Neuro-fuzzy islanding detection in distributed generation

Islanding detection methods are divided into three main groups as remote, active and passive. Although passive schemes have larger Non Detection Zones (NDZ) relative to other schemes, they are more used in utilities due to their low costs and less PQ problems than other schemes. Passive Schemes are based on the measurements of passive system parameters. These parameters are measured at the point of common coupling (PCC). A new approach in passive techniques is the use of data-mining to classify the system parameters. In this paper, massive indices are collected by simulation of a practical distribution system in PSCAD/EMTP environment. These indices include voltage, frequency, current, active power and etc. The classifying process of these indices is done by the Adaptive Neuro-Fuzzy Inference System (ANFIS) in MATLAB and the resultant logics and boundaries are implemented by the fuzzy logic using MATLAB software. The results show the effectiveness of ANFIS in reducing the NDZ of passive islanding detection schemes. In addition, this technique can be easily implemented with minor changes to distribution systems with different penetration levels and types of Distributed Generation (DG) as well as different distribution system topology.