The use of statistical methods in the evaluation of power transformer faults with frequency response analysis

Abstract

Purpose: The purpose of this study is to present principal of the FRA and to explain indexes used for statistical interpretation of the results on the real example cases. Theory and Methods: In this study, FRA method, which is one of the new generation test methods used in diagnosis of power transformer faults, is discussed. Statistical indices used in the evaluation of FRA results are introduced and comparisons are made according to various criteria. FRA test results of two real cases are evaluated both with statistical methods and experimentally. Results: The obtained experimental results showed that; residual magnetism and measuring direction is significantly influence of the FRA test results. In addition, statistical methods; Correlation Coefficient (CC), Cross Correlation Function (CCF), Standard Deviation (SD), Absolute Sum of Logarithmic Error (ASLE), Absolute Average Difference (DABS), Standard Difference Area (SDA), and Lin's Concordance Coefficient (LCC) are successfully applied as numerical index in the evaluation of FRA test results for diagnosis of the power transformer faults. Considering all these situations, expert knowledge is required in the interpretation of FRA results as well statistical methods can be used. Conclusion: FRA gives more sensitive and successful results than other electrical tests in the diagnosis of winding faults in transformers and structural dislocation in the core, especially in determining the structural movement that may occurs in windings and core.