Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Data-Driven Method to Identify Major Contributors to Low-Frequency Oscillations

Youhong Chen, Debraj Bhattacharjee, Balarko Chaudhuri

TL;DR

This work introduces a purely data-driven approach to identify major contributors to poorly damped low-frequency oscillations in power systems by applying Extended Dynamic Mode Decomposition (EDMD) to PMU data at generation POIs. By constructing a reduced-order Koopman operator and computing data-driven participation factors, the method quantifies each plant's contribution to dominant oscillatory modes without requiring detailed system models, enabling effective mitigation strategies. The approach is validated on a 39-bus system with 100% IBRs, a WECC 179-bus system, and real ISO-New England PMU events, demonstrating accurate mode identification and source localization for both natural and forced oscillations. The framework offers a practical tool for planning-stage simulations and post-event analyses, with potential extensions to observables library design to handle high-nonlinearity dynamics.

Abstract

We present a purely data-driven method to pinpoint generation plants that significantly contribute to poorly damped oscillations as part of post-event analysis. First, Extended Dynamic Mode Decomposition (EDMD) is applied on PMU data from the point of interconnection (POI) of the plants to obtain the finite-dimensional Koopman operator. Then, modal analysis is performed on a reduced-order Koopman operator to extract spatio-temporal patterns. The data-driven eigenvalues and eigenvectors quantify each plant's contribution to critical oscillatory modes without requiring any system model information. We demonstrate the effectiveness of this method through simulated case studies on modified IEEE 39-bus and WECC 179-bus test systems by benchmarking the data-driven results against ground-truth models. Its performance is further validated using PMU data from real oscillation events in the ISO-New England system. This data-driven method offers a practical tool for both planning-stage simulations and post-event analysis of real oscillation events, enabling effective mitigation.

A Data-Driven Method to Identify Major Contributors to Low-Frequency Oscillations

TL;DR

This work introduces a purely data-driven approach to identify major contributors to poorly damped low-frequency oscillations in power systems by applying Extended Dynamic Mode Decomposition (EDMD) to PMU data at generation POIs. By constructing a reduced-order Koopman operator and computing data-driven participation factors, the method quantifies each plant's contribution to dominant oscillatory modes without requiring detailed system models, enabling effective mitigation strategies. The approach is validated on a 39-bus system with 100% IBRs, a WECC 179-bus system, and real ISO-New England PMU events, demonstrating accurate mode identification and source localization for both natural and forced oscillations. The framework offers a practical tool for planning-stage simulations and post-event analyses, with potential extensions to observables library design to handle high-nonlinearity dynamics.

Abstract

We present a purely data-driven method to pinpoint generation plants that significantly contribute to poorly damped oscillations as part of post-event analysis. First, Extended Dynamic Mode Decomposition (EDMD) is applied on PMU data from the point of interconnection (POI) of the plants to obtain the finite-dimensional Koopman operator. Then, modal analysis is performed on a reduced-order Koopman operator to extract spatio-temporal patterns. The data-driven eigenvalues and eigenvectors quantify each plant's contribution to critical oscillatory modes without requiring any system model information. We demonstrate the effectiveness of this method through simulated case studies on modified IEEE 39-bus and WECC 179-bus test systems by benchmarking the data-driven results against ground-truth models. Its performance is further validated using PMU data from real oscillation events in the ISO-New England system. This data-driven method offers a practical tool for both planning-stage simulations and post-event analysis of real oscillation events, enabling effective mitigation.

Paper Structure

This paper contains 18 sections, 23 equations, 11 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Extended Dynamic Mode Decomposition (EDMD)
  3. Proposed Methodology
  4. Selection of Observables
  5. Filtered Mode of Interest
  6. Calculation of Reduced-Order $\Tilde{M}_K$
  7. Data-Driven Modal Analysis
  8. Application Steps
  9. Case Study on Modified IEEE 39-bus System
  10. Scenario #1
  11. Scenario #4
  12. Case Study on WECC 179-bus Test System
  13. Natural Oscillation
  14. Forced Oscillation
  15. Case Study with Real PMU Data from ISO-New England
  16. ...and 3 more sections

Figures (11)

  • Figure 1: Modified IEEE 39-bus test system with all synchronous machines replaced by IBRs. Different mix of GFL and GFM are considered as summarized in Table \ref{['tab:gfl_gfm']}
  • Figure 2: EDMD-based analysis for Scenario # 1: (a) time-series $P$ and $Q$ at POI of each IBR plant, (b) data-driven participation factor of each IBR plant
  • Figure 3: EDMD-based analysis for Scenario # 4: (a) time-series $P$ and $Q$ at POI of each IBR plant, (b) data-driven participation factor of each IBR plant.
  • Figure 4: The scheme diagram of WECC 179-bus system. Maslennikov2016
  • Figure 5: EDMD-based analysis for Sim6ND: (a) time-series $P$ and $Q$ at POI of each SM generation plant, (b) data-driven participation factor of each SM.
  • ...and 6 more figures