Research On Transformer Fault Diagnosis Method Based On Grey Correlation And Entropy Weight Fusion

At present,power transformers,as important power equipment,may cause power outages and even explosions if the faulty parts of the transformer are not repaired and repaired in time due to the health problems and potential operational hazards.Reliability has extremely serious consequences.Therefore,in order to ensure the safe and reliable operation of the power system,an effective method should be developed to diagnose the latent faults inside the transformer.Combined with the practical application effect of engineering in China for more than 30 years,the oil dissolution analysis and diagnosis technology(DGA)has been widely used to detect the latent fault of electrical equipment using insulating oil as dielectric.The inherent diagnosis is based on detecting whether the transformer is The essence of the failure.However,the traditional DGA method has limited diagnostic range and low diagnostic accuracy when used in a single use.For example,the gas ratio method can achieve a correct rate of over 90%for overheat or discharge fault diagnosis.However,if the location where the fault occurs is further found,it cannot be accurately judged.Therefore,relying on the above single diagnostic technology to determine the type of transformer failure,its accuracy is low,so it can not meet the engineering needs.In response to this problem,this paper combines traditional DGA diagnostic techniques with artificial intelligence methods.Using gray correlation analysis to deal with the advantages of small sample and bad information samples and the advantages of entropy weight method to deal with the weighted problem.Analyze the dissolved gas data samples of oil in the operation transformer,and propose a transformer fault diagnosis based on grey correlation and entropy weight fusion.method.The method firstly establishes 8 kinds of transformer fault states,and selects 8 groups of DGA oil chromatographic data that have been determined by the suspension cover to determine the fault type of the transformer to be standardized to obtain standardized sample data of 8 kinds of transformer fault states;then adopt the dimensionless standardization method Standardized sample data,the standard fault state model of 8×5-order transformer is obtained;and the entropy weight of the normalized sample data of 8 transformer fault states is solved to determine the weight of the five fault indicators;finally,the detected transformer data samples are dimensionlessly standardized.After processing,the standard weights of the five fault indicators obtained by the combination are substituted into the gray weighted association formula,and the gray weighted relevance of the data samples to be detected is obtained;then the gray weighted relevance of the data samples to be detected is arranged in descending order.That is,the indicator in the standard failure mode corresponding to the maximum gray weighted correlation degree is the type of transformer fault diagnosed by the method in this paper.finally,the transformer corresponding to the sample to be detected which is detected by the method is subjected to the hood inspection,through three examples The inspection result indicates that The results obtained by the method are consistent with the results obtained by the hood inspection.The results show that the method has a good engineering and practical prospects.