An Enhanced Geometric Featured Extraction Method for Real-time Oscillation Detection in Power System

This paper proposes a real-time oscillation monitoring framework based on an enhanced Geometric Feature Extraction (GFE) algorithm. The method utilizes high-resolution synchrophasor measurements from Phasor Measurement Units (PMUs) and employs a sliding window approach to identify oscillatory behavior in power systems. To improve detection robustness, the algorithm integrates signal-to-noise ratio (SNR) filtering and adaptive estimation of the Radius of Trajectory (RoT), which characterizes geometric patterns in reconstructed phase-space trajectories. A real-time alarming index is computed for each sample, and an oscillation alarm is triggered when the proportion of oscillatory samples within a window exceeds a predefined threshold. The proposed method was validated using actual PMU data recorded during a forced oscillation event caused by a Thyristor-Controlled Series Capacitor (TCSC) in the KEPCO power system. Experimental results demonstrate that the algorithm accurately captures the initiation, persistence, and attenuation of oscillatory behavior, offering both high detection sensitivity and low computational overhead suitable for real-time deployment.