Study On Infrared Fault Diagmosis Method Of Electrical Equipment Based On Deep Learning

The infrared detection method,having no contact with the device,is convenient and fast.The method does not affect the normal operation of the system.Therefore,it has become an effective means of detecting the status of power equipment.However,the current infrared image diagnosis of power equipment mainly relies on manual analysis and judgment of infrared pictures.The method is backward and the efficiency is low.In order to improve the level and efficiency of power grid equipment fault diagnosis and reduce the burden on staff,the method of using machine learning to automatically identify the type of substation equipment is studied.The method of using deep learning to automatically divide the structure of the equipment is studied.On this basis,the method of automatic extraction of equipment temperature and automatic judgment of equipment status is studied to realize infrared intelligent diagnosis of substation equipment faults.First,the automatic recognition method of the power equipment in infrared images is studied.Establish infrared image database of transformer equipment.Use labelimg software to make VOC data set of substation equipment.Choose the deep learning algorithm R-FCN,using the training set of power equipment for model training.At the same time,using transfer learning,Resnet101 and OHEM methods to improve the model,effectively improving the recognition accuracy of various substation equipment.Experiments and comparison of different methods show that the improved R-FCN algorithm has high recognition accuracy,which can obtain relatively complete and accurate positioning of the target device.This method is feasible in the identification and positioning of infrared images with complex background,provideding a good basis for the subsequent division of the device structure.Then,the automatic method of the structure division of the power equipment in the infrared image is studied.Establish infrared image database of power equipment.Use labelme software to create a data set of each substation equipment structure.Choose the deep learning algorithm Mask RCNN.The model training is carried out by using the training set of each structure of the substation equipment,and finally the structure division models of the arrester,current transformer,voltage transformer and breaker are obtained.According to the output result of the recognition model,the corresponding substation equipment structure division model is selected,and then the image is input,and finally the structure division diagram of various types of equipment is obtained.The experiment and the comparison of different methods show that the structure area division obtained by this method is basically consistent with the actual structure area of the equipment,which provides a basis for the automatic diagnosis of thermal faults of substation equipment.Finally,the method of thermal fault diagnosis of substation equipment is studied.On the basis of the above research,combined with the specifications and standards of infrared diagnosis of substation equipment,the fault diagnosis criteria of different types of equipment are studied,and the flow of equipment fault diagnosis is determined.