Open Access

ARTICLE

Fault Identification and Location in Distribution Networks Based on Grayscale Preprocessing and Graph Transformer

Jilin Wang¹, Haipeng Chen^2,*
1 State Grid Zhangye Power Supply Company, Zhangye, China
2 Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Northeast Electric Power University, Ministry of Education, Jilin, China
* Corresponding Author: Haipeng Chen. Email:
(This article belongs to the Special Issue: AI-Enabled Resilient Distribution Networks and Active Distribution Systems)

Energy Engineering https://doi.org/10.32604/ee.2026.079112

Received 14 January 2026; Accepted 11 March 2026; Published online 27 July 2026

Abstract

With the advancement of digital power inspection, distribution network images are often affected by noise, perspective changes, and occlusions, which can easily cause false detections during fault location in complex scenes. To address this issue, this paper proposes a fault identification and zoning location method based on grayscale feature enhancement and graph transformer for operational images collected by drones and ground vehicles. First, through weighted grayscale transformation, denoising, Otsu’s threshold segmentation, and the introduction of an attention mechanism, key texture and edge features such as cracks, ablation, and deformation are captured. Subsequently, a graph structure is constructed based on the distribution network topology, and device node attributes and spatial dependencies are incorporated into the encoding to jointly mode fault type discrimination and fault zoning location. Using the IEEE-33 system as a case study, the proposed model achieves a fault-identification accuracy of 93.76% and 98.25% fault-zone location accuracy. It also maintains high location accuracy (≥97%) even in the presence of 1–6 missing nodes and 30–45 dB Gaussian white noise, demonstrating good generalization and noise robustness. Research shows that fusing robust visual priors in the grayscale domain with graph-level global attention can effectively improve the performance of distribution network fault location in complex scenarios.

---

Keywords

Distribution network; computer vision; graph transformer network; fault location

---