Ultraviolet Imaging of External Insulation Discharge of Electrical Equipment Based on Diffusion Model Detection

An ultraviolet imaging method for the automatic detection of external insulation discharge in electrical equipment is proposed, based on a diffusion model. This study aims to enhance detection accuracy and efficiency by designing an anomaly detection framework that includes a training model, a fine‐tuning model, and an inference model. By incorporating an unsupervised learning diffusion model, the method effectively addresses the issue of rare anomalies, achieving efficient and accurate anomaly detection and localization. Experimental results demonstrate that the proposed method excels in both image‐level and pixel‐level anomaly detection, significantly outperforming existing methods and providing reliable assurance for the safe operation of electrical equipment. This study highlights the practical significance of using deep learning models for automatic ultraviolet imaging detection, saving substantial manpower and resources. The method also offers the advantage of high sensitivity and a broad detection range, making it suitable for both day and night operations without interference from background light. Additionally, the implementation of a multi‐level comparative analysis further improves the accuracy and reliability of anomaly detection, ensuring the effective identification and maintenance of potential issues in electrical equipment. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.