Determining fault's type and accurate location in distribution systems with DG using MLP Neural networks

Abstract

Finding and designing new methods for determining type and exact location of faults in power system has been a major subject for power system protection engineers in recent years. Fault locating in transmission networks is not very hard and complicated due to low impedance of faults. This job is usually done by distance relays. But, in distribution networks, because of high impedance of fault and its vast variety and also simplicity of protective devices, determining the exact location of faults is very complicated. On the other hand, penetration of distribution generation into distribution networks reinforces the necessity of designing new protection systems for these networks. One of the main capabilities that can improve the efficiency of new protection relays in distribution systems is exact fault locating. In this paper, a new approach for determining the exact fault type and location in distribution systems including distributed generation using MLP neural networks is presented. In the suggested method, after determining the fault type, by normalizing the fault current of the main source, the corresponding trained neural network has been activated and the exact location of occurred fault has been derived. The presented method has been implemented on a sample distribution network, simulated by DIgSILENT Power Factory 13.2, and its performance has been tested. The simulation results show high performance and accuracy of the method and substantiate that it can be used in modern heuristic protection schemes in distribution systems.