Predicting transformer service life using simplified Perks' equation and Iowa curves

Facing the aging of its large electric transformer assets, PJM, a regional transmission organization (RTO) in eastern US, is testing a new method to model and predict transformers' service lives systematically and objectively. The approach is introduced in this paper. A Bayesian probabilistic method is used to model the uncertainty in the three parameters of Perks' distribution derived from a simplified form of Perks' equation to represent the hazard function of transmission transformers. This method is a major improvement from traditional Iowa curves. It gives a continuous family of transformer life distributions as well as the likelihood rather than a limited number of distribution patterns. The Perks' distribution can assume all the three distribution patterns of Iowa curves: right, left, and symmetrically modal. The method is illustrated in detail by applying to a single vintage data set of transformers. PJM is in the process of implementing this method to the statistics on hundreds of large electric transformers in PJM