Rough set theory for data mining for fault diagnosis on distribution feeder

Distribution feeder faults cause power outages, therefore it is crucial to diagnose and thus locate the fault quickly to reduce the duration of the outage. In practice, feeder patrols usually identify the fault locations by referencing the regional distribution of the calls reporting trouble, the abnormal observations of the feeders that have been reported in the calls, and the observed conditions in the surrounding environments. Feeder patrols in Taiwan have recorded each fault on a table that includes time, date, month, year, address, equipment at fault, causes or accidents, and etc. The database has accumulated a large base of information for many years. This study aims to use rough set theory as a data-mining tool to derive useful patterns and rules for distribution feeder faulty equipment diagnosis and fault location. In particular, the historical data of distribution feeder faults of Taiwan Power Company was used for validation and the results show the practical viability of the proposed approach.