EWDGA and Markov process based failure rate estimation of transformer internal latent fault

This paper proposed a novel failure rate estimation model of transformer internal latent fault. The concept of EWDGA is introduced from the thermodynamic view, which is proved to reflect the extent of fault development effectively. Multi-state Markov process model based on EWDGA is proposed which integrated the historical and real-time data. Numerical examples testify the effectiveness of the model, which can distinguish the failure rate of different transformers in the same state but with different fault development extent. This research can help the decision of the transformer maintenance strategy.