Fault Diagnosis of Power Transformers Using GeNIe Modeler

Abstract

Fault diagnosis is an important task for technicians and engineers in detecting, isolating and identifying faults in systems. Previously, fault diagnosis and forecasting are mainly performed based on analytical models and expert’s experience. However, in practice, the derivation of an analytical model for a fault diagnosis process is difficult or impossible. In addition, as a given system has some degrees of uncertainty, there is a need of using a mathematical tool for handling this issue. Bayesian networks (BNs) are probabilistic graphical models that effectively deal with uncertainty and are widely used in fault diagnosis. Recently, there have been free and commercial tools for Bayesian network-based modeling and inference of system faults. Dissolved gas analysis (DGA) is a technique widely used in fault diagnosis of oil-immersed power transformers. This paper presents the use of Bayesian networks in GeNIe Modeler environment with DGA technique for conveniently deploying fault diagnosis models of oil-immersed power transformers.