Study On Partial Discharge Characteristics Of Typical Insulation Defects In Transformers And Defect Type Identification

Power transformer is one of the core equipments in power system.Its safe and stable operation provides an important guarantee for the safe and stable operation of power system.Oil paper insulation is the main insulation system of power transformer at present,when its internal due to the deterioration of insulation partial discharge will cause immeasurable losses to the entire power system,due to the formation and development of different defects of partial discharge mechanism is different,its partial discharge development characteristics are not the same.Therefore,the classification of partial discharge types with different defect types is of great significance for the rapid diagnosis of different defects in power transformers.This paper the method of industrial manufacturing insulating board,is made without sulfate pulp bleaching needle timber as raw materials,insulation board after handling of the insulating board different process,preparation into not board(hereinafter referred to as the pure cardboard)containing defects,including air gap defect and metal impurity defects of three different types of insulation board sample.To delve into different development characteristics of partial discharge,defect board insulation partial discharge test platform in laboratory were set up the paper,in order to get a better look to the partial discharge characteristics of different sample development rule,the ladder step up method and constant pressure method combining way of booster,pressure test to different defect types of specimen.The test results show that,due to the introduction of defects,the initial discharge voltage of the pure paperboard sample is the highest,the sample with air gap defects is the second,the sample with metal defects is the lowest,the average discharge quantityand discharge times of the three samples increase with the discharge time,and the partial discharge duration of the three samples is different.Second,in order to explore the different classification identification of the defect sample based on the shape parameters and the positive and negative half cycle outline differences parameter extraction,for different defect in the sample from the initial discharge to near breakdown phase discharge two-dimensional mapping of the16 characteristic parameters extraction of partial discharge,deep for subsequent use neural network to classification of sample laid the foundation data.Finally,a deep neural network was constructed to identify the partial discharge types of different defective samples.the characteristic quantities of partial discharge were divided into training set and prediction set according to a certain proportion,and the influence of different parameters of neural network on the accuracy of training set and prediction set was deeply explored,and the structure of neural network was determined.The results of partial discharge classification by single order neural network and deep third order neural network were compared.The experimental results show that the deep third-order neural network has higher recognition rate and is suitable for partial discharge recognition of different defect samples.