Automating the Diagnosis of Occurrences in Power Plants Using Data from DFR and Sequence of Events: an Expert System Based Methodology

This paper describes a novel methodology for automating the analysis of occurrences in power plants. The proposed scheme uses, as input features, oscillographic records from digital fault recorders (DFR) and sequence of events (SOE) from supervisory systems. The objective of the methodology is to help the engineer responsible for this kind of analysis by means of a pre-classification of the data. The analysts can then focus their attention to the most relevant occurrences. The tool described here is based on expert systems. It was developed and tested using data from real occurrences. The results show that the methodology is effective to automatically classify the most common situations in a power plant, thus helping the engineers deal with the large amount of data available today in the operation centers.