Title: A hybrid graph attention mechanism for load forecasting based on efficient spatiotemporal feature extraction

Authors: Jie Chen; Yajing Tang; Xiao Liu; Ling Wang; Mao Tan; Haodong Zhang

Addresses: School of Automation and Electronic Information, Xiangtan University, Hunan, 411105, Xiangtan, China ' School of Automation and Electronic Information, Xiangtan University, Hunan, 411105, Xiangtan, China ' School of Automation and Electronic Information, Xiangtan University, Hunan, 411105, Xiangtan, China ' Department of Automation, Tsinghua University, 100084, Beijing, China ' School of Automation and Electronic Information, Xiangtan University, Hunan, 411105, Xiangtan, China ' School of Automation and Electronic Information, Xiangtan University, Hunan, 411105, Xiangtan, China

Abstract: Load forecasting is fundamental to optimising and dispatching power systems. Despite the efficiency of existing load forecasting methods, they fall short in extracting more comprehensive feature representations. To solve this problem, this paper presents a spatiotemporal load forecasting method that integrates key factors, including historical load patterns, footfall, and meteorological conditions. Our method leverages residual graph convolutional networks (ResGCN) and long short-term memory (LSTM) as the primary models for forecasting. Firstly, we pinpoint the most significant variables by correlation analysis. Subsequently, we extract spatiotemporal features from load graphs and input these features into our forecasting model. The model integrates a local-global graph attention (LGGA) mechanism to incorporate local and global information, enhancing the understanding of load data. Additionally, we employ a convolutional block attention module (CBAM) to fine-tune the feature representations, thereby improving model sensitivity. Experimental results demonstrate the superiority of our method.

Keywords: load forecasting; spatiotemporal feature; graph attention mechanism; graph convolutional network; LGGA; local-global graph attention; CBAM; convolutional block attention module; ResGCN; residual graph convolutional network.

DOI: 10.1504/IJAAC.2026.152107

International Journal of Automation and Control, 2026 Vol.20 No.2, pp.169 - 192

Received: 27 Jun 2024
Accepted: 18 Sep 2024
Published online: 09 Mar 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article