Data-Driven Coherency Identification for Generators Based on Spectral Clustering

The wide-area measurement system provides a new data acquisition and supervisory control tool for a power system, and the data acquisition level is increased dramatically with its development in the smart grid environment. Huge data associated with the power system operation are acquired, which are beneficial for enhancing situational awareness of a power system concerned. Identifying the coherency among synchronous generators using real-time signals from phasor measurement units (PMUs) is one of the major tasks of situational awareness in power system operation. Given this background, a data-driven coherency identification methodology is proposed based on the spectral clustering algorithm. First, several trajectory dissimilarity indices for the rotor angle and rotor speed trajectories of generators as measured by PMUs are presented based on the trajectory similarity theory. Second, a decision-making method based on the Gini coefficient and Kendall rank correlation coefficient is presented for integrating multiple indices describing trajectory dissimilarities. Third, the spectral clustering algorithm is presented to identify the coherency of synchronous generators, and silhouette is presented for determining a reasonable number of coherent groups. Finally, oscillation events happened/simulated in two actual power systems, i.e., Guangdong power system in China and Western Interconnection power system in North America, are utilized to demonstrate the effectiveness of the proposed data-driven coherency identification methodology.