Application of neural-network computing in intelligent alarm processing (power systems)

Abstract

A neural-network approach was used in building and testing an experimental intelligent alarm processor (IAP) which analyzes the multiple alarms associated with a power system problem and identifies the particular problem causing these alarms. The test results, although preliminary, suggest that the neural-network approach could be the answer to the problem of developing a generic intelligent alarm processor that can be implemented by utilities with minimal customization effort. When there is no missing alarm data, the experimental IAP correctly interprets 100% of the system problems. When there is missing alarm data, and if such missing data will not result in an alarm combination identical to one associated with another system problem, the experimental IAP successfully makes the right guess. This is similar to the behavior of a human system operator. >