Study On Condition Assessment Approach For Transformer Based On Information Fusion

Transformers are key equipments in power plants and substations and their own operation conditions always affect the safe and stable operation of the entire power system.At present,the service life of many transformers in our country is close to or even far greater than their design life,which leads to an increase in failure rates of some transformers year by year.In order to ensure the safe and reliable operation of the transformers and save the costs of maintenance,the maintenance method of transformers in the grid companies is gradually changing from accident-based maintenance,time-based maintenance to condition-based maintenance.Condition-based maintenance is based on the results of condition assessment.Therefore,it is extremely important to implement the condition assessment for transformers.Transformer condition parameters are numerous and the mechanism is complex.There are many uncertain factors in the condition assessment process.Therefore,an approach for transformer condition assessment based on information fusion is studied in this thesis.The main contents of this thesis are as follows:(1)An approach of transformer operation condition assessment based on improved evidence theory fusion is proposed.Firstly,the deterioration degrees are calculated through preprocessing the operation condition assessment indexes.Secondly,the improved analytic hierarchy process method is used to calculate the constant weights of the indexes,and the equilibrium function is introduced to calculate the indexes degradation variable weights.Then,the condition parameters are layered and the comprehensive connection degree of each layer is calculated as an independent body of evidence.Finally,the evidence bodies modified by the sensitivity factors and the credibility coefficients are fused through the fusion rule.Through the comparative analysis of the experimental results,the feasibility and effectiveness of the operation condition assessment method proposed in this thesis are verified.(2)A transformer fault diagnosis approach based on principal component analysis and adaptive relevance vector machine is presented.Firstly,the gas content and ratio are used as fault feature information,and are processed by principal component analysis to reduce dimension,realizing the fusion of feature layer.Secondly,an adaptive relevance vector machine model based on particle swarm optimization and cross validation is presented to realize the automatic optimization of kernel parameters in sample training.Through the comparative analysis of the experimental results,the feasibility and effectiveness of the fault diagnosis method presented in this thesis are verified.(3)The software of transformer condition assessment is designed and implemented.Firstly,based on the basic requirements of the software,the four main functions of the software are determined: storing and reading data,operation condition assessment,fault diagnosis,and exporting assessment report.Secondly,a three level system architecture of the software is designed by MATLAB and Access database.Then,the realization method of software function is given.Finally,the practicability of the software is verified by showing the actual operation effect.