Data mining framework for power quality event characterization of iron and steel plants

In this paper, a power quality (PQ) knowledge discovery and modeling framework has been developed for both temporal and spatial PQ event data collected from transformer substations supplying iron and steel (I&S) plants. PQ event characteristics of various I&S plants have been obtained based on clustering and rule discovery techniques. The data are collected by the PQ analyzers, which detect the voltage sags, swells, and interruptions according to the IEC Standard 61000-4-30. The constructed clustering strategy ensures feasible system monitoring by reducing unmanageable number of PQ events collected by the distributed PQ measurement systems into event clusters count. An abstraction for event representation has been developed, through which representative feature bags are constructed for each event to be used in the similarity decisions. The developed model has been applied satisfactorily to PQ event data obtained from 15 major transformer substations supplying heavy industry zones of the transmission system up to a five-year time period and from two additional transformer substations supplying some other industrial zones, for comparison purposes. The developed PQ data mining framework, which is used to identify PQ event distributions based on the event descriptions given in the IEEE Std. 1159, provides a useful analysis and evaluation infrastructure for taking countermeasures against the most probable event occurrences, specific to those feeders of I&S plant transformer substations.