14 June 2024 in IEEE Access
As renewable energy sources are integrated into power systems, the complexity of their management increases. For the analysis of modern power systems, clustering algorithms have become indispensable tools. The representation of an electric power transmission system involves conceptualizing it as a graph—a mathematical structure depicting a set of nodes and edges. In the context of power systems, these nodes correspond to buses, and the edges represent branches. Weighting may be assigned to the edges to denote the strength of their connections. In this paper, spectral clustering and hierarchical agglomerative clustering have been employed to analyze the Italian transmission network. Spectral clustering relies on the eigenvalues and eigenvectors derived from the Laplacian matrix associated with the network. The hierarchical agglomerative clustering, based on Thévenin impedance to define the distances, merges the two closest clusters at each step. The primary objective of this research is to identify distinct areas within the network based solely on its topology. In pursuit of this goal, admittances and Thévenin impedances have been employed as weights in the graph representation. By utilizing these approaches, the research aims at uncovering meaningful structures and patterns within the power transmission system based on the inherent connectivity and strength of the network elements. The techniques have been used on two real network models to demonstrate the performance of the algorithms with different grid configurations.
25 July 2025 in Sustainable Energy, Grids and Networks
25 July 2025 in Sustainable Energy, Grids and Networks
1 July 2026 in Electric Power Systems Research
5 November 2025 in TechRxiv
GGGiorgio Maria Giannuzzi
Monitoring electromechanical oscillations is critical for maintaining stability of modern power systems, especially in the presence of increasing penetration of inverter-based resources. Existing data-driven techniques, such as Dynamic Mode Decomposition (DMD) and the combination of Principal Component Analysis with Prony analysis (PP), have shown good performance during oscillatory event but their limited robustness under ambient conditions limit their applicability. To overcome these limitations, this paper evaluates three advanced signal processing approaches: Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT), Tufts-Kumaresan (TK), and Stochastic Subspace Identification (SSI). Furthermore, an energy-based triggering mechanism is also introduced to enhance situational awareness. All the approaches are benchmarked against existing methods on both simulated test networks and real-world oscillatory events recorded in the European power system. The robustness of modal parameter identification under ambient conditions and the reliability of the trigger alarm are validated through the analysis of a real short-circuit event and a 15-minute interval of normal grid operation. The results demonstrate that ESPRIT, TK, and SSI yield more accurate modal parameter estimates and trigger alarms earlier than DMD and PP, confirming the suitability of the proposed framework for real-time stability monitoring in modern power systems.