A hybrid approach for fault location in power distributed networks: Impedance-based and machine learning technique

Abstract

Fault location (FL) is one of the challenges in distribution networks. Many impedance-based methods have been developed to address this challenge. However, these impedance-based methods are not yet able to provide a unique answer, and they only can approximate the distance of the fault from the reference point, which creates challenges in systems that have a large number of separate lines. In this paper, using the impedance method, as well as a deep neural network, a new method for FL is suggested, which can provide a unique answer. In this approach, the FL is determined in less than 6 s, and the accuracy of 99 percent. The deep memory neural network is then used to accurately detect the exact fault location between candidate points. This method also identifies the line where the fault occurred. The performed simulations confirm the validity of the proposed method.