Research on Tree Line Grounding Fault Identification Technology Based on Improved YOLOv8 Model

Aiming at the frequently occurring tree line ground fault recognition challenges in power systems, the article uses the improved YOLOv8 model for the study. By integrating the KAN neural network architecture and adopting the EIoU loss function to improve the bounding box prediction accuracy, the accuracy of the model is improved, which makes up for the accuracy limitation problem faced by the traditional YOLOv8 model in the power system image recognition task. The experimental results show that a significant increase of 1.7% in precision, 9% in recall, and 6.9% in average precision (mAP_0.5) is achieved while maintaining the original processing speed, which verifies the effectiveness of the method in improving the power system image data processing precision and fault feature recognition effectiveness.