Data mining model and Gaussian Naive Bayes based fault diagnostic analysis of modern power system networks

Abstract

In this paper, different machine learning (ML) algorithms like Gaussian Naïve Bayes (GNB), K-Nearest Neighbours (KNN), Decision Tree (DT), Support Vector Machine (SVM), Random Forest (RF) and Adaptive Boosting Classifier (ABC) are used to classify the different types of faults on the transmission line. The faults and non-faults on the transmission line are created by simulating the transient behaviour of the long length transmission line by using PSCAD/EMTDC software. While simulating the faults and non-faults on the transmission line, fault resistance, fault inception angle, fault location and fault type are varied within their practical range. During the simulation, currents and voltages of three phases on both sides of the TL are measured using Current Transformers (CT) and Potential Transformers (PT) . Measured currents and voltages are pre-processed to produce the time series and frequency series parameters data set by using Fourier Transform in full cycle, half cycle, and quarter cycle from the fault instant. Feature reduction techniques select very important parameters among the calculated time series and frequency series parameters. After completing the data generation and pre-processing and application of feature reduction techniques, the machine learning algorithms are applied to the data in python software to get the comparative analysis. Noise analysis is also carried out in this work to assess the performance of the ML algorithms in dealing with the nonlinear data due to CT, PT saturation, and measurement errors.