Eigen-analysis based power quality event data clustering and classification

In this work, an eigen-analysis based power quality (PQ) event data clustering and classification method has been developed, which is aimed to serve the needs of the smart-grid applications. With the proposed clustering approach, huge amount of PQ event data, which corresponds to voltage sags, swells and interruptions, are classified into finite number of classes and spatial classification of those clusters provides characterization of specific parts of the grid. The proposed method is based on k-means clustering of the feature space, which is selected as the voltage rms values, suggested by the IEC 61000-4-30 Standard. To reduce the number of optimum clusters and to increase clustering efficiency, two eigen-analysis based transformations, principle-component-analysis (PCA) and linear-discriminant-analysis (LDA), have been applied on feature space before k-means clustering. Eigen-analysis has reduced the clustering distances and provided more efficient clustering and PCA+k-means algorithm has given the best clustering in terms of PQ event characterizaton.