A New Approach for Fault Classification Based on S-Transform and Artificial Neural Network

Gabriella P. Santos¹; Gustavo G. Santos¹; Thiago S. Menezes¹; José Carlos M. Vieira Júnior¹; Pedro Inácio N. Barbalho¹

¹ Universidade de São Paulo

Resumo

Traditionally, fault location techniques in power distribution systems have used fault classification methods as input. Recently, due to the insertion of Distributed Generation (DG), the need for new fault classification methods, that considers the uncertainties associated with such generation, has become imperative. Thus, this study proposes the classification of faults in an unbalanced distribution system, considering variations in the fault incidence angle, resistance, distance and type also in the presence of DG. For this, the voltage and current signals measured at the substation were processed applying the S-transform (ST) and reduced using the Principal Component Analysis (PCA). After that, an Artificial Neural Network (ANN) performed the fault classification. The effectiveness of the proposed methodology was evaluated, using computational simulations, in a modified version of the IEEE 34 bus system. The software PSCAD simulated the system and MatLab performed the fault classification. In general, the proposed approach presented reliable results, achieving accuracies higher than 99%, even for different tests considering DG contributions. Thus, the presented fault classification is a promising approach for distribution systems.

Palavras-chave: artificial neural network; distributed generation; fault classification; principal component analysis; S-transform