Transformer fault diagnosis based on probabilistic neural networks combined with vibration and noise characteristics

When the transformer is running, vibration which generated in core and winding will spread outward through medium of metal, oil, air. The magnetic field changes with variation excitation source state core, so corresponding noise change. Therefore, contain a lot information. If information can be associated fault characteristics transformer, it significant to evaluate running signal, improve intelligence, safety, stability operation. Based on this, modeling simulation multi-point grounding, DC bias, short-circuit between silicon steel sheets are first carried out this paper, distribution under different faults given. Second, diagnosis method based proposed. In process implementation, signals taken as sample data, probabilistic neural network algorithm used effectively predict fault. Finally, effectiveness proposed scheme verified by identifying faults-the PNN applied transformer.