Fault type determination with artificial neural networks in electricity distribution systems

In the decline in electrical events, the energy-related and stable event is important. In this study, artificial neural networks are used to select real short circuits in four different electrical designs. These four different classes of power lines, reference and overhead are mixed lines. Insufficient calculations and training results of the deficiencies in the images consisting of malfunctions that occur in the small of this electricity, as well thought and old model computational from insufficient predictions. The three networks and their trainings differ in their success. The network type used is Feed Forward Back Propagation network type, Cascade Connected Feed Forward Back Propagation network type, Elman Feedback network type. His trainings are Levenberg-Marquardt (LM), Powell/Beale Restarts with Conjugate Gradient (CGB), One Step Secant (OSS), Variable Learning Rate Backpropagation (GDX), Momentum Gradient Descent (GDM), Scaled Conjugate Gradient (SCG), and NRP (Resilient backpropagation). This training fed back propagation network type gave more successful results.