Work

Identifying Mis-Wiring Induced Earth Faults in Low Voltage Distribution Network using Real-Imaginary Part Separated Regression

Year: 2026

Type: article

Abstract: Wiring faults involving the neutral and ground wire of user in low-voltage distribution networks (LVDNs) are prevalent yet notoriously difficult to diagnose, posing a significant threat to the reliable operation of residual current protection devices (RCDs). The extensive application of the Internet of Things (IoT) meter has provided critical data support for the monitoring and identification of users associated with wiring fault. Existing studies primarily utilize multi-variables regression and thresholding segmentation to identify anomalous users; however, they still suffer from inadequate identification accuracy, high computational overhead and limited transferability in different LVDNs. Considering the timeliness and exactness requirements for fault localization, this paper proposed a real-imaginary part separated regression (RIPSR) model to accelerate the computational speed by mapping complex-domain operations into the real domain while fully preserving the complex-valued information. Furthermore, to address the time-varying characteristics of leakage data and the limited discriminative capability of rigid thresholding, a sliding window mechanism is adopted to extract the temporal variation trajectories of each user’s weight coefficient. Subsequently, a density-based spatial clustering of applications with noise (DBSCAN) is employed to cluster users. Simulation studies on a 204-user LVDN platform indicate that, within the low false positive rate region below 2%, the proposed method improves recall by 21.2% while reducing the computation time by 65% compared with complex-valued regression. Its robustness is further tested under conditions representative of practical engineering environments.

Source: IEEE Internet of Things Journal

DOI: 10.1109/jiot.2026.3738796

Authors: Xiaofei Feng, Liu Guanke, Chong Yang, Bin Li, Sheng Su, Paul Barendse

Institutions: Changsha University of Science and Technology, Guangdong Power Grid Company (China), University of Cape Town

Language: English

Topic: Electrical Fault Detection and Protection

Subfield: Electrical and Electronic Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: hybrid

Estimated APC paid: $2,645