A new intelligent power quality disturbance classification in renewable and decentralized hydrogen-based energy systems using SwResNET hybrid model

Küçüker A., Baraklı B., Bayrak G., BAŞARAN K., Balaban G.

Renewable Energy, cilt.250, 2025 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 250
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.renene.2025.123251
  • Dergi Adı: Renewable Energy
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Aquatic Science & Fisheries Abstracts (ASFA), CAB Abstracts, Communication Abstracts, Compendex, Environment Index, Geobase, Greenfile, Index Islamicus, INSPEC, Pollution Abstracts, Public Affairs Index, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Anahtar Kelimeler: Hydrogen energy-based distributed generation, Power quality disturbances, Residual networks, Swin transformer learning, Vision transformers
  • Manisa Celal Bayar Üniversitesi Adresli: Evet

Özet

In this study, a scalogram image-based Swin-Residual Network (SwResNET) hybrid method is proposed for the identification of power quality disturbances (PQDs) in a hydrogen energy-based distributed generator (HEBDGs). The proposed approach involves the creation of PQD scalogram images by applying spectrogram analysis to power signal data. This process generates a two-dimensional image that represents the frequency and time characteristics of the signal. These spectrogram images are then input into a SwResNET hybrid model for learning. The SwResNET hybrid model extracts features from the scalogram images and classifies the input signal based on the presence or absence of power quality disturbances. This paper used 21 different PQD events in HEBDGs for classification purposes. Furthermore, the proposed method was tested under noisy conditions. The data achieved from simulated results of the HEBDG system in Matlab/Simulink and empirical data collected in the laboratory collectively demonstrate that the proposed methodology exhibits exceptional performance in terms of 98.22 % accuracy and resistance to noise, surpassing existing state-of-the-art approaches.