Data driven approach for fault detection and Gaussian process regression based location prognosis in smart AC microgrid

Modern intelligent digital relays embedded with phasor measurement units have facilitated easy access and collection of electrical signals in a microgrid. In this paper, a model for detection, location and isolation of faults in AC microgrid is presented. The real time voltage, frequency and current data are processed via simple formulations to detect the fault. The tripping command is promptly communicated to the incumbent line relays within fault tolerance time, after fault detection. The proposed scheme assures reliable protection in grid connected and islanded mode of operation with added security in case of primary protection failure. Post tripping, the exact location of fault in any distribution line from utility grid is predicted using a fault locator module. This module is designed and developed by gaussian process regression technique. Further, various machine learning techniques used for fault location has been explored and compared to establish the superiority of GPR method over others. Also the features extracted for model training is quite simpler, which imposes less computational burden on central protection computer. An IEEE 15 bus distribution system with synchronous and solar photo voltaic based DG is simulated in EMTP-RV platform to generate the electrical data. Further the required calculations are performed in MATLAB 2020a