| In this paper,two kinds of electrical equipment such as disconnecting switch and transformer in the complex substation are mainly analyzed.Two algorithms are proposed for the two kinds of equipment identification,which are the identification and status identification algorithm of the disconnector and the infrared image of the transformer Fault diagnosis method.The main research methods for the identification and status recognition of isolating switch are as follows:Firstly,based on the traditional digital image processing technology,the identification information of the isolating switch is extracted.Secondly,analyze the texture features and color features that are characteristic of the isolating switch and extract the suitable features.Finally,based on the machine learning technology,an information fusion method is proposed,which integrates the feature vectors for information fusion and inputs SVM to classify the target into two classes.The model of the SVM training is used to judge the switch state of the isolating switch.The comparison between the single-feature image classification and the multi-feature fusion image classification shows that the proposed algorithm is more accurate.The main research method of fault identification for transformer infrared imaging is to set up the target detection model of VGGNet fault based on the emerging deep learning framework.In the pre-training images,the fault location is effectively calibrated,and then the training samples are used for training.After the training model is obtained,the testing samples are used for testing.Into the parameter adjustment training to get the fault accuracy.Finally,an improved full-connection network structure is proposed to make the new model more accurate than the old model.Finally,improve fault recognition accuracy.This paper fully combines the advantages of traditional image processing technology and the advantages of deep learning algorithm intelligent recognition.Based on the cross-knowledge of multi-domain such as feature engineering,data mining and machine learning,this paper propose a kind of intelligent identification solution of isolation switch identification and status identification and transformer fault diagnosis,provides a new way of thinking for the construction of "smart grid".And the algorithm in this paper has been shown to be effective by experiments.Comparative tests show that the improved algorithm has better recognition effect and better accuracy in fault isolation of the transformer and transformer. |
