Data-driven koopman operator-based detection and location for multi-source forced oscillations

• The forced oscillations were detected using the calculated phase trajectories of the eigenfunctions of the Koopman operator. • A modal contribution factor, derived from the Koopman operator, was developed to locates the oscillation source. • A data-driven framework for forced oscillation detection and source location was proposed, leveraging the Koopman operator. The identification and localization of oscillation sources (OSs) is crucial for the effective suppression of forced oscillations in power systems. Herein, we introduce a novel data-driven methodology, leveraging the Koopman operator, for the detection and localization of forced oscillations. The proposed method begins by identifying forced oscillations based on their distinct time-domain characteristics via phase trajectories. A Koopman operator-based modal contribution factor is then defined to quantify the sensitivity of each mode to state variables. The Koopman operator is extracted from measured dynamic data via sparsity promoting dynamic mode decomposition (SpDMD), and the OS locations are determined by computing the modal contribution factor. The accuracy of this method is validated via simulation studies and tests with real-world data, demonstrating its applicability to multisource scenarios.