On Predicting the Times to Failure of Power Equipment

Across power systems, large classes of identical devices can be found which support the system operation (transformers, breakers, switches, utility poles, etc.) The problem of their operational management is often aggravated by in-service failures and associated additional costs. Part of asset management strategy is to learn the failure characteristics of classes of devices in service and attempt to formulate the preventive replacement strategy based on that information. The paper presents an algorithm based on Bayesian learning which enables predictions of times to failure of identical devices to be refined with accumulated experience.