[Paper Review] Data-Driven Diagnostics of Mechanism and Source of Sustained Oscillations

Xiaozhe Wang, Konstantin Turitsyn|arXiv (Cornell University)|Jun 2, 2015

Power System Optimization and Stability33 references3 citations

TL;DR

This paper proposes a data-driven method to diagnose the mechanism and source of sustained oscillations in power systems using PMU measurements. By identifying distinct statistical signatures—particularly kurtosis and power spectral density—across weakly damped oscillations, limit cycles, and forced oscillations, the method accurately classifies oscillation types in real time, even at low SNR, enabling targeted control actions to restore system stability.

ABSTRACT

Sustained oscillations observed in power systems can damage equipment, degrade the power quality and increase the risks of cascading blackouts. There are several mechanisms that can give rise to oscillations, each requiring different countermeasure to suppress or eliminate the oscillation. This work develops mathematical framework for analysis of sustained oscillations and identifies statistical signatures of each mechanism, based on which a novel oscillation diagnosis method is developed via real-time processing of phasor measurement units (PMUs) data. Case studies show that the proposed method can accurately identify the exact mechanism for sustained oscillation, and meanwhile provide insightful information to locate the oscillation sources.

Motivation & Objective

• To address the critical challenge of distinguishing between three key oscillation mechanisms—weakly damped oscillations, limit cycles from Hopf bifurcations, and forced oscillations—in power systems.
• To overcome the limitations of conventional waveform analysis, which cannot differentiate mechanisms despite similar time-series appearances.
• To develop a real-time, data-driven diagnostic framework using PMU measurements to identify the root cause of sustained oscillations.
• To provide actionable insights for locating oscillation sources, enabling appropriate countermeasures such as PSS tuning, operational reset, or disturbance isolation.

Proposed method

• Develops a unified mathematical framework to model all three oscillation mechanisms: weakly damped oscillations as modulated sinusoids, limit cycles as deterministic periodic signals with noise, and forced oscillations as external periodic excitation.
• Derives analytical expressions for statistical signatures—specifically kurtosis and power spectral density—under each model to enable discrimination between mechanisms.
• Uses kurtosis as a primary discriminator: weakly damped oscillations exhibit high kurtosis due to amplitude modulation, while limit cycles and forced oscillations show lower, distinct kurtosis values.
• Employs power spectral density to further differentiate between limit cycles (broadband, nonlinear features) and forced oscillations (sharp, discrete peaks at excitation frequency).
• Applies the method in real-time to PMU data streams, enabling online diagnosis without requiring full system model knowledge.
• Validates the method using numerical case studies on a modified PSAT test system, demonstrating robustness under low signal-to-noise ratios.

Experimental results

Research questions

• RQ1 Can statistical signatures such as kurtosis and power spectral density reliably distinguish between weakly damped oscillations, limit cycles, and forced oscillations in PMU data?
• RQ2 How does the proposed method perform in identifying oscillation mechanisms when signal-to-noise ratio is low and time-series waveforms appear similar across mechanisms?
• RQ3 Can the method provide actionable information for locating the source of oscillations, particularly for forced oscillations?
• RQ4 To what extent does the method outperform traditional waveform analysis techniques in diagnosing oscillation mechanisms?
• RQ5 How robust is the diagnosis framework under realistic system conditions with noise and measurement uncertainty?

Key findings

• The proposed method accurately identifies the oscillation mechanism—weakly damped, limit cycle, or forced—using only PMU data, even at low signal-to-noise ratios.
• Kurtosis values differ significantly across mechanisms: weakly damped oscillations show high kurtosis due to amplitude modulation, while limit cycles and forced oscillations exhibit lower, distinguishable kurtosis levels.
• Power spectral density analysis successfully differentiates limit cycles (broadband, nonlinear content) from forced oscillations (sharp, discrete spectral lines).
• The method enables source localization by detecting phase coherence and frequency alignment with specific system components, particularly effective for forced oscillations.
• Case studies on a modified PSAT test system demonstrate high diagnostic accuracy under realistic conditions, including noise and load fluctuations.
• The framework is computationally efficient and suitable for real-time deployment in wide-area monitoring systems.