An Infrared Detection Method For Key Equipment In Substation Based On Image Processing And Neural Network

The infrared detection technology has been widely used in the fault diagnosis of power grid equipment with the advantages of high efficiency and live working.The infrared thermograph in substation has become more popular,and the industry authorities formulate national standards for infrared detection work and constantly updated,various provinces and cities have set up guidance manual for infrared detection in succession,which greatly promoted the infrared detection technology in power equipment fault diagnosis application and it speeds up the transformation speed of maintenance mode from ex post maintenance to condition based maintenance,improves the efficiency of power grid operation and maintenance,and has important positive significance for the safe operation of power system.In the daily infrared detection work of substations,on the one hand,a large number of infrared image samples of the equipment have been accumulated,and manual processing by the operation staff alone will result in a huge amount of work,and misjudgments and missed judgments may be caused due to subjectivity;On the other hand,the voltage-heating type faults that occur in some devices do not perform well in the infrared thermal image.In addition,due to the influence of the imaging quality of the infrared image,it is difficult to directly determine the operating status of the device through the image.At the same time,there are many types of equipment in substations,different types of equipment have different types of failures,and the required processing methods are also different.There is no standard processing method with applicability.In this paper,aiming at the problems in the infrared detection work mentioned above,the structure,type of failure and infrared image features of several key equipment such as insulator strings,arresters,breakers,voltage transformers,current transformers,transformer high-voltage bushings and disconnecting switches in 500 kV substations are analyzed first.Then,the infrared image standard processing flow which has applicability to these types of devices is proposed,and corresponding optimization is performed according to the specific structural characteristics of each type of device to achieve the segmentation and extraction of the target region.Finally,the temperature feature vector is constructed based on the image processing results and equipment characteristics,and the BP neural network method is used to identify the equipment failure.Finally,in order to further improve the diagnostic efficiency and realize the automatic selection of image processing methods and detection networks,this paper uses the convolutional neural network to classify the images and establishes the final infrared detection method and flow for key equipment in substations.The result proves that the image processing method,fault recognition network and image classification network proposed in this paper have high accuracy both in the individual and the whole,which is helpful to further improve the efficiency and practicality of infrared detection work in fault diagnosis of power grid equipment.