基于双图结构的时空图卷积网络短期风电功率预测模型Symbol`@@

doi:10.6040/j.issn.1672-3961.0.2025.030

摘要/Abstract

摘要： 针对传统风电功率预测方法忽略时空特征交互的问题,提出一种融合双图结构与注意力机制的时空图卷积网络模型(spatial-temporal graph convolutional network with attention, STGCN-A)。基于最大信息系数构建相关性矩阵,形成基于统计相关性的空间图,结合欧氏距离构建地理邻近性空间图,实现风电机组间的双图建模;采用图卷积网络(graph convolutional network, GCN)提取空间特征,结合门控循环单元(gated recurrent unit, GRU)深度挖掘时间依赖关系,引入注意力机制(attention mechanism, AM)对时间步进行动态加权,以增强时空特征中关键信息的表示能力。在实际风电数据集中开展对比试验,结果表明,该模型在均方根误差E_RMS、平均绝对误差E_MA和决定系数R²评价指标上优于传统方法,表现出较高的预测精度,具有较好的实际应用潜力。

关键词: 风电功率, 图卷积网络, 门控循环单元, 注意力机制, 最大信息系数

Abstract: To address the limitations of traditional wind power prediction methods that ignored the interaction of spatial-temporal features, a spatial-temporal graph convolutional network with attention(STGCN-A)was proposed. A correlation matrix was constructed by the maximal information coefficient to form a statistical correlation-based spatial graph, while an Euclidean distance-based geographic proximity spatial graph was built to achieve dual-graph modeling among wind turbines. Spatial features were extracted through a graph convolutional network(GCN), and temporal dependencies were captured by a gated recurrent unit(GRU). An attention mechanism(AM)was introduced to dynamically weight different time steps, enhancing the representation of critical information in spatial-temporal features. Comparative experiments on real wind power datasets demonstrated that the proposed model outperformed traditional methods in terms of root mean square error(E_RMS), mean absolute error(E_MA), and coefficient of determination(R²). The results indicated higher prediction accuracy and strong potential for practical applications.

Key words: wind power, graph convolutional network, gated recurrent unit, attention mechanism, maximal information coefficient

中图分类号:

TM614

郑哲明,孔玲玲,何印. 基于双图结构的时空图卷积网络短期风电功率预测模型Symbol`@@[J]. 山东大学学报 (工学版), 2026, 56(2): 130-138.

ZHENG Zheming, KONG Lingling, HE Yin. Short-term wind power prediction model based on spatial-temporal graph convolutional network with dual-graph structure[J]. Journal of Shandong University(Engineering Science), 2026, 56(2): 130-138.

