Estimating the remaining BDV for large generator stator winding by using nondestructive insulation diagnostic data based on statistical method

The actual stator bars of 18 kV/300 MW generator are used for accelerated multi-stress aging experiments to investigate the relationship between the remaining breakdown voltage (BDV) and the nondestructive insulation diagnostic data. At the end of aging experiment, AC breakdown test is done to the bars after their nondestructive data has been measured. The BDV data fits Weibull distribution. The correlation between dissipation factor, the PD parameter and breakdown voltage has been analyzed, among which the skewness of average discharge has the most correlation with breakdown voltage. Regression analysis is carried on using step-by-step multiple regression method and a mathematical expression is extracted, which can be used to predict the breakdown voltage.