Statistical Analysis of the Partial Discharge Data for the Classification of Transformer Oil

Abstract

Statistical analysis of the Partial Discharge (PD) data is proposed for the quality assessment of the transformer oil. Initially, the PD data is represented as X and Y marginal histograms and the Histogram Similarity Measures (HSM) are adopted for the supervised classification of the transformer oil. Three different HSM techniques: cross-correlation, Chi-square and the Kolmogorov-Smirnov tests are employed. 45 transformer oil samples collected from the State Electricity Board are used for the classification of the oil samples to Class 1, Class 2 and Class 3 as per IEEE C.57.106-2006 standard. The proposed statistical method using the Chi-square test classifies the oil samples with an accuracy of 100%, proving its superior performance over any other physical, chemical or electrical tests suggested in the literature.