A Novel Ann Training Approach for Supervised and Constructive Learning Applied to Fault Classification & Short-circuit Zoning in Rural Primary Distribution Systems

In this paper, artificial neural networks (ANN) are proposed to classify and establish zoning faults (single-phase, bi-phase and tri-phase short circuits) in primary distribution systems. The proposed classification and zoning models are oriented toward the stochastic tracking of failures via emergency brigades. This work, instead of training a big ANN model, proposes to decompose the problem into two smaller ones in order to take advantage of the available previous knowledge and improved training. The classification model is based on an ANN with supervised learning (SL). The zoning representation is based on a simple binary perceptron with constructive learning (CL). In this article, the theoretical elements underpinning the proposal are detailed and justified. The performance of the proposed models is verified through various real primary network applications.