Semimonthly
ISSN 1000-1026
CN 32-1180/TP
Semimonthly
ISSN 1000-1026
CN 32-1180/TP
ZHANG Junkai
College of Information Science and Engineering, Northeastern University, Shenyang110819, ChinaHU Xuguang
College of Information Science and Engineering, Northeastern University, Shenyang110819, ChinaLIU Yaobo
College of Information Science and Engineering, Northeastern University, Shenyang110819, ChinaXU Qing
College of Information Science and Engineering, Northeastern University, Shenyang110819, ChinaMA Dazhong
College of Information Science and Engineering, Northeastern University, Shenyang110819, ChinaSUN Qiuye
College of Information Science and Engineering, Northeastern University, Shenyang110819, ChinaCollege of Information Science and Engineering, Northeastern University, Shenyang110819, China
Short-term residential load forecasting can provide real-time and flexible power demand information for virtual power plants, which is helpful for virtual power plants to realize efficient utilization of energy and optimize electricity market transactions. With the increasing prominence of correlation among residential loads, traditional forecasting methods, which primarily rely on time-series forecasting based on individual residential historical load, fail to satisfy the comprehensive demands for load interconnectivity in large-scale virtual power plant. Based on this, a short-term residential load forecasting method based on dynamic association graph attention networks for the virtual power plant is proposed. Firstly, a hybrid correlation analysis method is proposed to describe the linear and nonlinear relationship between residential loads, and a weight pruning threshold mechanism is further proposed to derive the hybrid correlation matrix of residential loads. Secondly, a dynamic association graph structure is constructed based on the hybrid correlation matrix, and a temporal graph attention mechanism is proposed to deeply learn the spatial-temporal association characteristics of residential loads, achieving the objective of short-term residential load forecasting. Finally, the effectiveness of the proposed method is verified by actual residential load data from a specific region.
This work is supported by National Natural Science Foundation of China (No. 62303103), Scientific Research Project of Liaoning Provincial Department of Education (No. JYTQN2023161) and Liaoning Provincial Natural Science Foundation of China (No. 2023-BSBA-140) .
| [1] | ZHANG Junkai, HU Xuguang, LIU Yaobo, et al. Short-term Residential Load Forecasting Based on Dynamic Association Graph Attention Networks for Virtual Power Plant[J]. Automation of Electric Power Systems,2024,48(21):120-128. DOI:10.7500/AEPS20240328002 |
Address：No.19 Chengxin Avenue, Nanjing 211106, Jiangsu Province, P. R. China Postcode：211106 ServiceTel：025-81093050
Publish: 025-81093071 Fax:025-81093040 E-mail：aeps@alljournals.cn
Copyright:Automation of Electric Power Systems ® 2025 All Rights Reserved Support:Beijing E-Tiller Technology Development Co., Ltd. ICP:苏ICP备09008660号