Cable Fault Detection Based on Improved Deep Convolutional Neural Network

Background: The high-voltage cable is a critical component in power transmission systems, making regular inspections essential for the timely detection of potential hazards, schedule maintenance, and avoiding safety accidents. Objective: This paper aims to use deep learning algorithms to improve the precision and timeliness of cable fault detection, thereby ensuring safe and secure power system operation. Methods: Automatic cable fault detection based on YOLOv8s was conducted in the study in order to assist the power sector in automatically detecting cable faults. Results: PConv and BiFPN networks were added to the backbone network to improve the feature fusion performance of the model. To enhance the model's identification capabilities, the WIoU loss function was modified. Conclusion: The proposed method allows for the rapid detection of cable faults by analyzing three common fault types: "thunderbolt," "wear," and "break." By deploying this approach on edge computing devices mounted on UAVs, automatic inspection of power faults can be effectively achieved.