Design And Implement Of Insulator Flashover Detection System For Substation Based On Computer Vision

With the rapid development of the railway industry in entire country,the railway line is spread all over the country,and the environment of the traction substation is complex and changeable.Therefore,the insulator of the substation will be affected by climate such as rain,haze,frost,etc,and it is easy to accumulate dust,which causes the insulator’s resistance to change.There will be a flashover phenomenon along the surface discharge or breakdown air.If flashover occurs frequently,it will cause major safety accidents such as substation tripping and fire,so flashover is one of the common safety hazards in traction substations.Therefore,it is of great significance for the safety of traction substations to carry out research on the flashover phenomenon of insulators in substations,identify flashovers at the alarm site,record the flashover frequency,analyze the influence factors of weather changes on insulator flashovers,and predict the probability of occurrence of flashovers.The system designed by the thesis is mainly divided into three parts: remote interaction part,server part and local data processing part.The remote interaction part mainly provides staff with functional operating windows of the system,as well as preview video,alarm information,predictive analysis and other data displays.A database is deployed on the server side,which is responsible for deploying the data information of the storage system and the communication part of the control system.The local data processing terminal is responsible for monitoring on-site cameras and real-time analysis of the retrieved insulator video to monitor the occurrence of flashover.In terms of insulator flashover monitoring technology,here we have collected and organized monitoring videos about insulators in the substation site.In the program,the opencv library is used to disassemble the video frame by frame and package it,and input the classified video package into the C3 D neural network to extract features Information,and then directly add the output to the MIL classification model for training to obtain an abnormal video detection learning model.In the system,the SDK library provided by the manufacturer is used for secondary development of the camera,and the real-time monitoring video of the substation site is called;after the real-time video is split into video packets and input into the learning model,the hierarchical background subtraction method is used to extract the frame Prospect of the abnormal video package output by the model,calculate the distance between the point and the centroid of the edge of the abnormal object;calculate the distance information using the wavelet transform algorithm,calculate the normalized ratio of the filtered numerical signal and the calibration signal;the insulator flashover is through Judging by comparing the size of the threshold and the ratio.The system is about to record the flashover frequency of the substation insulator,collect the environmental information data set on the site of the substation,and use the LSTM network model training data to predict the probability of the occurrence of the flashover phenomenon and analyze each environmental impact factor.The computer vision-based substation insulator flashover monitoring system based on computer vision is designed to identify,record and analyze the entire process of substation insulator flashover.After field experiments,the system can stably,efficiently and in real time complete the task of monitoring the flashover of insulators in traction substations,with a recognition rate of 97%,and successfully predicting the occurrence of flashover.Facts have proved that the research and system development of this paper have positive practical significance for strengthening the safety of substations.