An efficient learning based RFMFA technique for islanding detection scheme in distributed generation systems

This paper proposes an effective hybrid system for islanding detection of distributed generation (DG). The proposed control scheme is a united execution of both the Random Forest (RF) and Moth-Flame Optimization (MFO) named as RFMFO. The fundamental objective of the proposed work is to diminish the Non-Detective Zone (NDZ) to as close as could be allowed and to keep the output power quality unaltered. Moreover, the issue of setting the discovery thresholds in the current methods is overwhelmed by this strategy. For intelligent islanding detection, Rate of Change of Frequency (ROCOF) is used; by this methodology at the target DG location is utilized as the input sets for an RF system. So as to extricate various features among islanding and grid disturbance, the precision of the RF is prepared by MFO algorithm. In the proposed work, the RF is utilized to classify the islanding and non-islanding events subject to the extracted features. A few conditions and diverse loading, switching operation, and network conditions are resolved to approve the practicality of the proposed method. The proposed technique is implemented in MATLAB/Simulink working platform. The sampling point of voltage swell is 0-1.4x 10(4) N and 2.7 x10(4) - 4 x10(4) N, voltage sag has the sampling point of 1.4x 10(4)-2.7 x 10(4) N and the PCC has the sampling point of 2.1 x 10(4) - 4 x10(4) N. The performance of the DG is evaluated by using comparative analysis with the current methods. (C) 2020 Elsevier B.V. All rights reserved.