Fault Classification in Microgrids using Deep Learning

In this work, two neural network models i.e. Long - Short Term Memory (LSTM) Networks and Convolutional Neural Networks (CNN) are employed to classify faults in microgrids. We used Matlab/Simulink to model a modified IEEE-13 bus feeder and simulate 11 types of faults to generate training and testing data. Additive White Gaussian Noise (AWGN) and Additive Impulsive Gaussian Noise (AIGN) are added to the data to make it closer to real-world data. The data is pre-processed using Discrete Wavelet Transform (DWT) and Multi-Resolution Analysis (MRA). The investigation showed that the LSTM network out-performed the CNN classifier and achieved high accuracy in classifying the faults using only one signal cycle of post fault voltage.