Substation One-Line Diagram Automatic Generation Based On Image Recongnition

With the development of smart grids, lots of technologies are applied to solve the problems of power systems. The substation one-line diagram automatic generation is the core of power systems. Currently, the traditional methods are not suitable for solving problems. In this paper, we introduce the newly deep learning technologies to generate substation one-line diagram automatically. Firstly, our algorithm adopts YOLOV3 method to detect symbols, which is a multiscale object detection algorithm, it is very effective for small objects like the Disconnector. Secondly, the PPOCR method is used to detect position and content of text, it can recognize various numbers and letters. At last, the Rooted Tree Algorithm and Hightower Algorithm are introduced to generate layout and topological relationship between symbols. Extensive experiments prove that our method obtains the best performance, the symbol recognition accuracy can reach 98%, and the text recognition accuracy can reach 97%, the accuracy of circuit diagram regeneration and construction is effectively improved.