A Deep Learning-Based Substation Siting Assistance Method

With the acceleration of urbanization and industrialization, the structure of power grids has become increasingly complex, raising the demand for scientific and intelligent substation siting methods. Traditional approaches rely heavily on expert knowledge and rule-based reasoning, which struggle to fully exploit spatial information from remote sensing imagery<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[3][4]</sup>. This paper proposes SSA-Net (Substation Siting Assistance Network), a deep learning-based method for substation siting. SSA-Net consists of three core modules: Buildable Area Segmentation, Power Facility Detection, and Candidate Region Evaluation. By integrating spatial and semantic features of remote sensing images, the model generates a heatmap reflecting site suitability, providing a decision-making reference for planners. Experiments conducted on remote sensing images of a prefecturelevel city demonstrate the model’s generalization and practicality, with superior semantic segmentation accuracy and spatial localization compared to traditional models.