Measurement-based coherency detection through Monte Carlo Consensus Clustering

This paper proposes a novel data-driven algorithm for measurement-based coherency detection in power systems, which is based on Monte Carlo Consensus Clustering (M3C). Unlike the clustering techniques conventionally adopted for coherency detection, M3C automatically identifies the optimal and stable number of coherent groups, despite the time-varying phenomena affecting power system operation. The proposed methodology is tested and validated with the IEEE 39-bus system. Results are compared with other existing clustering techniques, where the Friedman test with post-hoc analysis is performed on returned clustering scores to assess the statistically significant difference between the clustering techniques. This comparison highlights the advantages of the proposed approach, which does not require time-consuming analysis aimed at preliminary tuning and adjourning the algorithm parameters. •Coherency detection requires the number of clusters as input, reducing automation capabilities.•The number of coherent areas in power systems is not fixed.•Monte Carlo Consensus Clustering automatically returns the best number of clusters.•Monte Carlo Consensus Clustering is applied to coherency detection.•The number and composition of coherent areas is accurately estimated.