Artificial Neural Network Solution for Anomalous Operation Detection in Power Systems

Nathan Elias  Maruch Barreto; Ciro  Monteiro Baer; Mateus Jaensen Daros; Marlon Alexsandro  Fritzen; Guilherme  Schneider de Oliveira; Rafael  Rodrigues; Alexandre  Rasi Aoki

doi:10.48011/sbse.v1i1.2158

Nathan Elias Maruch Barreto Federal University of Paraná
Ciro Monteiro Baer Federal University of Paraná
Mateus Jaensen Daros Federal University of Paraná
Marlon Alexsandro Fritzen Federal University of Paraná
Guilherme Schneider de Oliveira Federal University of Paraná
Rafael Rodrigues TECPAR
Alexandre Rasi Aoki Federal University of Paraná

DOI: https://doi.org/10.48011/sbse.v1i1.2158

Keywords: Artificial neural networks, Power system simulation, Power system faults, Fault detection, Phasor measurement units

Abstract

This paper presents an anomalous operation detection system for power systems using the artificial neural network approach while discussing its advantages and disadvantages. The initial data for the proposed technique is a set of simulated post-fault bus voltages and currents obtained in a sampling rate so as to emulate a phasor measurement unit network. Several types of faults are dealt with, such as three-phase to ground, two-phase, two-phase to ground and single-phase to the ground as well as line and load contingencies. All fault and steady-state simulations were performed on MATLAB using Graham Rogers’ Power System Toolbox. The artificial neural network was designed on MATLAB, using an architecture proper for pattern recognition with supervised learning and obtaining high accuracy predictions within a short amount of time. The test system used in all simulations is the IEEE 39-Bus New England Power System, which presents 10 generation units, 21 loads and three distinct areas alongside transient and sub transient models, with phasor measurement units in 14 buses. Future works are discussed, showing the possibilities for feature engineering in this type of problem, fault type detection and fault location in operation using analogous dataset and neural network structures.