Classifier For Fault Detection Research Articles

Leveraging a micro synchrophasor for fault detection in a renewable based smart grid—A machine learned sustainable solution with cyber-attack resiliency

The advent of renewable distributed generation has led to the rethinking of the conventional protection systems, especially during fault. A sustainable fault detection algorithm is needed to enhance the distribution system's resiliency and safety. Keeping this point of view, the paper presents a micro synchrophasor or µPMU-based fault detection algorithm with fewer probabilities of cyber-attack. In the µPMU, after obtaining the current signals, the three-phase sequence components are evaluated and the angular sum of the positive and zero sequences is recorded. The maximum angular sum is fed to a trained machine learning (random forest) classifier for fault detection. The algorithm has a high resistance to cyber-attacks and is strongly immune to noise. The method has 98.91% accuracy, 99.89% precision and exhibits a detection time of 8.5 ms. The method is cost effective as it leverages a µPMU for fault detection algorithm, reducing the need of additional hardware and software. The method also proves to be superior than other fault detection methods in terms of accuracy, precision, detection time and the capability to handle noise and cyber-attacks. All the simulations are done in MATLAB/SIMULINK for a renewable based IEEE 13 node distribution test feeder.