Research On Remaining Useful Life Prediction Of Power Transformer Based On Data-driven Method

The development of modern information technology provides better technical means for the whole life cycle management of power equipment.Power transformer is one of the most important equipment in power system.If the fault occurs,it will have obvious negative effects on the safe and stable operation of power system.Under the condition of complete data,the proper condition maintenance or preventive maintenance for it can reduce the probability of failure,not only can reduce the maintenance cost of the power transformer,but also have a significant positive effect on the safe and stable operation of the power system.But the current research mostly focuses on the transformer health management in one aspect or a certain stage,the research results are isolated,poor integration.Firstly,sort out the main problems existing in the research of transformer,and gives the solution based on prognostics and health management(PHM).Lastly,the transformer health management cycle is defined according to the PHM method system.A composed method based on deep belief nets and D-S evidence theory is proposed to solve the problem of deep fault diagnosis of power transformer.Deep belief nets is used to solve multidimensional data feature extraction and classification,and D-S evidence theory solve the problem of uncertainty.Therefore,a multilevel decision fusion model is proposed.The simulation results based on DGA,partial discharge data,historical fault data and family history of quality show that the proposed method is effective for fault diagnosis of power transformer which has a large number of multi-source information.Combining the remaining useful life(RUL)prediction results with the reasonable maintenance model can improve the availability of the power transformer and reduce the maintenance cost.Based on the distribution of degradation data of oil paper insulation in power transformers,a RUL model of power transformer based on multivariate Weibull distribution is proposed.First of all,realize the modeling for multi variable and multi parameters degradation based on statistics Weibull parameters degradation model.Then,use the maximum likelihood estimation method to estimate model parameters based on the performance degradation data to obtain the reliability function of power transformer.At last,use the results of failure diagnosis as the failure index to predict the RUL.The simulation results show that the method can accurately reflect the transformer degradation state and the RUL.