DEPQD-F: A Deployable and Explainable Framework for Power Quality Disturbance Classification in Smart Grids
8th Global Power, Energy and Communication Conference, GPECOM 2026, Naples, İtalya, 3 - 05 Haziran 2026, ss.1133-1138, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/gpecom70462.2026.11578840
- Basıldığı Şehir: Naples
- Basıldığı Ülke: İtalya
- Sayfa Sayıları: ss.1133-1138
- Anahtar Kelimeler: CNN-Transformer, Edge Deployment, Explainable AI, Multi-Label Classification, Power Quality Disturbances, Smart Grid Monitoring
- Gazi Üniversitesi Adresli: Evet
Özet
Despite sustained advances in deep learning (DL)based power quality disturbance (PQD) classification, with reported accuracies now exceeding 99%, the field faces a critical translation gap: very few proposed systems have been validated under real-world conditions or demonstrated feasibility on constrained embedded hardware. This paper presents DEPQD-F (Deployable and Explainable PQD Framework), a modular, end-to-end framework that systematically addresses six evidencebased deployment gaps identified in the prior literature. DEPQDF integrates a noise-aware data pipeline, a lightweight hybrid CNN-Transformer feature extractor, a multi-label classification head for compound disturbances, an XAI explanation module based on Grad-CAM and attention visualization, an edge deployment pathway with latency benchmarking, and a standardized evaluation protocol aligned with IEEE Std 1159-2019. Each module directly targets a specific gap that has prevented laboratoryaccurate PQD classifiers from achieving operational deployment in smart grid environments. The framework is presented as a design blueprint for future experimental realization.