Neural Network Modeling of Distribution Transformer with Internal Winding Faults using Double Fourier Series

An efficient transformer model is required to characterize the transformer internal faults for its condition assessment, which is experimentally very costly. This paper discusses the application of Neural Network (NN) techniques in the modeling of a distribution transformer with internal short-circuit winding faults. A transformer model can be viewed as a functional approximator constructing an input-output mapping between some specific variables and the terminal behaviors of the transformer. Neural network model takes fault specification and energized voltage as the inputs and the output voltage or terminal currents as the outputs. A major kind of neural network, i.e. back-propagation feed-forward network (BPFN), is used to model the faults in distribution transformers. The NN models are trained offline using training sets generated by a field based model, i.e. Double Fourier Series based field (DFSF) models. These models are implemented using MATLAB. The comparison between some simulation cases and corresponding experimental results shows that the well-trained neural networks can accurately simulate the terminal behavior of distribution transformers with internal short circuit faults.