Localizador de faltas utilizando medição em regime permanente e redes neurais artificiais

Abstract

The distribution system grows continuously thus increasing the probability of the emergence of faults in its lines. In order to solve this problem was developed a program that can, through the RMS values of the electrical quantities of the main bar of a distribution system indicate possible predictions of the location of the occurrence of a system fault. For the implementation of the program was used three softwares: simulation of faults, programming in Python, and a software of artificial neural nets, the KNIME, for data prediction. With the application of faults in all buses and the use of artificial intelligence it was possible to acquire electrical quantities such as voltage, current, power and power factor and create Artificial Neural Networks. As results we obtained the predictions of the fault sites and their errors and then it was possible to conclude that the RNA can efficiently determine the location of the fault, becoming an attractive project for since, one can reduce with fault locators, energy costs, travel costs and time without electricity, thus increasing customer satisfaction rates