3D Point Cloud Segmentation Based Large-Scale Transmission Line Wildfire-Induced Tripping Risk Assessment Technology and Application

• Published in: 2023 2nd International Conference on Clean Energy Storage and Power Engineering (CESPE) pp. 24 - 30
• Main Authors: Zhao, Jiguang, Zhang, Keying, Wu, Xinqiao, Liu, Lan, Qin, Ping, Lu, Mingxiang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 15.12.2023
• Subjects: deep learning; Hazards; Point cloud compression; point cloud segmentation; Power system stability; Power transmission lines; Three-dimensional displays; transmission line corridors; Voltage; wildfire risk assessment; Wildfires
• DOI: 10.1109/CESPE60923.2023.00016

In recent years, wildfires have occurred frequently, leading to increasing power line outages. Among these, outages of ultra-high voltage lines account for over 20%, severely affecting the power grid's safety and stability. Although extensive research has been conducted on the mechanisms of wildfire-induced line tripping, these delicate wildfire-tripping models have issues with parameter extraction and require strict parameter precision, which hinders their large-scale application. This paper proposes a large-scale wildfire hazard assessment method for transmission lines based on point cloud segmentation. Firstly, an improved local feature sampling mechanism based on unbalanced data characteristics is proposed to enhance RandLA-Net performance for transmission line point clouds by reinforcing the embeddings for categories with fewer samples. Then, based on the segmentation results, the parameters for wildfire hazard assessment are automatically extracted to establish a wildfire tripping risk hazard assessment model. Finally, a large-scale engineering verification is conducted on ultra-high voltage lines of 500kV and above in the Southern Power Grid. The verification results demonstrate the effectiveness of the proposed method.