| As an important infrastructure,the substation undertakes the important responsibility of ensuring the demand for electricity,which is of great significance to the development of social economy.At present,there are a series of problems in substation secondary loop terminal block wiring inspection,including long inspection period,single verification means,scattered data management,and inability to accurately detect equipment hidden dangers.These problems have seriously affected the efficiency and accuracy of substation secondary loop inspection and brought potential risks and hidden dangers to substation operation and equipment safety.Therefore,in order to improve the efficiency of secondary loop inspection and the level of intelligent inspection,it is necessary to design a system that integrates the detection and recognition of design drawing text and the on-site verification of terminal block,so as to improve the efficiency of substation secondary loop connection verification,so as to solve the problems such as low efficiency of manual verification and non-negligible secondary loop information,so as to reduce the risk of substation operation.The research of this system mainly includes three parts: detection and recognition of terminal block design drawing and drawing text,on-site connection loop verification,model transplantation and system interface design.Specific research contents are as follows:1)Test and identify the text of terminal block design drawing.First,in order to identify the terminal block and connection line in the drawing,CBAM attention mechanism was added to the original YOLOv5 target detection algorithm,which improved the average detection accuracy of the algorithm to 95.5%.Secondly,if the drawing bar of terminals is not disassembled and OCR is directly used for text recognition,the error rate is very high.Therefore,the Open CV machine vision algorithm is used to split the terminal block and the connecting line,and the terminal block is divided into a terminal group starting with the terminal name,which improves the recognition accuracy of the drawing terminal block text.Finally,text in the drawings was identified using Paddle OCR optical character recognition technology with an accuracy of 93.2%.In general,the accuracy of text detection and recognition of design drawings by the above algorithm model reaches 88.0%.2)Check the on-site wiring circuit.First of all,because there are some characters that are difficult to identify in the actual wiring,the Paddle OCR text recognition technology is optimized by adopting the original recognition model for transfer learning.After retraining and learning,the text recognition accuracy of field pictures reached 79.4%.Secondly,by fetching the documents identified by the design drawing and comparing with the actual wiring situation,it can determine whether the wiring is correct and complete the verification of the wiring loop.Finally,in order to solve the text recognition errors caused by obscuring and blurring,a fuzzy matching method is proposed for connection verification,which improves the accuracy of connection verification to 79.5%.3)Transplant the model and design the system interface.Since the substation needs to check the wiring of the secondary loop terminal block at the work site,the above algorithm model is transplanted to the tablet computer.The system uses QT application development framework to design the operating interface of the system.It can directly call its own camera to take on-site photos,and can carry out text recognition and connection verification on the pictures.The system not only reduces the manual operation,but also meets the requirement of high recognition accuracy. |
PDF Full Text Request