EUVIMEDEUVIMED entdecken
Crossref

Wavelet-Transform and Convolutional Neural Network Based Fault Detection and Classification in Smart Distribution Grids with High DER Penetration

Prajapati Jignasha A Prajapati Jignasha A

International Journal of Scientific Research in Science Engineering and Technology · 10.10.2023 · 398

DOI: 10.32628/ijsrset23587119 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Scientific Research in Science Engineering and Technology, online erschienen am 30.10.2023. Verfasst von Prajapati Jignasha A Prajapati Jignasha A. Die Arbeit nennt 23 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Rapid penetration of distributed energy resources (DER) reshapes fault behaviour in distribution feeders. Bidirectional power flow, low fault current contribution from inverter-based generators and increasingly frequent islanded operation degrade the dependability of conventional overcurrent protection. This paper presents a two-stage fault detection and classification scheme that combines the discrete wavelet transform (DWT) for feature extraction with a one-dimensional convolutional neural network (1-D CNN) for classification. Per-phase current signals are decomposed up to four levels using the Daubechies-4 mother wavelet; the resulting detail coefficients form a compact time-frequency signature that emphasises the high-frequency transient at fault inception. The signature feeds a 1-D CNN with three convolutional blocks and a softmax head over 11 classes (no-fault, three single line-to-ground faults, three line-to-line faults, three double line-to-ground faults and one three-phase fault). The method is trained and tested on a modified IEEE 13-node feeder simulated in PSCAD with photovoltaic, wind and battery storage at four nodes. Across 21,180 fault and no-fault snapshots the proposed scheme attains a test accuracy of 99.05% at 30 dB signal-to-noise ratio (SNR) and 95.4% at 15 dB SNR, outperforming wavelet-based SVM, ANN and random-forest baselines by 4–6 percentage points across the noise range and detecting faults within 4.2 ms of inception on average. The approach is robust to varying fault resistance (0.1–50 Ω), fault inception angle and DER output level.

Zitieren

APA

A, P. J. A. P. J. (2023). Wavelet-Transform and Convolutional Neural Network Based Fault Detection and Classification in Smart Distribution Grids with High DER Penetration. International Journal of Scientific Research in Science Engineering and Technology, 398. https://doi.org/10.32628/ijsrset23587119

Vancouver

A PJAPJ. Wavelet-Transform and Convolutional Neural Network Based Fault Detection and Classification in Smart Distribution Grids with High DER Penetration. International Journal of Scientific Research in Science Engineering and Technology. 2023:398. doi:10.32628/ijsrset23587119

BibTeX
@article{A2023,
  author = {A, Prajapati Jignasha A Prajapati Jignasha},
  title = {Wavelet-Transform and Convolutional Neural Network Based Fault Detection and Classification in Smart Distribution Grids with High DER Penetration},
  journal = {International Journal of Scientific Research in Science Engineering and Technology},
  year = {2023},
  pages = {398},
  doi = {10.32628/ijsrset23587119}
}
Kostenlos starten →