Sensibility Analysis of a Fault Location Method Based on ANN, WPT and Decision Tree in Distribution Systems

Given the possibilities provided by smart grids in terms of communication infrastructure and information acquisition, there are new options on how to use the signals coming from meters to locate short circuits that occur in the system. This paper presents a framework for fault location in radial distribution systems based on machine learning algorithms and a multistage approach. A methodology in order to segment the system for proper identification of the outage region is presented. Studies are carried out involving the variation of the fault impedance, the fault incidence angle and the number and position of the meters. The IEEE 34-bus bar distribution feeder was considered for the tests. The results so far are promising, attesting and validating the presented methodology.