ESTUDO DA APLICAÇÃO DE REDES NEURAIS ARTIFICIAIS PARA IDENTIFICAÇÃO DE CURTO-CIRCUITOS NO SISTEMA ELÉTRICO DE DISTRIBUIÇÃO

Abstract

The algorithm of artificial neural networks (RNA), are computational models that can perform generalization, inferences, identification, and classification of information and patterns. Thus, in this work, a study was developed through the creation of a neural network classifying patterns to identify and classify the types of short circuits that occur in the electrical distribution system. Thus, a multilayer perceptron neural network was developed in the Matlab software with 3 hidden layers, 25 neurons in each hidden layer, and a hyperbolic tangent activation function. The PMC was trained using simulated short-circuit data in the ATPDraw software and presented an efficiency of 94.7% in the identification of short circuits in the validation stage. The trained network was also able to evaluate short circuits on an IEEE 9-bar test bus demonstrating the potential to be applied as an additional measure of network information in integrated operation centers (IOC).