Research on Spatial Information Supplementation for Substation Main Wiring SCD Based on Image Recognition

To improve the layout completeness and automatic drawing capability of substation main wiring diagrams, this paper proposes a method for extracting the positional layout of key components from substation connection drawings, aiming to supplement incomplete spatial attributes in Substation Configuration Description (SCD) models. The proposed method leverages a You Only Look Once version 9 (YOLOv9)-based deep learning framework to detect critical graphical elements such as main transformers, busbars, and bays, and extracts their spatial coordinates from engineering drawings. A structured spatial descriptor is designed to represent each detected object, enabling automatic integration with SCD files. Experimental results on a hybrid dataset consisting of real and synthetic 500 kV main wiring diagrams demonstrate a high detection accuracy (mAP@0.5 > 90%) and a spatial recovery rate of 96.4% in typical SCD structures. The method effectively enhances the completeness and usability of SCD models in substation layout systems.