A New Transmission Line Protection Approach Using Cumulants and Artificial Neural Networks

Abstract

This article presents a new methodology for transmission line protection that uses higher-order statistics (HOS), cumulants, and artificial neural networks (ANNs). The main objective is to design a distance relay algorithm. Results for the fault-detection and -classification stages are presented, as well as for fault location. The proposed method combines a large number of samples of cumulants with different features and the capability of ANNs to discriminate different patterns. In summary, HOS are used for feature extraction in the protection scheme. The ANNs receive these statistics as inputs, and they are responsible for the logical functioning of the protection system, deciding if a trip is needed after detecting, classifying, and locating a fault. The results have shown that the proposed approach is suitable for protection purposes. For the fault-detection stage, results have shown to be immune to the high presence of additive noise and also to the power-system frequency deviation. Moreover, the fault-classification stage is computed without the need of current information from the power system. Finally, the preliminary results for fault location are precise for a correct estimation of fault distance and determination of the fault zone. It must be highlighted that this new distance protection approach is essentially based on voltage signals, using current signals only for determining the direction of the fault. This fact represents an innovation in distance relaying.