Fast discrete s-transform and extreme learning machine based approach to islanding detection in grid-connected distributed generation

摘要

This paper presents an approach for islanding detection in distributed generation systems (DGs) using fast discrete s-transform (FDST) algorithm and bidirectional extreme learning machine (BELM) classifier. The system undertaken for this study comprises of two different kinds of DGs such as hydro turbine governor system and wind turbine generator. The analysis starts with extracting the non-stationary negative sequence voltage and current signals at DG end and then the instantaneous amplitude and frequency information are extracted through FDST algorithm. From this information, different distinguishing features are computed such as energy and standard deviation of amplitude to track the islanded event from different non-islanded events. The obtained features are examined through an extreme learning machine classifier to discriminate islanding and non-islanding events, under various operating conditions of distribution system. The accuracy of the proposed method is compared with other recently published techniques by various researchers to justify its improved performance for islanding detection.