Abstract
Fault of the distribution feeder on the power delivery would cause the power outage, locating the fault quicker so to reduce the outage time. Various methods that use pre-fault and post-fault voltage and current phasors/values for distribution feeder fault location have been developed. Fault location accuracy is still limited by the inherent imprecision on the estimation of feeder model parameters and uncertain arcing resistance at the fault location. In Taiwan, feeder patrols identified the fault locations by referencing to the regional distribution of the trouble calls, the abnormal observations of the feeders, complained or reported in the calls, and the conditions in the surrounding environments. Feeder patrols record each fault with a table including time, date, month, year, address, equipment of fault, causes or accidents and so on. The database collects the information of many years and accumulates a large base of records. Data mining offers tools for discovery of patterns, associations, changes, anomalies, rules, and statistically significant structures and events in data. The rough set theory of data mining is a viable approach for extraction of meaningful knowledge and making predictions for an individual data object (e.g. fault occurrence) rather than a population of objects. In particular, little research has been done to apply rough set theory for fault location in power delivery system. This study aims to use rough set theory of data mining for distribution feeder fault location. In this paper, the approach will be test on the historical data of distribution feeder faults occurred within the business area of Taipei City District Office of Taipower Company.