Data-Driven Post-Event Analysis with Real-World Oscillation Data from Denmark

TL;DR

This paper demonstrates that Extended Dynamic Mode Decomposition (EDMD) can accurately identify the main contributor to power grid oscillations using real-world PMU data, outperforming traditional methods in a post-event analysis setting.

Contribution

The study introduces the application of EDMD grounded in Koopman operator theory for post-event analysis of power grid oscillations using real PMU data, highlighting its accuracy and effectiveness.

Findings

EDMD accurately identified the oscillation source in Denmark.

Conventional methods like DEF did not clearly identify the source.

Validation with real data confirms EDMD's potential for targeted mitigation.

Abstract

This paper demonstrates how Extended Dynamic Mode Decomposition (EDMD), grounded in Koopman operator theory, can effectively identify the main contributor(s) to oscillations in power grids. We use PMU data recorded from a real 0.15 Hz oscillation event in Denmark for post-event analysis. To this end, the EDMD algorithm processed only voltage and current phasors from nineteen PMUs at different voltage levels across the Danish grid. In such a blind-test setting with no supplementary system information, EDMD accurately pinpointed the location of the main contributor to the 0.2 Hz oscillation, consistent with the location of the problematic IBR plant later confirmed by Energinet, where the underlying cause was a control system issue. Conventional approaches, such as the dissipating energy flow (DEF) method used in the ISO-NE OSL tool did not clearly identify this plant. This joint validation with Energinet, reinforcing earlier studies using simulated IBR-dominated systems and real PMU data from ISO-NE, highlights the potential of EDMD-based post-event analysis for identifying major oscillation contributors and enabling targeted SSO mitigation.