In the maintenance of power transmission systems, wire conductor inspection plays a crucial role in the early detection of abnormalities such as power outages and short circuits, thereby ensuring transmission quality. Drones equipped with LiDAR (Light Detection and Ranging) can survey target areas and generate 3D models of wire conductors, revealing sagging lines and potential tree intrusion into transmission corridors. While conventional LiDAR data is unclassified, common objects such as ground and buildings can be automatically classified using machine learning algorithms. However, power transmission lines lack such automated classification tools and are typically labeled manually. In this study, a deep learning network based on the RandLA-Net architecture was developed and trained to automatically segment power transmission lines in large-scale LiDAR point clouds, replacing manual methods that are inadequate for big data processing. RandLA-Net utilizes LiDAR point attributes such as intensity and elevation to identify and classify points corresponding to the power transmission line class. The training and evaluation process yielded promising results, with Precision, Accuracy, and F1-Score all exceeding 0.95.