Using neural network to interpret multiple alarms

Abstract

The problem of interpreting a large number of alarms in a power system control center under stress conditions is considered. A description is given of an experiment using a neural network to build and test an intelligent alarm processor, that can analyze the multiple alarms associated with a system problem and identify the particular problem causing the alarms. The offline test results, although preliminary, suggest that the neural network approach is promising in developing a generic intelligent alarm processor that could be implemented by various electric utilities with minimal customization effort.