HYBRID DEEP LEARNING MODEL FOR BYPASS DIODE FAULTS IN PHOTOVOLTAIC SYSTEMS

Creative Commons License

Akçam D. B., Özarslan Yatak M.

V. International BRITISH Congress on Interdisciplinary Scientific Research & Practices, London, İngiltere, 5 - 07 Şubat 2026, ss.705-710, (Tam Metin Bildiri)

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: London
  • Basıldığı Ülke: İngiltere
  • Sayfa Sayıları: ss.705-710
  • Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
  • Gazi Üniversitesi Adresli: Evet

Özet

The reliability, efficiency, and maintenance of photovoltaic (PV) systems have become increasingly

critical as installed capacity increases. Faults in PV arrays not only lead to production losses but also pose

serious safety risks, including hot-spot formation, module degradation, and fire risk. Especially, opencircuit

or short-circuit faults in bypass diodes, which serve as the module's protection mechanism, cause

permanent damage and efficiency losses by disrupting the module's electrical characteristics. Therefore,

early detection and accurate classification of bypass diode faults are fundamental requirements for the

sustainability of PV panels. This study proposes to classify the effects of bypass diode failures in PV panels

on the fundamental current-voltage and power-voltage characteristics. Different operating conditions and

failure scenarios were modelled in MATLAB/Simulink, and a dataset was generated from the resulting

data. Using the created dataset, a hybrid deep learning model based on 1DCNN- BiLSTM was developed.

The model's performance was evaluated. The results demonstrate that the proposed approach can

effectively distinguish bypass diode failures. The model developed in this study aims to contribute to early

failure detection and increased reliability in PV systems.