Islanding Detection Based on Artificial Neural Network and S-transform for Distributed Generators

Abstract

Recently, Distributed Generator (DG) connections in electrical power systems have been increasing, bringing benefits such as redundancy. However, it also presents some problems, such as islanding.The most commonly used methods to detect islanding are passive techniques, which are inexpensive and easy to be implemented. However, these techniques are very dependent on power unbalance between load and generation at the occurrence of an islanding to properly detect it. In order to mitigate this issue, new detection techniques based on machine learning have been receiving more attention in recent years. In this paper, a detection method based on the frequency spectrum of the voltage at the DG's terminals is proposed. The frequency spectrum is obtained by the S-transform. Afterwards an Artificial Neural Network (ANN) is used and the event is classified as being an islanding or not. The proposed method uses a common sample rate of 64 samples per cycle. Therefore, it might be easily implemented in digital relays already in operation. Additionally, the proposed islanding protection achieved a higher accuracy and a smaller detection time than conventional methods, with an average detection time of 26 ms.