Accurate power load forecasting is crucial to the safe and economic operation of modern power systems. Power load forecasting can be expressed as a multivariable time series forecasting problem with certain potential spatial dependence. However, most existing power load forecasting work fails to explore this spatial dependency relationship. Based on this, this paper proposes a short-term power load forecasting method based on the spatiotemporal graph attention network. A spatiotemporal graph-based attention network module is proposed, which uses a graph attention layer to adaptively capture potential spatial dependencies between users. At the same time, a gated convolutional attention layer is used to adaptively fit the electricity consumption of each user in the time dimension to improve the prediction accuracy of the network. Actual data experiments show that the overall prediction accuracy of the model proposed is significantly improved, especially in alleviating the problem of deteriorating long-range prediction accuracy to a certain extent, verifying the effectiveness and feasibility of the proposed method.
根据小世界网络生成的过程,提出基于小世界网络的邻接矩阵生成算法,如算法1所示.具体步骤如下:首先,根据多变量时序数据 X 计算任意两个变量间的相似性 S = X · XT,并进行归一化,得到相似性矩阵 S .随后,将邻近数k设为0,得到一个全断开的图结构;最后,逐渐增大邻近数k至N,在此过程中,图结构会逐渐变得稠密,此时ACC逐渐变大,h逐渐变小,分别计算对应的h( A )和ACC( A ),最终形成如图2右侧所示的h( A )和ACC( A )曲线。选择h( A )和ACC( A )曲线的交(ratio, k)作为邻近数k,根据选择到的邻近数k,进行图结构 A 的构建,具体定义公式如下:
其中,Neigh k (vi )为距离节点vi 最近的k个节点的集合.最后针对每一个节点,随机选择两个未与其相连接的节点,以概率p进行连接.
HSIAOY H. Household electricity demand forecast based on context information and user daily schedule analysis from meter data [J]. IEEE Transactions on Industrial Informatics, 2015, 11(1) : 33-43.
[2]
CHENY, XUP, CHUY, et al. Short-term electrical load forecasting using the Support Vector Regression (SVR) model to calculate the demand response baseline for office buildings [J]. Applied Energy, 2017, 195: 659-670.
[3]
CAIM, PIPATTANASOMPORNM, RAHMANS. Day-ahead building-level load forecasts using deep learning vs. traditional time-series techniques [J]. Applied Energy, 2019, 236: 1078-1088.
[4]
JURADOS, NEBOTÀ, MUGICAF, et al. Hybrid methodologies for electricity load forecasting: Entropy-based feature selection with machine learning and soft computing techniques [J]. Energy, 2015, 86: 276-291.
[5]
KUSTERC, REZGUIY, MOURSHEDM. Electrical load forecasting models: A critical systematic review [J]. Sustainable Cities and Society, 2017, 35: 257-270.
[6]
ZHANGD, AHMADT, SHAHW A J I A. Efficient energy planning with decomposition-based evolutionary neural networks[J]. 2020, 8: 134880-134897.
[7]
ZHANGN, LIZ, ZOUX, et al. Comparison of three short-term load forecast models in Southern California [J]. Energy, 2019, 189: 116358.
[8]
LEEY-S, TONGL-I. Forecasting time series using a methodology based on autoregressive integrated moving average and genetic programming [J]. Knowledge-Based Systems, 2011, 24(1) : 66-72.
YEZ B, ZHOUB X, LINN,et al. Mid-long term load forecasting based on equal dimension new information exponential smoothing model[J]. Power System Protection and Control,2012, 40(18) : 47-51.(in Chinese)
ZHAOC, DAIK C. Air conditioning load forecasting based on adaptive weighted least squares support vector machine[J]. Journal of Chongqing University,2016, 39(1): 55-64.(in Chinese)
[13]
CORIZZOR, CECIM, FANAEE-TH, et al. Multi-aspect renewable energy forecasting [J]. Information Sciences, 2021, 546: 701-722.
[14]
TASCIKARAOGLUA. Evaluation of spatio-temporal forecasting methods in various smart city applications [J]. Renewable and Sustainable Energy Reviews, 2018, 82: 424-435.
[15]
BOD Y, WANGX, SHIC, et al. Beyond low-frequency information in graph convolutional networks[DB/OL].
[16]
SILVAT C. Machine learning in complex networks: modeling, analysis, and applications[D].Beijing:Institute of Computing, Chinese Academy of Social Sciences,2012.
[17]
MAIERM, LUXBURGU V, HEINM, et al. Influence of graph construction on graph-based clustering measures[J]. NIPS’08:Proceedings of the 21st International Conference on Neural Information Processing Systems, 2008, 12:1025-1032.
[18]
MARIANIS, RENDUQ, URBANIM, et al. Causal dilated convolutional neural networks for automatic inspection of ultrasonic signals in non-destructive evaluation and structural health monitoring [J]. Mechanical Systems and Signal Processing, 2021, 157: 107748.
[19]
DAUPHINY N, FANA, AULIM, et al. Language modeling with gated convolutional networks[C]//Proceedings of the 34th International Conference on Machine Learning. Sydney, NSW, Australia, 2017.
[20]
HEK, ZHANGX, RENS, et al. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA, 2016.
[21]
HUJ, SHENL, SUNG. Squeeze-and-excitation networks[C]//Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA, 2018.
[22]
DEFFERRARDM, BRESSONX, VANDERGHEYNSTP. Convolutional neural networks on graphs with fast localized spectral filtering[C]//Proceedings of the 30th International Conference on Neural Information Processing Systems. Barcelona, Spain, 2016.
[23]
WUZ, PANS, LONGG, et al. Connecting the Dots: multivariate time series forecasting with graph neural networks[C]// Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York, NY, USA, 2020.
[24]
BROWNR, BRYANTP, ABARBANELH D I. Computing the Lyapunov spectrum of a dynamical system from an observed time series [J]. Physical Review A, 1991, 43(6) : 2787-2806.
[25]
SHIHS, SUNF, LEEH. Temporal pattern attention for multivariate time series forecasting [J]. Machine Learning, 2019, 108(8-9) : 1421-1441.
[26]
LAIG, CHANGW C, YANGY, 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. New York, NY, USA, 2018.
[27]
ZHANGX, LEIY, CHENH, et al. Multivariate time-series modeling for forecasting sintering temperature in rotary kilns using DCGNet [J]. IEEE Transactions on Industrial Informatics, 2021, 17(7) : 4635-4645.
[28]
WUZ, PANS, LONGG, et al. Graph wavenet for deep spatial-temporal graph modeling[C]//Proceedings of the 28th International Joint Conference on Artificial Intelligence. Macao, China, 2019.