Statistical insights into furan interpretation using a large dielectric fluid testing database

A large set of results for furan analysis of samples from oil-filled transformers above 500 kVA was statistically analyzed. Statistical correlations were calculated that may provide insight into the potential causes of the individual furans. Acid number, moisture saturation, and carbon monoxide were shown to be statistically significant predictor variables for total furan content. Various percentile values were also calculated for the distribution of and rate of change in furan concentration, and these are presented within subgroups subdivided by breather type and paper insulation type. Finally, equations are presented for using furan results to estimate the degree of polymerization and remaining life of the paper insulation.