Investigation of the Possibility of Applying Neural Networks for Selecting Methods of Remote Fault Location Based on Synchrophasor Measurements

According to the order of the Ministry of Energy of the Russian Federation No. 101, phasor measurement units (PMUs) are installed at large power facilities. Currently, they are used for the analysis of electromechanical processes, steady state modes, and supervisory control and data acquisition (SCADA). However, the research results of several authors [ 1 – 11 ] indicate the prospects of their application for the purposes of fault location (FL). Many fault location methods have been developed by research teams [ 1 , etc.], the accuracy of which depends on the configuration of overhead transmission lines (OTL) and various factors. The current industry guidelines contain recommendations for averaging the results of determining the distance to the fault point obtained by different methods, which, in the presence of methods with large errors in the sample, will lead to unsatisfactory results. At the same time, it is recommended to select FL methods based on the operating experience of a specific OTL, which may be absent. The ability to collect synchrophasor measurements using phasor data concentrators makes it possible to create centralized FL systems at the level of Interconnected Power System / Unified Power System. The implementation of many FL methods based on emergency mode parameters with automatic selection of the most accurate method for a specific OTL and case of its damage based on machine learning in such systems will reduce the human factor (in terms of selecting the FL method based on operating experience) and achieve the lowest FL error.