A novel SVM approach of islanding detection in micro grid

Passive islanding detection schemes are more used in utilities due to their low costs and less harmonic problems although having larger Non Detection Zones (NDZ) relative to other schemes. Passive system measurements at the point of common coupling (PCC) are the basis of this scheme. A new approach in passive techniques is the use of intelligent based methods in data-mining to classify the system parameters which affect the islanding detection. As a result, by finding an efficient and robust data-mining algorithm, the passive schemes problem, which is their relatively large NDZ, will be minimized. In this paper, massive measurements are collected by simulation of IEEE standard distribution system in PSCAD/EMTP environment. These indices include current, voltage, frequency, active power and frequency over active power. The classifying process of these indices is done by the Support Vector Machine (SVM) using MATLAB software. The results show the efficiency with good speed of SVM in passive islanding detection schemes. Also, this paper studies the best two indices which can be used for islanding detection. SVM technique can be easily implemented in the Micro Grid with different types and penetration levels of Distributed Generation (DG).