Detection and classification of short-circuit faults in distribution networks based on Fortescue approach and Softmax regression

Abstract

This paper proposes a method to detect and classify ten short-circuit faults in distribution networks, where the presence of distributed generators makes fault diagnosis a challenging problem. The main idea is to consider operating modes of distributed generators in analyzing fault characteristics via the Fortescue approach, and exploit the softmax regression to alleviate negative effects of transient data samples on the fault classification. The proposed method is developed in three main steps. First, the relationship between measurable currents and unavailable currents of the fault point is developed for the grid-connected mode or the islanding mode of distributed generators. Second, the Fortescue approach is used to formulate fault characteristics from the positive-, negative- and zero-sequence components of measurable currents. Third, the softmax regression is introduced to alleviate negative effects of transient data samples on the fault classification. The effectiveness of the proposed method is demonstrated via numerical examples on balanced and unbalanced distribution networks.