Islanding Detection with Empirical Mode Decomposition and Random Subspace Oriented Kth Nearest Neighbour

Abstract

In the recent past, introduction of renewable energy resources (RERs) as an alternative to conventional power production has increased the penetration of Distributed Generators (DGs) to the existing power system. While resolving the issue of inadequate power delivery, the independency of DGs' put forward another worry in front called unintentional islanding. In this context, the proposed work is willful for a solid inclusion to the contemporary literatures based on detection of non-intentional islanding. A novel application of empirical mode decomposition (EMD) along with random subspace oriented kth-nearest neighbour (RSOKNN) is suggested to identify the unintentional islanding problem. At first, the voltage signal is collected from the PCC of the studied model and undergone mode decomposition process to extract five different features. Secondly, the RSOKNN machine learning model is utilized for efficient islanding identification. The proposed method is verified under ideal and noisy condition as well as compared to other competitive classifiers to determine the supremacy.