参考文献

[1] 陶冶. 供需两侧齐发力, 新能源产业成为培育新质生产力的重要引擎: 2023年我国新能源发展形势分析与2024年展望[J]. 中国能源, 2024, 46(增刊1): 11-18. TAO Ye. The joint efforts of supply and demand have made the new energy industry an important engine for cultivating the new quality productive forces development in China: analysis of China's new energy development situation in 2023 and development prospects in 2024[J]. Energy of China, 2024, 46(Suppl.1): 11-18.
[2] 曹胡辉, 裘智峰, 向劲勇, 等. 考虑中长期交易与短期调度衔接的风电消纳模型[J]. 电网技术, 2020, 44(11): 4200-4209. CAO Huhui, QIU Zhifeng, XIANG Jinyong, et al. Wind power accommodation model considering the link of medium and long-term transactions with short-term dispatch[J]. Power System Technology, 2020, 44(11): 4200-4209.
[3] WAN C, WANG J H, LIN J, et al. Nonparametric prediction intervals of wind power via linear programming[J]. IEEE Transactions on Power Systems, 2018, 33(1): 1074-1076.
[4] 武新章,梁祥宇,朱虹谕,等. 基于CEEMDAN-GRA-PCC-ATCN的短期风电功率预测[J]. 山东大学学报(工学版), 2022, 52(6): 146-156. WU Xinzhang, LIANG Xiangyu, ZHU Hongyu, et al. Short-term wind power prediction based on CEEMDAN-GRA-PCC-ATCN[J]. Journal of Shandong University(Engineering Science), 2022, 52(6): 146-156.
[5] ZHANG J H, YAN J, INFIELD D, et al. Short-term forecasting and uncertainty analysis of wind turbine power based on long short-term memory network and Gaussian mixture model[J]. Applied Energy, 2019, 241: 229-244.
[6] SHAHID F, ZAMEER A, MEHMOOD A, et al. A novel wavenets long short term memory paradigm for wind power prediction[J]. Applied Energy, 2020, 269: 115098.
[7] 薛禹胜, 陈宁, 王树民, 等. 关于利用空间相关性预测风速的评述[J]. 电力系统自动化, 2017, 41(10): 161-169. XUE Yusheng, CHEN Ning, WANG Shumin, et al. Review on wind speed prediction based on spatial correlation[J]. Automation of Electric Power Systems, 2017, 41(10): 161-169.
[8] 刘金朋, 邓嘉明, 高鹏宇, 等. 基于VMD-IMPA-SVM的超短期风电功率预测[J]. 智慧电力, 2024, 52(7): 24-32. LIU Jinpeng, DENG Jiaming, GAO Pengyu, et al. Ultra short term wind power prediction based on VMD-IMPA-SVM[J]. Smart Power, 2024, 52(7): 24-32.
[9] 宋立业, 鞠亚东, 张鑫. 基于改进MFO优化Attention-LSTM的超短期风电功率预测[J]. 电气工程学报, 2023, 18(3): 359-366. SONG Liye, JU Yadong, ZHANG Xin. Ultra-short-term wind power prediction based on improved MFO optimized Attention-LSTM[J]. Journal of Electrical Engineering, 2023, 18(3): 359-366.
[10] JIANG G Q, XIE P, HE H B, et al. Wind turbine fault detection using a denoising autoencoder with temporal information[J]. IEEE/ASME Transactions on Mechatro-nics, 2018, 23(1): 89-100.
[11] FU X B, GAO F, WU J, et al. Spatiotemporal attention networks for wind power forecasting[C] //2019 Inter-national Conference on Data Mining Workshops(ICDMW). Beijing, China: IEEE, 2020: 149-154.
[12] AFRASIABI M, MOHAMMADI M, RASTEGAR M, et al. Advanced deep learning approach for probabilistic wind speed forecasting[J]. IEEE Transactions on Industrial Informatics, 2021, 17(1): 720-727.
[13] LIU Y Q, QIN H, ZHANG Z D, et al. Probabilistic spatiotemporal wind speed forecasting based on a variational Bayesian deep learning model[J]. Applied Energy, 2020, 260: 114259.
[14] YU J J Q, GU J T. Real-time traffic speed estimation with graph convolutional generative autoencoder[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 20(10): 3940-3951.
[15] QI Y L, LI Q, KARIMIAN H, et al. A hybrid model for spatiotemporal forecasting of PM_2.5 based on graph convolutional neural network and long short-term memory[J]. Science of the Total Environment, 2019, 664: 1-10.
[16] KHODAYAR M, WANG J H. Spatio-temporal graph deep neural network for short-term wind speed forecasting[J]. IEEE Transactions on Sustainable Energy, 2019, 10(2): 670-681.
[17] ZHOU J, CUI G Q, HU S D, et al. Graph neural networks: A review of methods and applications[J]. AI Open, 2020, 1: 57-81.
[18] SUN C, CHEN Y Y, CHENG C. Imputation of missing data from offshore wind farms using spatio-temporal correlation and feature correlation[J]. Energy, 2021, 229: 120777.
[19] 冯俊磊, 吕卫东, 段雪艳, 等. 基于模态分解和TCN-BiLSTM的风电功率预测[J]. 电子测量技术, 2024, 47(14): 49-56. FENG Junlei, LYU Weidong, DUAN Xueyan, et al. Wind power prediction based on mode decomposition and TCN-BiLSTM[J]. Electronic Measurement Techno-logy, 2024, 47(14): 49-56.
[20] 乔宽龙, 董存, 车建峰, 等. 时空特性下基于图卷积神经网络的风电集群功率短期预测方法[J]. 太阳能学报, 2024, 45(5): 95-103. QIAO Kuanlong, DONG Cun, CHE Jianfeng, et al. Short-term prediction method of wind power clusters based on graph convolution neural network under spitio-temporal characteristics[J]. Acta Energiae Solaris Sinica, 2024, 45(5): 95-103.

