Data Analytic Tool for Clustering Identification based on Dimensionality Reduction of Frequency Measurements

This work presents a data analytic tool for clustering analysis based on Dimensionality Reduction (DR) of power system measurements. The proposed method is applied to frequency measurements of the ENTSO-E dynamic model of continental Europe and the results are compared with other conventional DR approaches. After considerable reduction of the raw measurements, a phasor metric for identification of coherency groups of generators is proposed. The recommended measure stands for its simple implementation, interpretation and fast computation. To illustrate the effectiveness of the clustering approach and the coherency of the metrics, a particular study case following the outage of a representative generation unit in France is presented.