A Robust Fault Detection and Location Prediction Module Using Support Vector Machine and Gaussian Process Regression for AC Microgrid

Better reliability of power supply is assured with the inculcation of distributed generators (DGs) in a distribution network. Smart sensors and latest grid communication protocols have played a crucial role in the development of intelligent microgrids (MGs). Conventional protection schemes do not provide reliable performance when implemented in MGs. This article proposes an approach which requires root mean square value of one cycle three -phase voltage and current measurements during fault. These data are treated as inputs for developing a fault isolation and locator module. This module is supposed to be available at central protection system, and is designed using machine learning (ML) based techniques viz. Gaussian process regression for fault location prediction and support vector machine for fault identification. Effectiveness of the proposed methodology is evaluated by considering practical grid scenarios with load variation and different DG penetration level. Furthermore, the robustness of the proposed model is assessed by performing sensitivity analysis with consideration of variation in line parameters and load as well as effect of DG correlation. A 7-bus meshed ac MG test system consisting of three DGs and two grid sources is modeled in SIMULINK platform, and is used to demonstrate the proposed module. Data analytics tools of MATLAB 2020a has been explored to develop an ML-based fault isolation and location module for MGs. The proposed scheme has also been validated with real-time MG data obtained from OPAL Real time (OPAL-RT) real-time simulator OP-4510. The accuracy in predicted results proves that the proposed scheme is pertinent for real-time practical applications.