Deep Reinforcement Learning-based Path Planning Method for Transmission Lines

The rising demand for renewable energy gridconnection requires transmission lines to cross complex and ecologically sensitive terrains. Manual planning often fails to balance path efficiency, safety, and ecological impact. This paper proposes a deep reinforcement learning-based method for transmission line path planning. A 3D multi-objective tensor is constructed using terrain slope, vegetation cover, and ecological sensitivity data. Key innovations include the direction-sensitive TLA network with two-stage attention for feature fusion, the DESS module for search efficiency via spatial continuity, and the CRM for adaptive weighting of environmental constraints. The model supports safe, efficient, and eco-friendly grid planning under the “dual carbon” goals.