Akademik Veri Yönetim Sistemi
JOURNAL OF ELECTRICAL ENGINEERING & TECHNOLOGY, 2025 (SCI-Expanded)
The detection and classification of high-voltage (HV) equipment in electrical substations is essential for effective condition monitoring and fault prevention in power systems. Traditional visual-based methods often struggle under adverse lighting or environmental conditions, reducing their reliability in critical scenarios. To overcome these limitations, this study leverages infrared (IR) imaging, which provides stable thermal signatures even in poor visibility, offering a robust alternative for equipment identification. The key contribution of this work is a hybrid feature fusion strategy that integrates deep learning and traditional machine learning techniques. High-level semantic features are extracted via a lightweight deep neural network SqueezeNet, while low-level handcrafted features are derived using scale-invariant feature transform. These complementary features are fused and classified using a stacking ensemble method, enhancing both robustness and accuracy. The proposed approach was validated on IR imagery and achieved a high classification accuracy of 99.44% across five HV equipment types: circuit breakers, power transformers, surge arresters, disconnectors, and wave traps. This study advances the current literature by focusing on AI-based identification of HV equipment using IR imaging, addressing an underexplored yet practically relevant area. The developed system offers a scalable and technically viable solution for automated monitoring in real-world substation environments.