Voltage sag and short interruption edge-side tracing model based on a spiking spatiotemporal causal graph convolutional network
AIP Advances · 01.11.2026 · 16 (11)
DOI: 10.1063/5.0333663 ↗Kurz gefasst
Fachartikel in der Zeitschrift AIP Advances, online erschienen am 05.11.2026. Verfasst von Chenda Jiang und 1 weiteren Autor:innen. Die Arbeit nennt 33 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Sanming City Science and Technology Plan Guiding Project.
Abstract (Originalsprache)
The widespread deployment of renewable energy in modern power systems has rendered voltage dips and short-term interruptions critical bottlenecks for power quality. Existing solutions face three major limitations: inadequate adaptation to edge computing resources, high confusion rates in disturbance type identification and source localization, and considerable inference latency, which precludes simultaneous satisfaction of millisecond-level accuracy and low computational overhead. To address these issues, this paper proposes a pulse spatiotemporal causal graph convolutional network (SSTCGCN) for edge-based collaborative localization and classification of disturbance sources. The scheme innovatively integrates pulse coding with causal graph convolution and adopts a dual-branch differentiation architecture to reduce misclassification risks among disturbance types. Through optimization of the lightweight causal discovery module and the pulse encoding mechanism, parameter redundancy is reduced by 40%, while dynamic adjacency matrix optimization enhances topological adaptability. A case study on a 32-node industrial park power grid demonstrates that SSTCGCN achieves 99.15% (±0.42) accuracy in joint disturbance source localization and type discrimination, outperforming traditional combined methods by 2.92%. The inference delay is only 12.68 ms (±1.57 ms), representing a 17.23% reduction over standard SNN networks, and computational power consumption is lowered by 76.01% relative to the CTGCN model. Under the tested edge computing condition (200 MFLOPS), the proposed method attains an edge-side adaptation index of 0.92 (±0.02) with high robustness, effectively alleviating core technical bottlenecks within the specified grid scale and disturbance type range. It offers a feasible technical reference for the digital transformation and lightweight implementation of medium-sized distribution networks.
Förderung
- Sanming City Science and Technology Plan Guiding Project
Zitieren
APA
Jiang, C., & Li, J. (2026). Voltage sag and short interruption edge-side tracing model based on a spiking spatiotemporal causal graph convolutional network. AIP Advances, 16(11). https://doi.org/10.1063/5.0333663
Vancouver
Jiang C, Li J. Voltage sag and short interruption edge-side tracing model based on a spiking spatiotemporal causal graph convolutional network. AIP Advances. 2026;16(11). doi:10.1063/5.0333663
BibTeX
@article{Jiang2026,
author = {Jiang, Chenda and Li, Jin},
title = {Voltage sag and short interruption edge-side tracing model based on a spiking spatiotemporal causal graph convolutional network},
journal = {AIP Advances},
year = {2026},
volume = {16},
number = {11},
doi = {10.1063/5.0333663}
}
EUVIMED