EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Performance Comparison of GRU and CNN-1D Model for Predicting Rain-Induced Ku-Band VSAT Link Degradation

Fahmi Khairul Ichsan, Makhsun, Kahfi Heryandi Suradija, R Yudha Mardyansyah, Naufal Ananda

Jurnal JEETech · 09.11.2026 · 7 (2) · 252-265

DOI: 10.32492/jeetech.v7i2.7207 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Jurnal JEETech, online erschienen am 09.11.2026. Verfasst von Fahmi Khairul Ichsan und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ku-Band Very Small Aperture Terminal links in tropical regions were vulnerable to rain-induced attenuation, which reduced the signal-to-noise ratio and could interrupt communication services. This study compared a one-dimensional convolutional neural network and a gated recurrent unit for predicting the maximum attenuation of individual rainfall events on an operational satellite link in Sorong, Indonesia. The 2-minute signal-to-noise ratio measurements were aggregated over every five consecutive samples to match the 10-minute rainfall resolution, after which both datasets were merged using identical timestamps. A dynamic seven-day 95th-percentile signal baseline was constructed, attenuation was derived from the baseline deficit, and rainfall events were detected using a 0.1 mm threshold with a 30-minute dry-gap tolerance. Sequence and event-level features were chronologically divided into training, validation, and testing subsets. Eight combinations of sequence length and pre-event context were evaluated. Both models achieved strong generalization. The best gated recurrent unit configuration, using 24 event steps and 12 pre-event steps, produced a validation R-squared of 0.9895 and a test R-squared of 0.9832, with a root mean square error of 0.1056 dB and a mean absolute error of 0.0697 dB. Its attenuation-severity classification accuracy reached 94.48%, compared with 91.41% for the best convolutional model. The results indicated that gated temporal memory more consistently represented the event-based dynamics of tropical rain attenuation and provided a suitable foundation for an operational early-warning system.

Zitieren

APA

Ichsan, F. K., Makhsun, Suradija, K. H., Mardyansyah, R. Y., & Ananda, N. (2026). Performance Comparison of GRU and CNN-1D Model for Predicting Rain-Induced Ku-Band VSAT Link Degradation. Jurnal JEETech, 7(2), 252-265. https://doi.org/10.32492/jeetech.v7i2.7207

Vancouver

Ichsan FK, Makhsun, Suradija KH, Mardyansyah RY, Ananda N. Performance Comparison of GRU and CNN-1D Model for Predicting Rain-Induced Ku-Band VSAT Link Degradation. Jurnal JEETech. 2026;7(2):252-265. doi:10.32492/jeetech.v7i2.7207

BibTeX
@article{Ichsan2026,
  author = {Ichsan, Fahmi Khairul and Makhsun and Suradija, Kahfi Heryandi and Mardyansyah, R Yudha and Ananda, Naufal},
  title = {Performance Comparison of GRU and CNN-1D Model for Predicting Rain-Induced Ku-Band VSAT Link Degradation},
  journal = {Jurnal JEETech},
  year = {2026},
  volume = {7},
  number = {2},
  pages = {252--265},
  doi = {10.32492/jeetech.v7i2.7207}
}
Kostenlos starten →