Research On Online Monitoring And Intelligent Diagnosis Method Of Mechanical State On GIS Breaking Equipment

The Gas Insulated Switchgear(GIS)is the important equipment in the power transmission system.GIS is more and more widely used in power system because of its small size and strong anti-electromagnetic interference ability,and the mechanical failure of GIS breaking equipment is an important reason that threatens its safe and stable operation.Aiming at the low accuracy of traditional GIS breaking equipment mechanical fault diagnosis methods,this paper proposed an intelligent diagnosis method based on wavelet time-frequency spectrum of vibration signals and deep learning,which provides a deeper.faster and more accurate method for feature extraction and fault diagnosis of GIS breaking equipment mechanical fault.The main contents include:(1)A test platform for mechanical characteristics of GIS breaking equipment was built.By analyzing the structure and working principle of GIS breaking equipment,the selection and arrangement of sensors in fault simulation test were obtained.According to the principle of fault and its intuitive appearance,five kinds of common mechanical fault test simulation methods were given,and the test platform was used to complete the acquisition of vibration signal and characteristic data in the process of opening and closing of the breaking equipment under various faults.(2)Combined with the working process of the breaking equipment,the changing relationship between mechanical failure and GIS dynamic characteristics was analyzed.Then,the wavelet time-frequency spectrum of vibration signals was obtained by using the wavelet transform algorithm.The features of vibration signals in the time-frequency domain under different mechanical fault states were extracted,and the GIS mechanical fault wavelet time-frequency spectrum database was constructed.The results show that the wavelet time-frequency spectrum of vibration signals in different states is obviously different,which has the basic conditions to be used as samples of the next intelligent diagnosis algorithm.(3)Aiming at the low recognition accuracy of traditional feature extraction methods,an intelligent diagnosis method based on VGG16 for the mechanical state of GIS breaking equipment was proposed.The improved generative adversarial network was used to expand the fault samples,and the mechanical fault sample database was obtained.When the amount of wavelet time-frequency spectrum used as training samples is sufficient,the accuracy of GIS mechanical fault intelligent diagnosis algorithm based on VGG16 can reach 98.8%in the closing process and 97.2%in the opening process,which is significantly improved compared with traditional diagnosis methods.For SVM,BPNN and VGG16 models,the recognition accuracy of wavelet time-frequency spectrum as samples is higher than that of the original vibration signals.The lift rates are 5%,18%and 24%in the closing signal and 8%,19%and 25%in the opening signal,respectively.