Research On Partial Discharge Development Characteristics And Defect Recognition Of Oil-impregnated Paper Bushing Under Typical Defects

In recent years,with the continuous expansion of the UHV network construction scale,the number of oil-impregnated paper bushings in use has further increased,and the bushings have been in relatively harsh operating conditions for a long time,resulting in frequent bushing failures.The failure has caused a huge impact on the reliable operation of the transformer and even the power supply system.After investigation by relevant personnel,it was found that the partial discharge caused by the poor manufacturing process of the main insulation of the casing(such as conductive impurities,uneven cutting of the aluminum foil edge,air gap,insufficient drying and impregnation,burrs on the surface of the catheter,etc.)is one of the main reasons which caused the casing to fail.At present,for this research,one is to find the simple cause of the faulty physical bushing,and the other is to use a typical electrode structure to conduct partial discharge research on oil-impregnated paperboard.Unlike the single oil-impregnated paperboard structure,the main insulation of bushing is formed by alternately winding multiple layers of insulating paper and aluminum foil on the catheter.Due to the structural differences,the partial discharge characteristics will be different,and the existing research results may not be fully applicable to the bushing.Therefore,further research is needed to consider the partial discharge of the actual structure of the main insulation of bushing to provide a reference for the diagnosis and evaluation of the bushing insulation state.Based on the design principle of oil-impregnated paper bushing,this paper designs and prepares four defect models from the perspectives of similar structure and same materials.The partial discharge test was carried out for each defect model,and the discharge process was divided according to the characterization parameters(discharge power,average discharge amount,maximum discharge amount).And the spectral features of different discharge stages were analyzed at the same time.The texture feature extraction method was used to extract the spectrogram characteristics of each defect,and the variation rules of the characteristics of different discharge stages were analyzed.The feature quantities with monotonous change trend were selected to characterize the development process of partial discharge of defects,and at the same time,to provide a reference basis for evaluating the severity of partial discharge.In addition,this paper also carried out research on casing defect recognition based on image texture features and multi-class support vector machine.The main contents included the construction of texture feature space,the use of kernel principal component analysis(KPCA)for feature space dimensionality reduction,design of multi-class support vector machine model,selection of kernel function and parameter optimization.And the recognition results before and after dimensionality reduction were compared and analyzed.The research results show that:(1)For the defects of metal particles,the discharge process is divided into the initial discharge stage,the discharge development stage and the severe discharge stage according to the change trend of discharge power and average discharge amount.The angular second moment,correlation,contrast and entropy extracted based on the gray level co-occurrence matrix can be used as a reference basis for characterizing the discharge process and evaluating the severity of the discharge;(2)For the tip defects,the discharge process is divided into the discharge initial stage,the discharge development stage,the discharge stable stage and the discharge dangerous stage according to the change trend of the maximum discharge amount and discharge power.The mean,entropy,energy and variance extracted through the local binary model can be used as the characteristic quantities to characterize the discharge process,and at the same time provide a reference basis for evaluating the severity of the discharge;(3)For air gap defects,the discharge process is divided into the initial discharge stage,the discharge development stage and the severe discharge stage according to the change trend of the maximum discharge amount and the average discharge amount.The roughness and contrast based on Tamura texture feature extraction can better characterize the discharge process,and at the same time can provide a reference for evaluating the severity of discharge;(4)For burr defects,the discharge process is divided into the discharge initial stage,the discharge development stage and the severe discharge stage according to the change trend of discharge power,average discharge amount and maximum discharge amount.The large gradient advantage,gradient standard deviation,inverse differential moment and inertia extracted through the gray gradient co-occurrence matrix can provide a reference basis for the characterization of the discharge process and the evaluation of the discharge severity;(5)It is feasible to identify the defect type of oil-impregnated paper bushings based on image texture features and multi-class support vector machines,and it has achieved good recognition results.Moreover,the feature quantity after KPCA dimension reduction still has a high recognition rate,indicating that the feature space after dimension reduction can better characterize the original feature space.