Transient Thermal Modeling of Split-Winding Transformer based on Finite Difference Method and Artificial Neural Networks

Nowadays, multi-winding transformers are widely used in traction systems and steel manufacturer. Multi-winding transformer that is used in these systems has especial geometry (that is usually called as split-winding transformer) and encounters with serious magnetic and thermal difficulties. According to the special geometry of split-winding transformer, studding the transient thermal behavior of their windings is very important. In this paper, a finite difference based method is introduced for mathematical modeling of windings' thermal behavior in transient state. After validating the mathematical model with experimental results, a typical loading is applied to the split-winding transformer and windings' temperatures are sampled in different time periods.Afterwards, by introducing artificial neural networks and training their parameters with thermal data gathered in the first section of the paper, two novel dynamic models have been proposed for windings thermal analysis in the split-winding transformer. As it has been shown, the introduced neural network based thermal models have good precision in transient thermal modeling of split-winding transformer.