Synchrophasor-based data mining for power system fault analysis

Phasor measurement units can provide high resolution and synchronized power system data, which can be effectively utilized for the implementation of data mining techniques. Data mining, based on pattern recognition algorithms can be of significant help for power system analysis, as high definition data is often complex to comprehend. In this paper three pattern recognition algorithms are applied to perform the data mining tasks. The deployment is carried out firstly for fault data classification, secondly for checking which faults are occurring more frequently and thirdly for identifying the root cause of a fault by clustering the parameters behind each scenario. For such purposes three algorithms are chosen, k-Nearest Neighbor, Naïve Bayes and the k-means Clustering.