Comparative study of statistical breakdown in mineral, synthetic and natural ester oils under AC voltage

This paper reports on a comparative study of statistical Breakdown Voltage (BDV) in mineral oils, natural esters (vegetable oils) issued from different origins, and one synthetic ester through the measurements of 32 BDV using spherical electrode under AC voltage according to main lines of IEC 60156 specifications. The Normal and Weibull distribution functions are used to analyze and estimate the BDV probability of these oils. The hypothesis testing is then used to select the distribution functions suitability. We mainly analyze the histogram, frequency, mean, standard deviation and estimate the BDV probability of the sample population. The considered liquids are three different mineral oils (MO3', MO4' and MO5), one synthetic ester (SE1) and three different natural esters (VO2, VO3 and VO4). It's shown that the measured BDV data follow better Normal than Weibull distribution. Basing on the BDV at the levels of risk 1%, 10% and 50% probability for U1%, U10% and U50%, the dielectric withstand of ester oils, under AC, homogeneous field and small gap, is slightly better compared to mineral oils. PCA (Principal Component Analysis) method is also used to analyse the BDV variance of tested oils.