2.765

2022影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于周期性建模的时间序列预测方法及电价预测研究

徐任超 阎威武 王国良 杨健程 张曦

徐任超, 阎威武, 王国良, 杨健程, 张曦. 基于周期性建模的时间序列预测方法及电价预测研究. 自动化学报, 2020, 46(6): 1136−1144 doi: 10.16383/j.aas.c180712
引用本文: 徐任超, 阎威武, 王国良, 杨健程, 张曦. 基于周期性建模的时间序列预测方法及电价预测研究. 自动化学报, 2020, 46(6): 1136−1144 doi: 10.16383/j.aas.c180712
Xu Ren-Chao, Yan Wei-Wu, Wang Guo-Liang, Yang Jian-Cheng, Zhang Xi. Time series forecasting based on seasonality modeling and its application to electricity price forecasting. Acta Automatica Sinica, 2020, 46(6): 1136−1144 doi: 10.16383/j.aas.c180712
Citation: Xu Ren-Chao, Yan Wei-Wu, Wang Guo-Liang, Yang Jian-Cheng, Zhang Xi. Time series forecasting based on seasonality modeling and its application to electricity price forecasting. Acta Automatica Sinica, 2020, 46(6): 1136−1144 doi: 10.16383/j.aas.c180712

基于周期性建模的时间序列预测方法及电价预测研究

doi: 10.16383/j.aas.c180712
基金项目: 国家重点研究发展计划基金(2019YFB1705702), 国家自然科学基金(60974119, 61533013)资助
详细信息
    作者简介:

    徐任超:上海交通大学工学硕士. 主要研究方向为深度学习, 时间序列预测和图像处理. E-mail: xurenchao@yeah.net

    阎威武:工学博士, 上海交通大学副教授. 主要研究方向为机器学习, 大数据, 图像处理和智能制造. 本文通信作者. E-mail: aas_yanwwsjtu@sjtu.edu.cn

    王国良:工学博士, 上海工程技术大学讲师. 主要研究方向为过程建模与控制, 火电机组先进控制. E-mail: glwang@sues.edu.cn

    杨健程:上海交通大学自动化系工学硕士. 主要研究方向为深度学习, 医学图像处理和3D视觉. E-mail: jekyll4168@sjtu.edu.cn

    张曦:工学博士, 教授级高级工程师. 主要研究方向为统计过程控制, 火电机组先进控制. E-mail: zhangxi2@csg.cn

Time Series Forecasting Based on Seasonality Modeling and Its Application to Electricity Price Forecasting

Funds: Supported by National Key Research and Development Program of China (2019YFB1705702) and National Natural Science Foundation of China (60974119, 61533013)
  • 摘要: 时间序列数据广泛存在于人类的生产生活中, 通常具有复杂的非线性动态和一定的周期性. 与传统的时间序列分析方法相比, 基于深度学习的方法更能捕捉数据的深层特性, 对具有复杂非线性的时间序列有较好的建模效果. 为了在神经网络中显式地建模时间序列数据的周期性和趋势性, 本文在循环神经网络的基础上引入了周期损失和趋势损失, 建立了基于周期性建模和多任务学习的时间序列预测模型. 将模型应用到欧洲能源交易所法国市场的能源市场价格预测中, 结果表明周期损失和趋势损失能够提高神经网络的泛化能力, 并提高预测时间序列趋势的精度.
  • 图  1  一个展开的基础循环神经网络

    Fig.  1  The unfold structure of RNN

    图  2  LSTM的内部结构示意图

    Fig.  2  The schematic diagram of LSTM

    图  3  GRU的内部结构示意图

    Fig.  3  The schematic diagram of GRU

    图  4  能源市场价格的高波动性

    Fig.  4  High volatility of electricity price

    图  5  能源市场价格的日级别周期性

    Fig.  5  The daily seasonality of electricity price

    图  6  能源市场价格的周级别周期性

    Fig.  6  The weekly seasonality of electricity price

    图  7  加入了周期损失和趋势损失的GRU模型的预测结果

    Fig.  7  The results of electricity price forecasting with seasonal loss and trend loss

    图  8  GRU在有无周期损失时的隐藏状态均值的比较

    Fig.  8  Hidden states of GRU with and without the seasonality

    表  1  循环神经网络的超参数设置

    Table  1  The hyperparameters of RNN

    超参数具体取值
    隐层大小64
    优化器RMSProp, 配合梯度裁剪
    初始学习率0.001
    批大小64
    训练轮数12
    延迟窗宽14
    下载: 导出CSV

