Design, Development and Testing of a Support Vector Machine based Relay for Islanding Detection

The modernization of power electronics has made the integration of distributed generators more viable in distributed power systems. However, there is a high chance of unintentional island formation in a DG penetrated power system and it is not preferred to energize the local loads during the formation of unintentional islands. Therefore the detection of unintentional islands is an important concern in the protection system. An islanding detection method based on Support Vector Machine, a supervised learning technique is used in this paper. The voltage and current signal at the point of common coupling are used as the parameters for classification. The Discrete Wavelet transform, a signal processing tool is performed on the voltage and current signals for extracting the features that can be further used to train the SVM to distinguish islanding from other test scenarios. The test system used for complete analysis is simulated in PSCAD. The proposed islanding detection technique is validated for several operating conditions and various test cases are considered.