Scholarlink

<p>To address the challenge of analyzing low-frequency oscillations in power systems caused by the high penetration of renewable energy, this paper proposes an integrated GFC-FD (Gaussian Field-Chern Class-based Frequency Domain Dynamics) algorithmic framework. This framework first introduces a double-tree complex wavelet Gaussian mixture model (DT-GMM) based on Riemannian manifolds to robustly extract multiscale time-frequency features of oscillation signals. Second, a frequency-domain dynamic fractional-order fusion (FreDF) component is designed to transform the features into the frequency domain and combine them with fractional-order calculus for dynamic prediction, thereby enhancing the sensitivity and accuracy of oscillation detection. Finally, an energy dissipation analysis method based on canonical field theory is established. By abstracting the power grid as a canonical field and utilizing topological invariants such as the Chern characteristic class, the method achieves physically precise localization of the origin of oscillation energy and its key propagation paths. Simulation results on the IEEE 39-node standard system with a high proportion of renewable energy sources demonstrate that the proposed framework significantly outperforms comparison methods in key metrics such as pattern recognition accuracy, oscillation detection sensitivity, and energy path localization accuracy. Furthermore, with a total computation time of approximately 4.23&#xa0;s, the framework demonstrates the potential to address real-world engineering challenges.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Study on Oscillation Feature Extraction, Detection, and Localization Based on the GFC-FD Algorithm

  • 期刊论文

  • 发表日期：

    2026年8月1日

查看全文
  • Miao Yu,
  • Chenxuan Han

摘要

To address the challenge of analyzing low-frequency oscillations in power systems caused by the high penetration of renewable energy, this paper proposes an integrated GFC-FD (Gaussian Field-Chern Class-based Frequency Domain Dynamics) algorithmic framework. This framework first introduces a double-tree complex wavelet Gaussian mixture model (DT-GMM) based on Riemannian manifolds to robustly extract multiscale time-frequency features of oscillation signals. Second, a frequency-domain dynamic fractional-order fusion (FreDF) component is designed to transform the features into the frequency domain and combine them with fractional-order calculus for dynamic prediction, thereby enhancing the sensitivity and accuracy of oscillation detection. Finally, an energy dissipation analysis method based on canonical field theory is established. By abstracting the power grid as a canonical field and utilizing topological invariants such as the Chern characteristic class, the method achieves physically precise localization of the origin of oscillation energy and its key propagation paths. Simulation results on the IEEE 39-node standard system with a high proportion of renewable energy sources demonstrate that the proposed framework significantly outperforms comparison methods in key metrics such as pattern recognition accuracy, oscillation detection sensitivity, and energy path localization accuracy. Furthermore, with a total computation time of approximately 4.23 s, the framework demonstrates the potential to address real-world engineering challenges.