基于有权无向图同构的电气图元件识别方法研究  

Research on Electrical Component Recognition Method Based on Weighted Undirected Graph Isomorphism

作　　者：李红[1] 曹添植 韩少峰 何青[1] LI Hong;CAO Tianzhi;HAN Shaofeng;HE Qing(School of Energy,Power and Mechanical Engineering,North China Electric Power University,Beijing 102206,China)

机构地区：[1]华北电力大学能源动力与机械工程学院,北京102206

出　　处：《电力与能源》2025年第1期27-32,共6页Power & Energy

摘　　要：在新建变电站时,需要手动输入变电站物理配置描述(SPCD)等文件,这往往需要耗费大量的时间与人力。针对变电站二次系统电气原理图中信息的提取需求,以及电气图纸活化的同构算法存在的计算量大和存储数据多的问题,提出了一种利用有权无向图邻接矩阵特征值判断电气元件拓扑图同构的方法。首先对电气图元进行分割,计算电路线的长度,并通过分析电路线之间的相交关系,得到一个包围电气元件的矩形区域;然后提取该区域内的基本图元,并利用几何算法得到基本图元的连接关系,赋予邻接矩阵元素相应的权重,从而得到图的邻接矩阵;最后计算邻接矩阵的特征值,并通过特征值对比进行图匹配,从而完成电气元件的识别。通过改进绘图交换文件(DXF)结果的提取模块和数据的存储模块,提高了识别准确率。对DXF格式的电气图进行识别测试表明,所提出的算法不仅降低了现有同构算法的计算复杂度,减少了数据存储量,而且识别精度达到了87%。所提出的算法在计算复杂程度以及存储等方面均优于乌尔曼算法。利用该算法的提取结果,编写生成SPCD等文件的程序,可大大减少人工投入,促进变电站二次系统的智能化。When constructing a new substation,files such as SPCD need to be manually input,which often consumes a large amount of time and effort.By extracting information from the electrical schematic diagrams of the substation's secondary system,and thd fact that the existing isomorphism algorithms for activating electrical drawings face issues of large computational load and extensive data storage.This study proposes a method for recognizing electrical component topological graphs using the eigenvalues of the adjacency matrix of a weighted undirected graph.The method first segments the electrical components,calculates the length of circuit lines,and analyzes the intersection relationships between electrical lines to obtain a rectangular region surrounding the electrical components.The basic elements within this region are then extracted,and geometric algorithms are used to determine the connection relationships between the basic elements.The adjacency matrix is assigned weights for its elements to generate the graph's adjacency matrix.Finally,eigenvalues of the adjacency matrix are computed,and graph matching is performed using these eigenvalues to complete the recognition of electrical components.By improving the extraction module of the DXF file and the data storage module,the recognition accuracy is enhanced.Tests on electrical diagrams in DXF file format show that the proposed method not only reduces the computational complexity of existing isomorphism algorithms but also minimizes data storage requirements,with a recognition accuracy of 87%.The proposed algorithm outperforms the Ullmann algorithm in terms of computational complexity and storage,and the extracted results can be used to generate SPCD files,reducing manual input and promoting the intelligence of secondary systems in substations.

关 键 词：电气图纸 图纸活化 图同构 元件识别 邻接矩阵 合同变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]