Data-Driven Identification of Nonlinear Frequency Dynamics in the South Korean Power Grid Using Symbolic Regression
ACM SIGEnergy Energy Informatics Review · 01.09.2026 · 6 (3) · 257-268
DOI: 10.1145/3861130.3861150 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM SIGEnergy Energy Informatics Review, online erschienen am 11.11.2026. Verfasst von Felix Wang und 3 weiteren Autor:innen. Die Arbeit nennt 33 Literaturangaben.
Abstract (Originalsprache)
The transition toward renewable energy sources and the resulting decentralization of power generation pose significant challenges to grid stability, particularly due to the reduction of system inertia. While traditional physics-based models provide a robust theoretical framework, they often rely on rigid structural assumptions that may not fully capture the complex dynamics of modern power grids. This paper investigates the potential of Symbolic Regression (SR) as a data-driven approach to discover the governing differential equations of frequency dynamics directly from measurement data, without presupposing specific equations. We systematically evaluate and compare three distinct SR paradigms: SINDy (sparse regression), PySR (genetic programming), and SVISE (Bayesian variational inference). We use synthetic data with known dynamics to validate the models' abilities to recover the governing control structure of a noisy linear baseline model, and then apply the methods to empirical 1 Hz frequency data from the South Korean grid covering 382 days. A consistent hierarchical pattern emerges across all methods in which terms involving higher powers of the angular velocity ω dominate the discovered dynamics, with ω 2 and ω 5 carrying the largest nonlinear coefficients. Mixed terms such as θω and θω 2 also appear consistently across methods, following the same ω -dominated ordering at each polynomial degree. This stability across three fundamentally different search strategies and noise handling approaches provides evidence for nonlinear frequency dynamics in the South Korean grid that are not captured by classical linear models. Motivated by this hierarchy, we propose a six-term extension of the aggregated swing equation for the South Korean grid that retains the linear control terms and adds a parsimonious set of dominant, nonlinear contributions. The proposed model improves on the linear baseline by roughly a factor of two in mean forward-simulation RMSE.
Zitieren
APA
Wang, F., Doan, M., Wen, X., & Schäfer, B. (2026). Data-Driven Identification of Nonlinear Frequency Dynamics in the South Korean Power Grid Using Symbolic Regression. ACM SIGEnergy Energy Informatics Review, 6(3), 257-268. https://doi.org/10.1145/3861130.3861150
Vancouver
Wang F, Doan M, Wen X, Schäfer B. Data-Driven Identification of Nonlinear Frequency Dynamics in the South Korean Power Grid Using Symbolic Regression. ACM SIGEnergy Energy Informatics Review. 2026;6(3):257-268. doi:10.1145/3861130.3861150
BibTeX
@article{Wang2026,
author = {Wang, Felix and Doan, Maily and Wen, Xinyi and Schäfer, Benjamin},
title = {Data-Driven Identification of Nonlinear Frequency Dynamics in the South Korean Power Grid Using Symbolic Regression},
journal = {ACM SIGEnergy Energy Informatics Review},
year = {2026},
volume = {6},
number = {3},
pages = {257--268},
doi = {10.1145/3861130.3861150}
}
EUVIMED