Short Circuit Fault Detection and Categorization for Secondary Distribution Network Using Non-Contact Magnetic Measurements

초록

Reliable electric power distribution strongly depends on rapid fault detection. Overhead distribution conductors are particularly prone to short circuit faults mainly caused by growing vegetation and other environmental factors. This paper introduces an automatic and non-contact short circuit fault detection and classification technique using flux leakage measurements together with machine learning algorithms. Unlike traditional current or voltage transformer based approaches, the proposed technique uses flux leakage measurements from a distribution transformer. Multiple sensor positions are explored for optimal performance via simulating a 200 kVA distribution transformer using the finite element analysis method. Among the tested classifiers, the designed neural network algorithm achieved 98% accuracy across 10 distinct fault types while classifying 669 scenarios in less than 0.16 seconds. The results demonstrate the high efficiency and feasibility of flux leakage based sensing for real time distribution network protection.

키워드