A Multifeature-Based Approach for Islanding Detection of DG in the Subcritical Region of Vector Surge Relays

Anti-islanding protection is an important requirement which has to be considered prior to the integration of distributed generation into electricity grids. Conventional vector surge (VS) relays are usually used to detect islanding; however, there is a nondetection zone (NDZ) wherein islanding incidents are undetectable by VS relays. This paper proposes a multifeature-based technique for islanding detection in the subcritical region, defined as a subregion of the NDZ. In the proposed method, features are extracted from five network variables. The extracted features are then used as inputs to a support vector machine to classify the event as islanding or nonislanding. A test network was used to generate a large number of islanding and nonislanding events with different load types. The proposed method is tested with the most critical islanding cases associated with NDZ of VS relays. Furthermore, all possible combinations of deficit and excess of active and reactive power imbalance, which may exist during the occurrence of an island, are considered in the testing phase. Experimental results demonstrate that the proposed method can successfully detect islanding events in the subcritical region, where a VS relay is expected to fail.