Research on Partial Discharge Identification Method Based on Ultraviolet Images

The Ultraviolet (UV) radiation generated by partial discharge (PD) of power equipment has the advantages of noncontact and anti-interference, which is suitable for insulation state monitoring in complex power environment. This paper focuses on the automatic identification of UV images, constructs a tip discharge model, obtains the electric field distribution in the discharge area through COMSOL simulation, and analyzes the positional characteristics of the discharge source by combining with the UV imaging principle. On this basis, a lightweight convolutional neural network (CNN) structure is designed to complete the binary classification recognition of UV images. The experimental results show that the method can realize high recognition accuracy while maintaining low model complexity, providing technical support for UV image-assisted local discharge detection.