Research On Infrared Image Recognition And Fault Diagnosis Method Of High Voltage Switchgear

As an indispensable electrical equipment for protecting various power distribution devices in the power system,high-voltage switchgear is subjected to high voltage,the high current,mechanical stress and environmental factors during operation,which causes highvoltage switchgear to be accompanied by heat.Infrared diagnosis technology is an effective means of thermal abnormal fault diagnosis of high-voltage switchgear.The collected infrared images can reflect the overload,poor contact,loose joints and other faults in the equipment,which are of great significance for the timely detection,handling,and prevention of major accidents caused by high-voltage switchgear failures.In order to quickly and effectively diagnose the thermal anomalies of high-voltage switchgear,this paper starts from two practical problems: infrared image recognition and fault diagnosis.By analyzing the typical thermal anomalies of high-voltage switchgear and the characteristics of infrared images,aiming at the shortcomings of traditional algorithms in high-voltage switchgear in infrared image recognition and fault diagnosis,an infrared image database of high-voltage switchgear was established,and a new high-voltage switchgear identification and The thermal anomaly detection algorithm has achieved the following results:1)There are a lot of salt and pepper noise and Gaussian noise in the infrared image of the high-voltage switchgear during the acquisition and transmission process.Therefore,in view of the shortcomings of the traditional infrared image denoising method,an adaptive median filter algorithm with improved mean is proposed.The experimental result shows that the algorithm can effectively filter out the noise in the infrared image of high-voltage switchgear,at the same time it can maintain the clarity of the image,and it shows good filtering performance on the PSNR and MSE indicators.2)Aiming at the problems of poor edge accuracy and unclear contours in the infrared image segmentation process of the high-voltage switch target device,the traditional infrared image segmentation method of high-voltage switch device is studied,and the infrared image segmentation method for the target high-voltage switchgear based on Mask R-CNN algorithm is proposed.The experimental result shows that the segmentation method has high accuracy and strong generalization ability.The test accuracy rate is 86%,and it can be initially applied to infrared image segmentation of high-voltage switchgear.3)Aiming at the problem of detection accuracy reduction caused by the complex scene and uneven size of the target location in infrared image abnormal heating point detection of high-voltage switchgear,the improved YOLO v3 algorithm realizes the rapid detection,identification and positioning of abnormal high-temperature switchgear abnormal heating point.Meanwhile,a data set for abnormal heating points of infrared images of high-voltage switchgear was established.The experimental result shows that the detection method has fast recognition speed,high accuracy and strong generalization ability,and it can be preliminary applied to the detection of abnormal heating point targets of high-voltage switchgear.