Comparison of Different Machine Learning Algorithms and Neural Network for Detecting and Classifying Electric Faults

Abstract

Faults in power systems are common, and they may lead to the devastation of costly parts like motors, generators, and transformers. Additionally, these defects have the potential to result in deadly events including explosions, overvoltages, excessive currents that are chosen outages, and even deaths. A power protection system is required to quickly identify, classify, and locate the defects, reducing their effect, in order to remedy these problems. For the purpose of providing constant power supply, minimizing interruptions, and avoiding equipment damage, it is essential to analyze power system failures. This technical report provides a thorough technique for locating, classifying, and identifying power system issues.For accurate fault identification, classification, and fault location determination, the suggested method makes application of machine learning techniques like Convolutional Neural Networks (CNN) and K-Nearest Neighbors (KNN), Random Forest (RF), and Decision Trees. models. Additionally, the technique divides power quality into five different groups.. The goals of identifying defects in power systems, classifying them, pinpointing their locations, and grading power quality are all achieved by this thorough technical explanation. This study employs the CNN, K-Nearest Neighbors, RF, and decision tree algorithms developed in MATLAB/SIMULINK to identify, classify, and localize power system failures on the Test Network