Power system fault detection and classification using probabilistic approach

Abstract

AbstractFinal decision of any digital power system protection is made in several steps including: fault recognition, estimation of criterion values and decision making concerning the state of protected plant. Whatever methods (deterministic or probabilistic) are used by digital signal processing, the protection system must identify fault instant and type at first, which starts all further required procedures. Since fault appearance and parameters are of probabilistic nature, some of the decision tasks may be solved on this basis. In the paper an extension of sequential decision method (first proposed by Sakaguchi) for fault detection and classification is presented. The algorithm is based on probabilistic characteristics of criterion values and hypothesis testing. Proposed method is optimized in sense of speed and accuracy together with measurement algorithm. Some indications for proper choice of criterion values based on probabilistic concept are also given. Intensive electromagnetic/alternative transients program (EMTP/ATP) simulations have been made to obtain required data and to test the developed algorithm. Results of its implementation on a 400‐kV transmission line EMTP model are presented in comparison with an artificial‐neural‐network‐(ANN‐)based fault detection and classification method.