Pattern Recognition Technique based Islanding Detection Scheme in Grid-connected PV System

A single phase network which is having photovoltaic (PV) components in an island state of decentralized Distributed Generating (DG) systems is presented in this paper. An efficient island classification technology i.e. Support Vector Machine (SVM) is developed to provide remote monitoring and operation of DG systems connected to the network. The voltage and current measurements at the common coupling point are the main sources of data for the development of classification algorithms. The proposed technique is used to extract features and create a feature matrix for all possible islanding schemes. The features matrix is influenced by machine learning method to create a trained dataset to help classify island scenarios. Also, a PV system connected to a single-phase network in the Matlab/Simulink environment (2015b) to determine the islands has been simulated. The results indicate that 98.6% of the training and 100 effectiveness tests were completed in 05 milliseconds, which is much better than conventional methods.