] 杨成荫,赵苏璇,程立国,等.短期区域气候预测系统及回报试验[J].解放军理工大学学报(自然科学版),2016,17(3):289-295.
[5] 王定成,曹智丽,陈北京,等.日气温多元时间序列局部支持向量回归预测[J].系统仿真学报,2016,28(3):654-660.
[6] HUANG Chiyo,LIU Yuwei,TZENG Weichang,et al.Short term wind speed predictions by using the grey prediction model based forecast method[EB/OL].[2021-03-16].https://ieeexplore.ieee.org/abstract/document/5754856.
[7] 刘丽丹.基于概率图模型的天气预测研究[J].计算机技术与发展,2019,29(7):103-107.
[8] 张曦,李璠,付雪峰,等.随机学习萤火虫算法优化的模糊软子空间聚类算法[J].江西师范大学学报(自然科学版),2021,45(2):137-144.
[9] 陈开阳,徐凡,王明文.基于知识图谱和图像描述的虚假新闻检测研究[J].江西师范大学学报(自然科学版),2021,45(4):398-402.
[10] 赵嘉,谢智峰,吕莉,等.深度学习萤火虫算法[J].电子学报,2018,46(11):2633-2641.
[11] 陶晔,杜景林.基于随机森林的长短期记忆网络气温预测[J].计算机工程与设计,2019,40(3):737-743.
[12] MIGUEZ-MACHO G,STENCHIKOV G L,ROBOCK A.Regional climate simulations over North America:interaction of local processes with improved large-scale flow[J].Journal of Climate,2005,18(8):1227-1246.
[13] 李芬,杨程,赵晋斌,等.基于天气类型聚类和LSTM的PM2.5短期预测模型[J].水电能源科学,2021,39(3):199-202,151.
[14] ZAREMBA W,SUTSKEVER I,VINYALS O.Recurrent neural network regularization[EB/OL].[2021-06-16].http://www.arxiv.org/pdf/1409.2329.pdf.
[15] HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[16] KOUTNIK J,GREFF K,GOMEZ F,et al.A clockwork RNN[EB/OL].[2021-06-16].http://www.arxiv.org/pdf/1402.3511v1.pdf.
[17] CHUNG J Y,ANH S J,BENGIO Y S.Hierarchical multiscale recurrent neural networks[EB/OL].[2021-07-16].https://arxiv.org/pdf/1609.01704.pdf.
[18] CHANG Shiyu,ZHANG Yang,HAN Wei,et al.Dilated recurrent neural networks[EB/OL].[2021-06-16].https://arxiv.org/abs/1710.02224.
[19] 肖勇,郑楷洪,郑镇境,等.基于多尺度跳跃深度长短期记忆网络的短期多变量负荷预测[J].计算机应用,2021,41(1):231-236.
[20] SCHUSTER M,PALIWAL K K.Bidirectional recurrent neural networks[J].IEEE Transactions on Signal Processing,1997,45(11):2673-2681.
[21] GRAVES A,SCHMIDHUBER J.Framewise phoneme classification with bidirectional LSTM and other neural network architectures[J].Neural Networks,2005,18(5/6):602-610.
[22] 曾安,聂文俊.基于深度双向LSTM的股票推荐系统[J].计算机科学,2019,46(10):84-89.
[23] 杨龙,吴红斌,丁明,等.新能源电网中考虑特征选择的Bi-LSTM网络短期负荷预测[J].电力系统自动化,2021,45(3):166-173.
[24] TABRIZI S E,XIAO Kai,GRIENSVEN J V,et al.Hourly road pavement surface temperature forecasting using deep learning models[J].Journal of Hydrology,2021,603(1):126877.

备注/Memo

备注/Memo:
收稿日期:2021-09-25
基金项目:国家自然科学基金(52069014,61962036)和江西省杰出青年基金(2018ACB21029)资助项目.
通信作者:赵 嘉(1981—),男,安徽桐城人,教授,博士,主要从事大数据分析、人工智能理论、深度学习研究.E-mail:zhaojia925@163.com
更新日期/Last Update: 2022-03-25