8th International Conference on Advanced Power System Automation and Protection (APAP2019)

961 / 2019-04-30 22:08:39

A Deep Learning Model for Transformer Short Turn-to-Turn Fault

Deep Learning,Transformer Fault,Protective Relaying

A Deep Learning (DL) model is designed to identify transformer short Turn-to-Turn Fault (TTF). First, in order to train the DL model, different features in the literature are reviewed. Then, a Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) architecture based DL model is proposed for transformer TTF. Due to high sensitivity for very low TTF current, negative sequence and wavelet transform of differential current are used as input data to train the model. The transformer model is developed in PSCAD/EMTDC environment while the DL model is executed using Keras with Tensorflow at the backend.