A Synchro-phasor Assisted Optimal Features Based Scheme for Fault Detection and Classification

A novel and efficient methodology for comprehensive fault detection and classification by using synchrophasor measurement based variations of a power system is proposed. Presently, Artificial Intelligence (AI) techniques have been used in power system protection owing to the greater degree of automation and robustness offered by AI. Evolutionary techniques like Genetic Algorithm (GA) are efficient optimization procedures mimicking the processes of biological evolution that have been shown to perform better than their gradient based counterparts in many problems. We propose a combined GA and Particle Swarm Optimization (PSO) approach to find the optimal features relevant to our fault detection process. As is evidenced by recent advances in multi-modal learning, it has been shown that this combined approach yields a more accurate feature optimization than that obtained by a single meta-heuristic. A systematic comparison of Artificial Neural Network (ANN) and Support Vector Machine (SVM) based methods for fault classification using the identified optimal features is presented. The proposed algorithm can be effectively used for real time fault detection and also for performing postmortem analysis on signals. We demonstrate its effectiveness by simulation results on real world data from the North American SynchroPhasor Initiative (NASPI) and signal variations from a test distribution system.