Power transformer insulation diagnosis under measurement originated uncertainties

This paper addresses the problem of assessing the fault symptoms of a power transformer's insulation system under measurement originated uncertainties, which arise from the imprecision of data sets used to identify different fault types. Such measurement originated uncertainties cause difficulties in interpreting measurement results obtained from practical power transformer insulation diagnosis methods using Dissolved Gas Analysis (DGA) and Polarisation and Depolarisation Current (PDC) measurement techniques. A support vector machine (SVM) algorithm has been developed in this paper to tackle any measurement originated uncertainties. Two case studies of interpreting DGA and PDC data sets are presented to demonstrate the applicability of the developed algorithm in practical power transformer insulation condition assessment.