| Nowadays,China ’s construction of power system has entered a new stage of comprehensive and rapid development.In the field of substations.According to the development planning requirements formulated by the national power system,the application of the intelligent substation needs Imperative under the circumstances that the traditional substation operation mode cannot meet current development.In the construction of intelligent substations,in order to solve the problem of intelligent detection of massive infrared image data of power equipment,this thesis first preprocesses the infrared images to isolate the defective power equipment and extracts the defective areas,this measure can improve the accuracy of subsequent algorithms.Based on the theory of deep learning,a method of defect recognition and classification for infrared images of power equipment based on deep learning is proposed.The content of this thesis is as follows:1)According to the characteristics of infrared images of power equipment,this thesis proposes a method for determining whether there are defects in power equipment based on the HSV color space and a method for extracting defect areas of power equipment in infrared image based on SLIC image segmentation algorithm.Through the above two algorithms,the infrared image of the defective power equipment is separated from the massive image data,and then the defective equipment area is extracted from the image.2)In this thesis,by discussing the relationship between the types of thermal defects in power equipment and the types of equipment,a new idea for converting complex thermal defect classification problems into equipment classification problems is proposed.Based on deep learning theory,a method of infrared image classification of defective power equipment based on convolutional neural network is proposed.By analyzing the network structure and working principle of VGGNet and the densely connected modules,a DV-CNN model is constructed for classification of defective power equipment.Through experimental comparative analysis of this model,the model has a TOP-3 classification and recognition accuracy rate of the infrared images of seven types of defective power equipment in the data set of this thesis reaching 90.74%.According to the special characteristics of the thermal defect types of current transformers,an improved CNN model is proposed to classify the thermal defect types of current transformers.Through experimental comparative analysis of this model,the accuracy rate of thermal defect classification and recognition reached 85.84%.The algorithm of power equipment defect detection proposed in this thesis has a good performance in the intelligent detection of power equipment defects.In practical applications,it can greatly improve the detection efficiency and reduce power equipment operation and maintenance costs.This research has made a positive exploration for the construction of intelligent substation. |
