Enhancing Fault Detection in Smart Microgrids using Centralized Artificial Neural Networks

Abstract

Smart microgrids (SMGs) face specific difficulties in detecting fault currents in various operational scenarios. Traditional protection methods are not accurate to detect faults in all scenarios. This paper presents a novel centralized artificial neural network (ANN) technique to detect faults in different operational scenarios with different fault resistances at different fault locations. The core idea involves fault detection by analyzing the three-phase voltage and current measurements detected by a centralized ANN within the microgrid. The three-phase measurement signals collected by phasor measurement units (PMU) are fed into ANN, to extract fault features. These features are then processed by fully connected layers in the ANN to identify fault type and phase. The proposed method is relatively simple to implement while providing a high degree of reliability. The proposed approach's performance and accuracy are extensively evaluated through simulations conducted on SMG in MATLAB/Simulink incorporating distributed energy resources (DER) types such as solar photovoltaic systems. The accuracy of the proposed approach is 97.14% where the fault can be cleared within 10.6 ms.