EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Deep Learning Approach to Power Quality Monitoring for Energy Efficiency and Carbon Reduction

Kevser İrem Danaci, Sıtkı Akkaya

Global Decarbonisation · 22.05.2026 · 2 · 111-142

DOI: 10.65582/gd.2026.007 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Global Decarbonisation, online erschienen am 22.05.2026. Verfasst von Kevser İrem Danaci und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

A classification system using deep learning for detecting power quality disturbances (PQDs), which are commonly experienced in electric power systems, has been proposed in this study. Because reliable and efficient classification methods are becoming increasingly critical as new forms of energy are integrated into the electric grid and as smart grid technology advances, accurate detection of PQDs will significantly contribute to the stability, efficiency, and robustness of current electric power systems. The SEED power quality disturbance dataset has been used for analysis; it contains 17 different types of disturbances, including pure sinusoidal signals. Both the raw 1D time domain signals and the 2D time frequency representations have been used to train the models (the time-frequency representations were created with the Short Term Fourier Transform, or STFT). The models used were convolutional neural network (CNN) architectures adapted for both 1D and 2D input data; both types of architectures were compared using accuracy, precision, and recall as measures of performance. The experimental results of the study show that for test accuracy, the 1D CNN model performed better than the 2D CNN model (i.e., 95% versus 82%, respectively). This research differs from most previous work because it simultaneously examines a common set of time domain and time-frequency feature representations while keeping all other test conditions constant. Results suggest that temporal representations of raw PQ signals may provide a classification advantage over time-frequency representations when used for PQD classification. Such benefits associated with using temporal representations will aid in the design of effective and accurate PQ monitoring systems, which, together with improved efficiencies from energy conservation, can help contribute towards global carbon emission reduction objectives by both reducing and eliminating wasted energy based on accurate and timely detection of power quality problems using intelligent monitoring techniques in modern smart grids.

Zitieren

APA

Danaci, K. İ., & Akkaya, S. (2026). A Deep Learning Approach to Power Quality Monitoring for Energy Efficiency and Carbon Reduction. Global Decarbonisation, 2, 111-142. https://doi.org/10.65582/gd.2026.007

Vancouver

Danaci Kİ, Akkaya S. A Deep Learning Approach to Power Quality Monitoring for Energy Efficiency and Carbon Reduction. Global Decarbonisation. 2026;2:111-142. doi:10.65582/gd.2026.007

BibTeX
@article{Danaci2026,
  author = {Danaci, Kevser İrem and Akkaya, Sıtkı},
  title = {A Deep Learning Approach to Power Quality Monitoring for Energy Efficiency and Carbon Reduction},
  journal = {Global Decarbonisation},
  year = {2026},
  volume = {2},
  pages = {111--142},
  doi = {10.65582/gd.2026.007}
}
Kostenlos starten →