Work

Topology-Adaptive ground fault location method for distribution networks

Year: 2026

Type: article

Abstract: • Topology-adaptive fault location for distribution networks with reconfiguration. • Wavelet-based panoramic traveling-wave features from multi-point measurements. • Topology encoding via random-walk and Laplacian positional embeddings. • Transformer encoder learns topology–feature correlations for localization. • 99% branch accuracy and < 20 m mean error; works up to 10 kΩ faults. With the large‐scale integration of distributed generation, the operational characteristics of distribution networks have become increasingly diverse. Frequent changes in network topology and complex variations in fault feature distributions reduce the reliability and accuracy of existing protection and fault location methods. To address this issue, this paper proposes a topology-adaptive fault location method for distribution networks. First, multi-dimensional fault information in traveling waves is extracted via a wavelet transform to generate panoramic traveling-wave patterns, and the impact of topology changes on traveling-wave feature distributions is analyzed in the time, frequency, and spatial domains. Second, a Transformer-based fault-location model is developed, where graph random-walk and graph Laplacian encodings are employed to represent topological information, and the Transformer encoder learns multi-point correlations to capture the relationships among panoramic patterns, system topology, and fault locations. Finally, by embedding topology variations into the feature-extraction process, the model learns essential fault features that are insensitive to topology changes yet indicative of the actual fault location, thereby enabling accurate fault localization. Simulation results demonstrate that, across datasets containing multiple topology types, the proposed method achieves a fault-location accuracy exceeding 99% and an average absolute location error of less than 20 m, showing strong adaptability to different topology variations. Moreover, it exhibits high sensitivity to weak faults and accurately locates 10 kΩ grounding faults.

Source: International Journal of Electrical Power & Energy Systems

DOI: 10.1016/j.ijepes.2026.111675

Authors: Feng Deng, Zhentong Wang, Yilin Chen, Junqin Liu

Institutions: Changsha University of Science and Technology, State Grid Corporation of China (China), State Grid Hunan Electric Power Company Limited

Language: English

Topic: Power Systems Fault Detection

Subfield: Control and Systems Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Industry, innovation and infrastructure, Affordable and clean energy

Open Access status: gold

Estimated APC paid: $3,920