Automatic Detection of Transformer Health Based on Bayesian Network Model

Abstract

Abstract In order to effectively reduce the increasing operation and maintenance costs of aging power systems and infrastructure, the authors propose a real-time monitoring method of transformer operation state based on dynamic Bayesian network modeling and prediction uncertainty. The transformer fault mode, fault mechanism, different standards and codes, as well as the current transformer operation status are converted into component status, and then these statuses are transmitted to the real-time monitoring system of transformer operation status, the overall risk probability of the transformer or the subsystem risk probability of focus can be calculated according to the Bayesian network, and the elements in the transformer that may cause system failure or have operational risk can be supplemented through appropriate data processing and interpretation. In addition, on the basis of Bayesian network framework, continuous time steps can be added for continuous real-time monitoring of operation status, and a real-time monitoring system of transformer operation status based on dynamic Bayesian network can be built.