Comparison of various neural network models applied to adaptive distance protection

Abstract

This paper elaborates settings and modification of the operating characteristic of adaptive distance relays which protect a transmission line between two terminals. Operating characteristic should have such shape to satisfy all conditions in the power system. However it is known that there is no combination of characteristics which can satisfy all possible conditions. So the only solution is modification of the operating characteristic. Solution of this problem involves two types of neural networks, multilayered perceptron and the self-organizing neural network, the so called Kohonen network. Here the authors consider a two-terminal transmission line, confirm that fault resistance and location of faults can produce erroneous relay function and finally suggest ways to ensure the generation of the correct signal for relay operation. With the simple simulation, the authors present the advantages of the particular models and the credibility of the developed models.