Partial Discharge State Recognition Of Oil Paper Insulation Based On Information Fusion

Oil-paper insulation is the main insulation structure of oil-immersed transformers,which will be affected by partial discharge and aging effects during long-term operation,resulting in deterioration of insulation performance.There is a coupling relationship between partial discharge and aging of oil-paper insulation.On the one hand,the insulation damage caused by partial discharge will accelerate the aging of oil-paper insulation.On the other hand,the aging of oil-paper insulation will aggravate the phenomenon of partial discharge,and the two influence each other.It is difficult to analyze and identify both at the same time through a single characteristic information source.Information fusion technology has excellent capabilities in processing multi-source information and solving complex classification problems.This provides the possibility to realize the comprehensive diagnosis of the transformer oil-paper insulation discharge state.In this paper,aiming at the partial discharge and aging problems of the oil-paper insulation structure inside the transformer,four types of oil-paper insulation samples with different aging degrees have been prepared,and three typical partial discharge models have been built.Carry out partial discharge experiment of oil-paper insulation based on pulse current method,obtain partial discharge data of oil-paper samples,and obtain gas product data of samples based on Dissolved Gas Analysis(DGA).The statistical parameter method and the ratio method are used to extract the partial discharge characteristic parameters of oil-paper insulation and the characteristic parameters of oil and gas respectively,and the dimensionality reduction is carried out by the factor analysis method.Use probabilistic neural networks to train the data.Based on the D-S evidence theory,the two types of information sources are fused to construct a comprehensive diagnosis system for the discharge state of oil-paper insulation.The research results show that the characteristic information of the partial discharge model and the aging state will be mixed with each other,which greatly increases the difficulty of classification.DS evidence theory can effectively correct misjudgments and resolve conflicts of evidence.The comprehensive recognition accuracy rate of oil-paper insulation discharge status after information fusion has been increased from 62.2% to 84.7%,indicating that the comprehensive diagnosis system for oil-paper insulation discharge status based on DS evidence fusion theory has good recognition results.The research on the identification of discharge mode and aging stage of oil-paper insulation in this paper can provide a certain reference for the non-destructive evaluation of transformer oil-paper insulation.