Um localizador de faltas em linhas de transmissão baseado em redes neurais artificiais.

Abstract

A fault location algorithm for transmission lines using Artificial Neural Networks (ANNs) and presented in this work. The architecture of adopted network was the multi-layer perceptron. The Neuranalysis and ATP software were used to build the database used in the training, testing and validation of RNAs. The entries for the ANNs are the samples of phase and zero sequence voltages and currents. The fault conditions were simulated for a 230 kV transmission line. Each file needed for the construction of the database and automatically generated in a format standardized and executed in batch mode. For fault location, the transmission and divided into 8 zones. Previous to localization, classification is done the type of lack training an ANN with the complete database. After classification, and the zone in which the fault is found is selected. To the location, 8 RNAs are trained for each type of fault, each with the specific zone data. The locator has a total of 37 RNAs, and presented satisfactory results for the analyzed problem.