摘要
特高压直流控制保护系统是特高压直流输电工程的“大脑”,控制保护系统设备采用分层分布式结构,不同层次、同一层次不同区域的装置之间均通过电缆、光纤进行大量的信号交换,构成了一套庞大而复杂的系统。此外,特高压直流控制保护系统设备的连接关系和信号交换等信息分散在不同载体上,信息检索工作量巨大。这些都使得特高压直流控制保护系统的运维检修难度大大增加。针对这些问题,提出基于深度学习的特高压直流控制保护系统图纸识别和可视化技术方案,采用自定义的控制保护系统图纸模型格式,将基于候选区域的快速卷积神经网络Faster R-CNN和光学字符识别Tesseract-OCR引擎结合,识别控制保护系统图纸中的信息并自动完成图纸建模,将模型存入数据库中,依托数据库便捷的查询方式,直观地展示控制保护系统的原理,提高数据检索效率。
The ultra-high voltage direct current(UHVDC)control and protection system is the brain of the UHVDC power transmission project,where the control and protection equipment uses hierarchical distribution structures.Massive signals are all exchanged through the fiber-optical cable among devices from different areas which are at different levels or at the same level,constituting a large and complex system.In addition,as the information of the connection relationship and signal exchange of the UHVDC control and protection system equipment are scattered on different carriers,the information retrieval workload is huge.All of these make the operation and maintenance of the UHVDC control and protection system more difficult.For this end,this paper proposes a drawing recognition and visual scheme for the UHVDC control and protection system based on deep learning.The scheme uses user-defined drawing model form of control and protection system,which combines the fast convolution neural network Faster R-CNN based on candidate districts and the optical character recognition Tesseract-OCR engine to identify the information of the control and protection system and complete the drawing modeling automatically,and the model is stored in the database,and the principle of the control and protection system is intuitively demonstrated based on the convenient query method of the database to improve the efficiency of data retrieval.
作者
孔祥平
李鹏
高磊
林俊
KONG Xiangping;LI Peng;GAO Lei;LIN Jun(Electric Power Research Institute of State Grid Jiangsu Electric Power Co.Ltd.,Nanjing 211103,Jiangsu,China;Jiangsu Future Smart Technology Co.,Ltd.,Nanjing 211100,Jiangsu,China)
出处
《电网与清洁能源》
2020年第2期29-37,共9页
Power System and Clean Energy
基金
国家电网公司科技项目(5210EF180013)。