EUVIMEDEUVIMED entdecken
Crossref

Learnable Transient Energy and Cross-Correlation Attention-Based Mamba for Fault Phase Selection in IoT-Enabled Distribution Networks

Huan Luo, Guowei Wang, Wei Sun, Weitao Li, Qiyue Li, Zhi Liu, Lijian Ding

IEEE Internet of Things Journal · 2026-11 · 13 (22) · 52293-52305

DOI: 10.1109/jiot.2026.3727554 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Internet of Things Journal, 2026-11. Verfasst von Huan Luo und 6 weiteren Autor:innen. Die Arbeit nennt 53 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Anhui Provincial Natural Science Foundation, 111 Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Fault phase selection is essential for protective relaying and active arc suppression in IoT-enabled distribution networks. In such systems, feeder terminal units continuously collect multichannel voltage and current waveforms, requiring accurate and robust phase selection. However, fault transients are often weak and nonstationary, making accurate phase selection challenging. To address these challenges, we propose NC-Mamba, which integrates learnable signal processing with Mamba, a selective state-space model (SSM) for sequence modeling. Specifically, we devise a Neural Instantaneous Energy Density Mechanism (NIEDM) that learns an adaptive transient energy operator to amplify subtle fault variations. We further introduce complex cross-correlation attention (CCA), which derives attention weights from complex cross-correlations of analytic voltage and current signals to fuse interphase magnitude and phase information. By embedding NIEDM and CCA into Mamba, NC-Mamba jointly encodes transient temporal dynamics and cross-phase dependencies. In addition, we release a real-world distribution network fault phase-selection dataset collected from field measurement devices. Experimental results under stratified fivefold cross-validation show that NC-Mamba achieves 96.2% $\pm ~0.8$ % accuracy and 94.1% $\pm ~1.1$ % Macro- $F1$ , outperforming both general time-series baselines and representative power-system-domain fault diagnosis models.

Förderung

  • National Natural Science Foundation of China
  • Anhui Provincial Natural Science Foundation
  • 111 Project

Zitieren

APA

Luo, H., Wang, G., Sun, W., Li, W., Li, Q., Liu, Z., & Ding, L. (2026). Learnable Transient Energy and Cross-Correlation Attention-Based Mamba for Fault Phase Selection in IoT-Enabled Distribution Networks. IEEE Internet of Things Journal, 13(22), 52293-52305. https://doi.org/10.1109/jiot.2026.3727554

Vancouver

Luo H, Wang G, Sun W, Li W, Li Q, Liu Z, et al. Learnable Transient Energy and Cross-Correlation Attention-Based Mamba for Fault Phase Selection in IoT-Enabled Distribution Networks. IEEE Internet of Things Journal. 2026;13(22):52293-52305. doi:10.1109/jiot.2026.3727554

BibTeX
@article{Luo2026,
  author = {Luo, Huan and Wang, Guowei and Sun, Wei and Li, Weitao and Li, Qiyue and Liu, Zhi and Ding, Lijian},
  title = {Learnable Transient Energy and Cross-Correlation Attention-Based Mamba for Fault Phase Selection in IoT-Enabled Distribution Networks},
  journal = {IEEE Internet of Things Journal},
  year = {2026},
  volume = {13},
  number = {22},
  pages = {52293--52305},
  doi = {10.1109/jiot.2026.3727554}
}
Kostenlos starten →