    表  2  周期损失和趋势损失的权重范围

    Table  2  Weights range of seasonal loss and trend loss

    权重取值范围
    $\lambda_S $0.05~0.15
    $\lambda_T^{\rm{MEAN}} $0
    $\lambda_T^{\rm{MAX}} $0.05~0.1
    $\lambda_T^{\rm{MIN}} $0.05~0.1
    $\lambda_T^{\rm{VAR}} $0
    下载: 导出CSV

    表  3  各种方法的能源价格预测效果对比

    Table  3  The result comparisons of different methods for electricity price forecasting

    模型RMSEMAE${MAE}^{\rm{MAX}}$${MAE}^{\rm{MIN}}$
    ARIMA6.414.775.154.82
    SVR4.913.714.273.34
    KRR5.143.753.813.78
    RNN5.09±0.243.75±0.193.72±0.283.78±0.19
    LSTM4.90±0.183.65±0.173.65±0.423.61±0.26
    GRU4.83±0.193.54±0.063.64±0.313.56±0.26
    GRU, $\lambda_S$ = 0.1, $\lambda_T^{\rm{MAX}}$ = 0, $\lambda_T^{\rm{MIN}}$ = 04.71±0.163.49±0.133.53±0.283.53±0.15
    GRU, $\lambda_S$ = 0.05, $\lambda_T^{\rm{MAX}}$ = 0, $\lambda_T^{\rm{MIN}}$ = 04.74±0.113.45±0.183.53±0.233.48±0.26
    GRU, $\lambda_S$ = 0 , $\lambda_T^{\rm{MAX}}$ = 0.1 , $\lambda_T^{\rm{MIN}}$ = 0.14.85±0.163.57±0.203.41±0.263.41±0.18
    GRU, $\lambda_S$ = 0 , $\lambda_T^{\rm{MAX}}$ = 0.05 , $\lambda_T^{\rm{MIN}}$ = 0.054.83±0.113.54±0.083.39±0.183.42±0.15
    GRU, $\lambda_S$ = 0.1, $\lambda_T^{\rm{MAX}}$ = 0.1 , $\lambda_T^{\rm{MIN}}$ = 0.14.68±0.083.45±0.033.35±0.133.33±0.12
    GRU, $\lambda_S$ = 0.05, $\lambda_T^{\rm{MAX}}$ = 0.05 , $\lambda_T^{\rm{MIN}}$ = 0.054.60±0.153.34±0.123.38±0.133.27±0.11
    下载: 导出CSV
  • [1] Budish E, Cramton P, Shim J. The high-frequency trading arms race: Frequent batch auctions as a market design response. The Quarterly Journal of Economics, 2015, 130(4): 1547−1621
    [2] He J, Li X, Liao L J, Song D D, Cheung W K. Inferring a personalized next point-of-interest recommendation model with latent behavior patterns. In: Proceedings of the 30th AAAI Conference on Artificial Intelligence. AAAI Press, 2016. 137−143
    [3] Ren Y F, Wu Y. Convolutional deep belief networks for feature extraction of EEG signal. In: Proceedings of the 2014 International Joint Conference on Neural Networks. Beijing, China: IEEE, 2014. 2850−2853
    [4] 张熙来, 赵俭辉, 蔡波. 针对PM2.5单时间序列数据的动态调整预测模型. 自动化学报, 2018, 44(10): 1790−1798

