Machine learning of rules for a power system alarm processor

An important function of a supervisory control and data acquisition (SCADA) system is the annunciation of alarms. The aim of alarm processing is to keep the number and annunciation rate of alarms manageable during emergencies. The majority of alarm processors have been implemented as hard-coded rule-based expert systems. The authors describe the use of machine learning to induce knowledge for an alarm processor from alarm sequences obtained via the SCADA from a power distribution system. An evaluation is made using data pertaining to a recent distribution system emergency