相关文章 15

[1]	赵峰,刘瑞,王英,陈小强,葛磊蛟,马爱平. 多尺度融合与动态自校正旋转的吊弦检测算法[J]. 山东大学学报 (工学版), 2026, 56(2): 1-10.
[2]	刘飞宇,张静,王亦楠. 基于渐近式特征融合的轻量化SAR舰船检测算法[J]. 山东大学学报 (工学版), 2026, 56(2): 52-59.
[3]	王禹鸥,苑迎春,何振学,何晨. 融合多特征和多头自注意力机制的高校学业命名实体识别[J]. 山东大学学报 (工学版), 2025, 55(6): 35-44.
[4]	周群颖,隋家成,张继,王洪元. 基于自监督卷积和无参数注意力机制的工业品表面缺陷检测[J]. 山东大学学报 (工学版), 2025, 55(4): 40-47.
[5]	李丰,文益民. 融合多尺度视觉和文本语义特征的图像描述生成算法[J]. 山东大学学报 (工学版), 2025, 55(3): 80-87.
[6]	王禹鸥,苑迎春,何振学,王克俭. 改进RoBERTa、多实例学习和双重注意力机制的关系抽取方法[J]. 山东大学学报 (工学版), 2025, 55(2): 78-87.
[7]	邹正标,刘毅志,廖祝华,赵肄江. 动态交通流量预测的时空注意力图卷积网络[J]. 山东大学学报 (工学版), 2024, 54(5): 50-61.
[8]	徐金华,罗义凯,李昱燃,李岩. 基于时频分解与深度学习的轨道客流预测[J]. 山东大学学报 (工学版), 2024, 54(2): 60-68.
[9]	李家春,李博文,常建波. 一种高效且轻量的RGB单帧人脸反欺诈模型[J]. 山东大学学报 (工学版), 2023, 53(6): 1-7.
[10]	王碧瑶,韩毅,崔航滨,刘毅超,任铭然,高维勇,陈姝廷,刘嘉巍,崔洋. 基于图像的道路语义分割检测方法[J]. 山东大学学报 (工学版), 2023, 53(5): 37-47.
[11]	宋佳芮,陈艳平,王凯,黄瑞章,秦永彬. 基于Affix-Attention的命名实体识别语义补充方法[J]. 山东大学学报 (工学版), 2023, 53(2): 70-76.
[12]	刘方旭,王建,魏本征. 基于多空间注意力的小儿肺炎辅助诊断算法[J]. 山东大学学报 (工学版), 2023, 53(2): 135-142.
[13]	武新章,梁祥宇,朱虹谕,张冬冬. 基于CEEMDAN-GRA-PCC-ATCN的短期风电功率预测[J]. 山东大学学报 (工学版), 2022, 52(6): 146-156.
[14]	侯月武,刘兆英,张婷,李玉鑑,孙长明. 基于改进的DUNet遥感图像道路提取[J]. 山东大学学报 (工学版), 2022, 52(4): 29-37.
[15]	梁晔,马楠,刘宏哲. 图像依赖的显著图融合方法[J]. 山东大学学报 (工学版), 2021, 51(4): 1-7.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed