Short Circuit Classification using the Discrete Fractional Fourier Transform and Artificial Neural Network

Abstract

The basic principle of the protection philosophy is to select, coordinate, adjust and allocate the various equipment and protective devices in an electrical system to keep a specific relationship between them. An abnormality in the system can be isolated or removed without affecting other parts of the system. Another concern linked to protection systems is the efficiency of the distribution network at critical moments: many consumers can remain without electricity supply after the protection system has operated. Thus, the time spent by maintenance teams in locating the point of occurrence of the fault and preparing a diagnosis of the problem and corrective or even preventive measures should be as little as possible.This paper presents a methodology for detecting and classifying short-circuit faults in power distribution systems. An artificial neural network is applied to categorize short circuit types. The pre-processing of signals is carried out through the fractional Fourier transform, a variation of the Fourier transform, which allows the representation of signals in the domains between time and frequency. The developed system showed accuracy in all tests performed, detecting faults, classifying and identifying the phase affected by single-phase and two-phase faults.