Machine Learning Technique for Faults Identification in Microgrid

Abstract: In this paper a machine learning technique-based method is presented for identifying faults in a microgrid. The machine learning method considered is multilayer neural network. The microgrid model includes renewable energy sources such as solar photovoltaic and wind generator. Both normal and fault situations for the microgrid are simulated when operating in grid-connected and islanded modes. The fault conditions include various faults experienced on a power distribution line. The fault currents and voltages are used for training a neural network with Levenberg-Marquardt method. The trained neural network is effective in identifying and classifying faults using trained and untrained data for both grid-connected and islanded modes of operation.

Keywords: Machine Learning, Microgrid, Neural Network, Protective Relaying, Fault Detection.