Study On Characteristics And Stage Identification Of Partial Discharge Process In Oil-paper Insulation

Transformer is the key device in grid system.Partial discharge is the main reason for the deterioration of oil paper insulation.Once the transformer accident is caused by insulation failure,the power grid will be cut off in a large area which may result in huge economic losses.At present,intelligent partial discharge monitoring has become an important part of a strong power grid.Therefore,it is of great engineering significance and theoretical value to study the partial discharge process characteristics and phase identification of oil paper insulation of transformer.In view of the partial discharge phenomenon of oil paper insulation inside the transformer,a partial discharge research platform is established.The models of air-gap discharge and surface discharge were designed and five aging oil-paper insulation samples were obtained by accelerated thermal aging.The step-by-step voltage rise test of oil paper insulation samples with two typical defects was carried out to track the partial discharge process of the samples.The development law of average discharge quantity,maximum discharge quantity,total discharge quantity,total discharge times and other parameters are analyzed.Based on the evolution of experimental data and the difference of discharge phenomena,the partial discharge stages of oil-paper insulation defects with different aging degrees were divided.The influence of aging degree on partial discharge development process was explained by means of micro detection and finite element simulation.Twenty-nine statistical characteristic parameters of phase resolved partial discharge atlas are extracted,and the new characteristic parameters are obtained by using the local linear embedding algorithm.The new characteristic parameters are input into probabilistic neural network to identify the development stages of air-gap discharge and surface discharge.Compared with the classification results of generalized regression neural network and back propagation neural network,the accuracy of the recognition method is evaluated.The results show that the process of air-gap discharge and surface discharge can be divided into initial stage,development stage,stable stage and near breakdown stage.It is found that the aging degree has a great influence on the later stage of discharge and a little influence on the early stage of discharge.The phenomenon is caused by the pore structure of the aged insulating board.Partial linear embedding-probabilistic neural network model is proposed to identify the development stages of partial discharge in oil-paper insulation withdifferent aging degrees.The model has faster recognition speed and higher recognition accuracy than the traditional generalized regression neural network model and back propagation neural network model.In this paper,the research on the characteristics of discharge process and stage identification of two typical defects of oil-paper insulation can provide some reference for the nondestructive state evaluation of oil paper insulation of transformer.