Development of transient fault management methodology

This paper presents the results of development studies to diagnostic and support to locate faults in distribution systems with recognition patterns using neural networks. It presents the developed methodology for input data processing for neural networks, as well as for fault type identification. In this methodology is necessary to get input data and were designed equipment prototypes to measure and transmit data's and voltage transducers. These equipments were installed in two feeders in the concession area of an electrical energy distribution company. The paper conclusion is that all results obtained indicate the suitability of neural networks to support transient fault management.