Article: Electric-power telecommunication alarm analysis method based on spatio-temporal graph neural networks Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.10 pp.1 - 20 Abstract: With the growing complexity of electric-power telecommunication networks, alarm messages have surged, challenging operation and maintenance. Traditional methods overlook spatio-temporal characteristics, hindering root cause identification and fault propagation analysis. This paper proposes a spatio-temporal graph neural network-based alarm analysis method, extracting spatial and temporal features to construct a graph model and node alarm subtrees based on alarm propagation. Subtree features incorporate node relationships and alarm rules for accurate analysis. To reduce computational overhead from model parameters, a swarm intelligence optimisation-based lightweight technique is introduced, enhancing deployment efficiency in resource-constrained environments. Experiments demonstrate superior performance in alarm correlation accuracy, compression, and fault localisation compared to traditional methods. The approach effectively minimises redundant alarms, boosts maintenance efficiency, and supports intelligent network operation. Inderscience Publishers - linking academia, business and industry through research

Title: Electric-power telecommunication alarm analysis method based on spatio-temporal graph neural networks

Authors: Yuhang Pang; Tongtong Zhang; Qiong Liu; Jiaxing Ren; Meiru Huo; Jianliang Zhang

Addresses: Information and Communication Department, China Electric Power Research Institute, Beijing, 100192, China ' Information and Communication Department, China Electric Power Research Institute, Beijing, 100192, China ' Information and Communication Department, China Electric Power Research Institute, Beijing, 100192, China ' Information and Communication Department, China Electric Power Research Institute, Beijing, 100192, China ' Department of Technology Development, State Grid Information and Telecommunication Corporation of SEPC, Shanxi, Taiyuan, 030021, China ' Department of Technology Development, State Grid Information and Telecommunication Corporation of SEPC, Shanxi, Taiyuan, 030021, China

Abstract: With the growing complexity of electric-power telecommunication networks, alarm messages have surged, challenging operation and maintenance. Traditional methods overlook spatio-temporal characteristics, hindering root cause identification and fault propagation analysis. This paper proposes a spatio-temporal graph neural network-based alarm analysis method, extracting spatial and temporal features to construct a graph model and node alarm subtrees based on alarm propagation. Subtree features incorporate node relationships and alarm rules for accurate analysis. To reduce computational overhead from model parameters, a swarm intelligence optimisation-based lightweight technique is introduced, enhancing deployment efficiency in resource-constrained environments. Experiments demonstrate superior performance in alarm correlation accuracy, compression, and fault localisation compared to traditional methods. The approach effectively minimises redundant alarms, boosts maintenance efficiency, and supports intelligent network operation.

Keywords: electric-power telecommunication network; spatio-temporal graph neural network; alarm subtree; particle swarm optimisation; PSO.

DOI: 10.1504/IJICT.2025.146095

International Journal of Information and Communication Technology, 2025 Vol.26 No.10, pp.1 - 20

Received: 06 Jan 2025
Accepted: 20 Feb 2025
Published online: 06 May 2025 *

Title: Electric-power telecommunication alarm analysis method based on spatio-temporal graph neural networks

Keep up-to-date