Evaluation of statistical methods used in the estimation of breakdown voltage distribution

Abstract

Multiple-level and up-and-down test methods are commonly employed in laboratories to obtain the breakdown voltage distribution of self-restoring insulation, which is applied for equipment evaluation and for risk calculation in insulation coordination studies. This study investigates the reliability of the methods and proposes a methodology for evaluating their performance. The accuracy and precision of the estimated distribution parameters were analysed considering the variation of the test parameters. Algorithms that simulate the random nature of the tests were developed and applied in simulations. Then, a case study of risk calculation was carried out. The results indicated that both methods estimate the critical flashover voltage with acceptable accuracy and precision, even for the minimum parameters defined by standards. However, estimating the standard deviation is the critical step of the tests. The proposed methodology allows parameter definition for a given reliability requirement in the case of the constant voltage method. In the case of the up-and-down method, however, high dispersion was observed for all analysed cases, which indicates that this method presents low reliability for calculating flashover risk and, therefore, its use for insulation coordination studies is not recommended.