Fault-pattern discovery in sequences of voltage sag events

A pattern discovery approach applied to sequences of voltage sags is proposed for fault modeling and failure prediction. It is assumed the existence of some failures that are the result of accumulated stress produced by overcurrents during previous faults or the existence of pattern sequences produced by the reclosing of the protection system in presence of a fault. So there may exist relationships between events. Events registered in a real power distribution system are analyzed to search those fault patterns. The events considered in this study are voltage sags generated by homopolar faults and registered by power quality monitors installed in the secondary of transformers in distribution substations. The ordered list (according to their time of ocurrence) of events corresponds to a sequence, and magnitude and duration are the attributes considered to describe the events. The aim of this work is to find patterns in a sequence of voltage sags that allow to recognize relationships between different types of sags occurred along time.