    Zhang Xi-Lai, Zhao Jian-Hui, Cai Bo. Prediction model with dynamic adjustment for single time series of PM2.5. Acta Automatica Sinica, 2018, 44(10): 1790−1798
    [5] Keogh E, Chu S, Hart D, Pazzani M. Segmenting time series: A survey and novel approach. Data mining in time series databases. 2004.
    [6] Box G E P, Jenkins G M. Time series analysis: Forecasting and control. Journal of the Operational Research Society, 1971, 37(2): 238−242
    [7] Engle R F. Autoregressive conditional heteroscedasticity with estimates of the variance of united kingdom inflation. Econometrica, 1982, 50(4): 987−1007 doi: 10.2307/1912773
    [8] Bollerslevb T. Generalized autoregressive conditional heteroscedasticity. Eeri Research Paper, 1986, 31(3): 307−327
    [9] Lu C J, Lee T S, Chiu C C. Financial time series forecasting using independent component analysis and support vector regression. Decision Support Systems, 2009, 47(2): 115−125 doi: 10.1016/j.dss.2009.02.001
    [10] Engel Y, Mannor S, Meir R. The kernel recursive least-squares algorithm. IEEE Transactions on Signal Processing, 2004, 52(8): 2275−2285 doi: 10.1109/TSP.2004.830985
    [11] Yang J C, Yan W W, Xu R C, Zhang X. A novel online kernel ridge to forecast next-day electricity price. International Journal of System Control and Information Processing, 2018, 2(4): 317−331 doi: 10.1504/IJSCIP.2018.097198
    [12] Zhang J, Man K F. Time series prediction using RNN in multi-dimension embedding phase space. In: Proceedings of the 2002 IEEE International Conference on Systems, Man, and Cybernetics. IEEE, 2002. 1868−1873.
    [13] Graves A, Mohamed A, Hinton G. Speech recognition with deep recurrent neural networks. In: Proceedings of the 2013 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP). IEEE, 2013. 6645−6649
    [14] 伦淑娴, 林健, 姚显双. 基于小世界回声状态网的时间序列预测. 自动化学报, 2015, 41(9): 1669−1679

    Lun Shu-Xian, Lin Jian, Yao Xian-Shuang. Time series prediction with an improved echo state network using small world network. Acta Automatica Sinica, 2015, 41(9): 1669−1679
    [15] Besse P C, Cardot H, Stephenson D B. Autoregressive forecasting of some functional climatic variations. Scandinavian Journal of Statistics, 2000, 27(4): 673−687 doi: 10.1111/1467-9469.00215
    [16] Aggarwal S K, Saini L M, Kumar A. Electricity price forecasting in deregulated markets: A review and evaluation. International Journal of Electrical Power and Energy Systems, 2009, 31(1): 13−22
    [17] Anbazhagan S, Kumarappan N. Day-ahead deregulated electricity market price forecasting using recurrent neural network. IEEE Systems Journal, 2013, 7(4): 866−87 doi: 10.1109/JSYST.2012.2225733
    [18] Clements A E, Hurn A S, Li Z. Forecasting day-ahead electricity load using a multiple equation time series approach. European Journal of Operational Research, 2016, 251(2): 522−530 doi: 10.1016/j.ejor.2015.12.030
    [19] Anbazhagan S, Kumarappan N. Day-ahead deregulated electricity market price forecasting using neural network input featured by DCT. Energy Conversion and Management, 2014, 78: 711−719 doi: 10.1016/j.enconman.2013.11.031
    [20] Rafiei M, Niknam T, Khooban M H. Probabilistic forecasting of hourly electricity price by generalization of ELM for usage in improved wavelet neural network. IEEE Transactions on Industrial Informatics, 2017, 13(1): 71−79 doi: 10.1109/TII.2016.2585378
    [21] Pascanu R, Mikolov T, Bengio Y. On the difficulty of training recurrent neural networks. In: Proceedings of the 2013 International Conference on Machine Learning. 2013.
    [22] Hochreiter S, Schmidhuber J. Long short-term memory. Neural computation, 1997, 9(8): 1735−1780 doi: 10.1162/neco.1997.9.8.1735
    [23] Greff K, Srivastava R K, Koutnik J, Steunebrink B R, Schmidhuber J. LSTM: A search space odyssey. IEEE Transactions on Neural Networks and Learning Systems, 2017, 28(10): 2222−2232 doi: 10.1109/TNNLS.2016.2582924
    [24] Cho K, Van Merriënboer B, Gulcehre C, Bahdanau D, Bougares F, Schwenk H, Bengio Y. Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv: 1406.1078v1, 2014.
    [25] Caruana R. Multitask learning. Machine Learning, 1997, 28(1): 41−75 doi: 10.1023/A:1007379606734
  • 加载中
图(8) / 表(3)
计量
  • 文章访问数:  1346
  • HTML全文浏览量:  207
  • PDF下载量:  331
  • 被引次数: 0
出版历程
  • 收稿日期:  2018-10-31
  • 录用日期:  2019-06-02
  • 网络出版日期:  2020-07-10
  • 刊出日期:  2020-07-10

目录

    /

    返回文章
    返回