Research On Association Rule And Set Pair Analysis For Power Transformer Condition Assessment In The Big Data Environment

Power transformer is the key hub equipment of electric system. The Condition BasedMaintenance (CBM) replacing the periodic maintenance is an inevitable development trendof the electric power industry. The precondition of CBM is the accurate assessment on theoperating state of transformer according to the equipment test data.With the continuousdevelopment of national smart grid construction, the state data generated by statemonitoring of power apparatus increase exponentially and the related analysis research ofthese big data about transformer device in regional power grid. Therefore, data processingand information mining technology around the fault diagnosis and state assessment must beof great significance on the business of smart grid equipment side.Currently, one of the key emphasis in the big data research of Gird Company is thecorrelation analysis in statistical on the huge amounts of data, such as search, comparison,classification and clustering, etc. However, most methods of power transformer conditionassessment for equipment side either ignore the correlation analysis and inner link of dataand information about transformer state detection, or never notice the uncertain problems oftransformer state information. To this end, this thesis presents a new method for transformerfault diagnosis in the big data environment of smart grid, which is the combination of setpair analysis and association rules.Firstly, this thesis selects the most representative status parameters which canaccurately and effectively reflect the state of health of the transformer as the fault types andfault symptoms based on relational database theory, related standards and practicaloperating experience. Use support of association rules to manage fault types and faultsymptoms, and analyze the mutual coupling relationship between the fault types and faultsymptoms. Then, reduce the group of fault symptoms of each type of fault types. Finally,establish state assessment model. On the method of determining the weight coefficient, thisthesis obtains constant weight coefficient of fault symptoms by comparing the calculatedconfidence of association rules, which solves the problem of experts’ subjective opinionaffecting the accuracy of the weight. Meanwhile, it can effectively avoid the defects that constant weighting factor under a single fault type does not accurately reflect the overallhealth of transformer by using variable weight coefficients and determining the weightingfactor score based on conditions of each fault type.Secondly, the operating status of power transformers is graded to determine theelement number and expression of connection degree which improve the evaluationaccuracy of transformer condition assessment and fault diagnosis system. Relativeimpairment grade and membership grade of fuzzy theory are introduced to build theidentical discrepancy contrary (IDC) evaluation matrix. The introduction of Fuzzy theoryhas some scientific basis which can effectively avoid the problem of expert opinion andsubjectivity of experience. For the expression of multivariate connection degree, themethod of dichotomy is used to deal with coefficient of diversity factor. Combined with theweight coefficient determined by association rules, the connection value of transformeroverall operational status and various types of faults can be obtained. Compared to thepartition of state levels, the health status of transformer can be assessed and the faults canbe diagnosed.Finally, the correctness and effectiveness of the method proposed in this article isverified through calculation examples and statistical result. Compared with the method thatassociation rule and set pair analysis is separately applied to the analysis of transformercondition assessment, the method proposed in this article have a higher rate of positivejudgment and it is also excellent in multiple fault diagnosis.