Fault Isolation and Location Prediction using Support Vector Machine and Gaussian Process Regression for Meshed AC Microgrid

Better reliability of power is assured with the inculcation of distributed generations in a distribution network. Smart sensors and latest grid communication protocols has led to the development of intelligent microgrids (MG). Protection of such grids are a key concern for power engineers, as conventional protection schemes fails while operating in microgrids. This paper presents, data analysis based fault isolation and its location prognosis. The proposed approach takes post fault, one cycle three phase voltage and current measurements as the inputs for developing a fault isolation module (FIM) and fault locator module (FLM). These modules are supposed to be available at central protection system (CPS), designed through meticulous data training, using Gaussian process regression (GPR) for fault location prediction and support vector machine (SVM) for fault identification. The effectiveness of proposed methodology is represented by considering practical grid scenarios such as varying load and DG penetration. A 14 bus meshed AC microgrid structure has been modelled in SIMULINK having three DGs and two grid sources. Data analytics tools of MATLAB 2018a has been explored to develop machine learning based protection strategy for microgrids.