Data Mining Approach to Fault Detection and Analysis in Grid-connected Micro-Grid using Bayesian Network

Abstract

The concept of distributed generation has enabled the development and integration of renewable energy into existing power systems. Different types and modes of operation of distributed generation represent serious obstacles for the protection system. It is necessary to develop a protection solution that would respond to all challenges and facilitate the integration of renewable energy sources. This work aims to present a Data Mining based approach to developing protection schemes. First, a thorough fault analysis of a system with high penetration of distributed generation has been done, including a set of simulations to generate the required data. Second, different feature selection methods for local-based decision-making have been analyzed and compared in order to obtain optimal feature subsets. Further, the potential of Bayesian Networks in fault detection has been explored with special focus in the area of communication-assisted configurations. The fault analysis and data collection were done by time domain simulations using MATLAB/Simulink. Feature selection and classification were conducted using the open sourced software package WEKA and the evaluation version of the software package BayesiaLab.

Date of Award	Dec 2013
Original language	American English
Supervisor	Wei Lee Woon (Supervisor)