IDENTIFICATION OF COHERENT GENERATORS USING PCA AND CLUSTER ANALYSIS

Abstract – Principal Component Analysis (PCA) has been recently introduced in the power system dynamics literature as a time domain coherency identification method. The method showed to be promising especially with the increased growth of Phasor Measurement Unit (PMU) installations in power networks. The method, how-ever, suffers from the inability of visualizing transformed information beyond three dimensions. In addition, visual inspection of three dimensional (3-D) results can be some-times misleading. In this paper, transformed data by the PCA method are further clustered by using the spatial distances between objects representing generators ’ re-sponses to a disturbance in the system. The objects are then linked together to build a multi-level hierarchical tree representing the dynamic behavior of network generators. The proposed technique enables adjustable (variable) clustering, thus adding more flexibility and applicability in different types of applications requiring network model order reduction and more efficient monitoring and control of power systems. The technique is illustrated on the New England test system. The results showed multi-level clus-tering of highly meshed network regardless of the number of principal components required to capture the original data variations