Study For Transformer Fault Diagnosis And Forecast Based On Data Mining

Power transformer is one of the expenxive device in power system, power transformer fault diagnose is vital to make the whole power system run normally, the failures of transformers can result in serious issues, such as service disruptions and severe economic losses.Transformer fault diagnose and fault prediction are the basis of the transformer condition based maintenance. In this paper we have done in-depth research in transformer data warehouse construction, transformer fault diagnosis and fault forecast.A data precessing method is provided through constructing the power transformer fault information data warehouse, based on analyzing and reorganizing the various existing data sources, the transformer failure diagnosis and early failure predication can be benefit from OLAP (On-Line Analysis Processing) and data mining technology.Aiming at the low diagnositic performance of single models caused by transformer testing data, a combination diagnosis model based on the Bayes classifiers and SVM is proposed in this paper. This model can solve the low efficientcy and accuracy when using support vector machines (SVM) to solve multi-classification question.Due to the randomness and uncertainty of power transformer fault diagnosis data, a novel method using selective Bayes classifier SRBC is proposed to estimate probability distribution directly from incomplete transformer fault data, and then estimates mutually information. This proposed. SRBC approach can solve the randomness and uncertainty problem of power transformer fault diagnosis data. The experimental results show that this method can obviously enhance the accuracy rate.Aiming at the low forecasting accuracy of traditional predictive approaches, a combinational model is proposed on the basis of SVM theory in this paper. During the process of the forecast, firstly several single forecast approaches are used to form a model group, and a set of data in time sequence on each dissolved gas are fitted by the model group. And then the fitted results by different traditional predictive models in time sequence act as the input of the support vector machine regression (SVMR) model, and the changeable weights combinational SVMR prediction model is obtained.Adopting .NET platform and the C/S structure, the transformer fault diagnosis system has been developed using the VC# language and SQL Server. This system mainly includes the transformer oil chromatograph forecast, the transformer fault diagnosis as well as the data warehouse OLAP analysis and so on several modules. This system has been put the operation in the Hebei Hengshui electric power company.