Online Low Frequency Oscillation Detection and Analysis System with an Ensemble Filter

Desong Bian, Zhe Yu, Di Shi, Ruisheng Diao, Zhiwei WangarXiv 2018 · DOI · arXiv

Abstract

The widespread deployment of phasor measurement unit (PMU) overpower systems makes it possible to monitor and analyze grid dynamics in real-time. Low-frequency oscillation is harmful to power system equipment and operation, and in the worst-case scenario may lead to cascading failures. Therefore, it is critical to detect and identify them as soon as they appear. This paper presents an online low-frequency oscillation detection and analysis (LFODA) system, which has the merit of significantly reducing the chance of false alarm via a voting schema and a time-serial filter. A novel algorithm based on density-based spatial clustering of applications with noise (DBSCAN) is proposed to classify oscillation modes as well as to group their corresponding buses/monitoring sites. Performance of the LFODA system is evaluated through experiments using both simulated and real-world PMU data.

Introduction

Modern power grids are evolving rapidly with the integration of renewable energy sources, electric vehicles, and micro-grids. While these advancements improve sustainability, they also increase grid complexity, making systems more susceptible to low-frequency oscillations (LFOs). These rhythmic swings in power can cause severe damage to generator shafts and, if unmitigated, lead to catastrophic cascading blackouts, such as the 1996 WSCC breakup.

To monitor these instabilities, grid operators rely on Phasor Measurement Units (PMUs). However, real-time monitoring faces two primary challenges: the immense computational burden of processing thousands of PMU channels and the high sensitivity of standard detection algorithms to signal noise, which frequently results in "false alarm" scenarios.

The LFODA System Architecture

The proposed Low-Frequency Oscillation Detection and Analysis (LFODA) system addresses these challenges through a multi-layered, robust monitoring framework. The core of the system is an ensemble filter that balances computational efficiency with detection accuracy.

The LFODA system architecture, illustrating the data preprocessing, tiered oscillation detection ensemble, and clustering stages.

Ensemble Detection and Voting Schema

The LFODA system utilizes a "Triple-Threat Ensemble" consisting of Prony’s Method, Hankel Total Least Square (HTLS), and a novel Extended Kalman Filter (EKF). To optimize performance, the system employs a sequential voting logic rather than running all algorithms in parallel. It begins with the computationally inexpensive Prony method. If no oscillation is detected, the process resets. If a potential event is flagged, the system triggers the more robust HTLS and EKF algorithms to confirm the finding. This tiered approach reduces total calculations by over 50%.

Noise Mitigation and Clustering

To further reduce false alarms caused by transient signal glitches, a time-series filter is implemented. This filter requires an oscillation to persist across consecutive time windows before an alert is triggered.

Once an oscillation is confirmed, the system must classify the affected areas. The researchers proposed using Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to group PMU channels by oscillation mode (local vs. inter-area). Unlike traditional methods, DBSCAN does not require a predefined number of clusters, making it ideal for large-scale geographic grids.

Visualization of the DBSCAN clustering approach used to group monitoring sites based on oscillation characteristics.

Experimental Results

The LFODA system was validated using both simulated WECC 179-bus models and real-world data from the State Grid Jiangsu Electric Power Company, monitoring over 1,000 channels.

  • Efficiency: The system processed 1-second data windows in under 700ms on standard hardware, meeting real-time requirements.
  • Reliability: In a two-year historical data test, the time-series filter reduced reported events from over 20 (mostly noise) to 10 legitimate, confirmed oscillations.
  • Field Performance: The system successfully identified real-world oscillations in the Jiangsu grid, providing damping factors and mode shapes consistent with historical records.

Conclusion

The LFODA system provides a scalable and reliable solution for real-time grid monitoring. By combining a tiered ensemble filter with unsupervised clustering, it significantly improves detection accuracy while halving computational costs. Future work will focus on hardening the system against sensor failures and integrating modules for precise oscillation source localization and cybersecurity defenses against malicious data injection.

Summary updated

Abstract

The widespread deployment of phasor measurement unit (PMU) overpower systems makes it possible to monitor and analyze grid dynamics in real-time. Low-frequency oscillation is harmful to power system equipment and operation, and in the worst-case scenario may lead to cascading failures. Therefore, it is critical to detect and identify them as soon as they appear. This paper presents an online low-frequency oscillation detection and analysis (LFODA) system, which has the merit of significantly reducing the chance of false alarm via a voting schema and a time-serial filter. A novel algorithm based on density-based spatial clustering of applications with noise (DBSCAN) is proposed to classify oscillation modes as well as to group their corresponding buses/monitoring sites. Performance of the LFODA system is evaluated through experiments using both simulated and real-world PMU data.
BibTeX
@misc{bian2018online,
  author = {Desong Bian and Zhe Yu and Di Shi and Ruisheng Diao and Zhiwei Wang},
  title = {{Online Low Frequency Oscillation Detection and Analysis System with an Ensemble Filter}},
  year = {2018},
  eprint = {1812.11266},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1812.11266},
}
Figure