:At present, 3D virtual reality technology for substations is one of the important research directions in the modern power industry. 3D modeling technology is a very important part, especially in buildings with special requirements such as substations. Surveying and mapping provides great convenience. However, three-dimensional modeling technology has the disadvantages that it cannot directly model the two-dimensional image of the equipment, and the scene modeling volume is too large, it is difficult to quickly establish a single-region equipment model, and the complexity of on-site surveying and mapping is large. Based on this, explored and constructed a deep learning 3D modeling system suitable for substation equipment operation and maintenance management, and improved Yolov5 visual inspection algorithm.The improved algorithm has a detection accuracy of 97.5% and a detection rate of 156fps, realizing a 3D modeling system Quick detection of device types in The three-dimensional modeling system based on intelligent detection realizes automatic identification of types and rapid modeling, and intelligent linkage with auxiliary equipment monitoring system and lidar, providing an intuitive and interactive monitoring man-machine interface for substation operation and maintenance personnel.