变电站图纸物理回路建模方法及实现  

Realization of physical circuit modeling method for substation drawings

作　　者：侯进[1,2] 王祥宇 郝彦超 文志龙 李昀喆 康萍萍 HOU Jin;WANG Xiangyu;HAO Yanchao;WEN Zhilong;LI Yunzhe;KANG Pingping(Laboratory of Intelligent Perception and Smart Operation&Maintenance,School of Information Science and Technology,Southwest Jiaotong University,Chengdu 611756,China;National Engineering Laboratory of Integrated Transportation Big Data Application Technology,Southwest Jiaotong University,Chengdu 611756,China;School of Tangshan,Southwest Jiaotong University,Tangshan 063000,China)

机构地区：[1]西南交通大学信息科学与技术学院智能感知智慧运维实验室,四川成都611756 [2]西南交通大学综合交通大数据应用技术国家工程实验室,四川成都611756 [3]西南交通大学唐山研究院,河北唐山063000

出　　处：《电力工程技术》2023年第1期193-200,共8页Electric Power Engineering Technology

基　　金：国家重点研发计划资助项目(2020YFB1711902);四川省科技计划资助项目(2020SYSY0016)。

摘　　要：目前变电站原始图纸信息保存单一、数字化解析程度不高,人工解析变电站原始图纸工作量庞大。为了解决该问题,提出基于可携带文档格式(portable document format,PDF)图纸智能化提取变电站设备及其连接关系信息、构建物理回路模型的方法,并根据智能变电站物理配置描述(substation physical configuration description,SPCD)文件对该方法进行具体实现。首先,对图纸中提取的图元信息进行处理;然后,通过字符串相似度匹配结合KR的贪心串覆盖(running Karp-Rabin greedy string tiling,RKR-GST)算法实现电气符号的识别,使用梯度提升决策树和逻辑回归(gradient boosting decision tree-logistics regression,GBDT-LR)混合算法进行变电站图像分类;最后,根据SPCD文件完成从原始图纸到物理回路模型的数字化描述。实验结果表明,在图元存在误差的情况下,电气符号匹配正确率达到93%,物理回路识别正确率超过90%。At present,the original drawing information of the substation is stored in a single way,and the degree of digital analysis is not high.In order to reduce the huge workload of manually analyzing the original drawing of the substation,a method is proposed to intelligently extract the information of the equipment and connection relationship of the substation based on the portable document format(PDF)drawing for building a physical circuit model,thus implementing the method according to the smart substation physical configuration description(SPCD)file.Firstly,the primitive information is extracted and processed from drawings.Then electrical symbol recognition is realized through string similarity matching running Karp-Rabin greedy string tiling(RKR-GST)algorithm,and substation images are classified by gradient boosting decision tree-logistic regression(GBDT-LR)hybrid algorithm based on features.Finally,the digital description from the original drawing to the physical circuit model is completed according to the SPCD file.Experiments show that the correct rate of electrical symbol matching is 93%,and the correct rate of physical circuit identification is more than 90%when there are errors in primitives.

关 键 词：智能变电站 二维图纸 配置描述文件 物理回路 图像处理 图纸活化 辅助设计 

分 类 号：TM933[电气工程—电力电子与电力传动] TP391[自动化与计算机技术—计算机应用技术]