Improved Islanding Detection With h‐EXPO‐SV ML Algorithm Supporting Positive Sequence Impedance Angle in an Active Distribution System

ABSTRACT

Islanding detection has become a serious concern due to the extensive integration of renewable energy sources with the main grid. This article addresses two significant challenges with islanding detection schemes: the non‐detection zone (NDZ) and system‐specific applicability. The purpose of this work is to provide an islanding detection technique that operates at zero NDZ and is adaptable to all types of renewable energy sources. This paper presents a novel passive islanding scheme to detect islanding events in an active distribution system based on the rate of change of the superimposed positive sequence impedance angle. The positive sequence impedance angle, in its derivative form, checks the threshold value in terms of phase angle with the assistance of the proposed machine learning (ML) algorithm, which further distinguishes islanding and non‐islanding zones. The proposed algorithm for identifying the islanding detection zone (IDZ) is named the heuristic expo support vector (h‐EXPO‐SV) ML algorithm, which is a combination of exponential distribution and support vector machine. The proposed controller is simple, robust, and requires less computational data. The proposed ML approach uses a positive sequence impedance angle to identify IDZ. The MATLAB/Simulink framework simulates the IEEE 13‐bus test system. When the power generated and load power are equal, the proposed approach detects islanding within 8 ms. By effectively distinguishing between islanding and non‐islanding occurrences, the suggested ML‐based approach aids in preventing preventable DG faults. Several test systems have evaluated the suggested algorithm's effectiveness, which is superior to that of traditional methods. The proposed h‐EXPO‐SV ML algorithm successfully differentiates between islanding and non‐islanding events, which helps to prevent avoidable DG malfunctions. The performance of the proposed algorithm has been tested in several test systems, and it is better than some conventional techniques.