LAUSR

This article offers an islanding detection approach based on dual-tree complex discrete wavelet transform (DTCDWT) and extreme gradient boosting method (XGBoost). Initially, the DTCDWT is used to extract the feature vector from collected negative sequence voltage and current signals at the respective distributed energy resource ends. Afterward, the nonlinear synthetic minority oversampling technique (SMOTE)-based XGBoost is used to classify the islanding (I) and nonislanding (NI) events with the following characteristics: high accuracy, good sensitivity, and faster speed. Here, the SMOTE is used as an offline prefeature processing approach to reduce the unbalancing/uncertainty of “I” and “NI” classes and helps the intelligent classifier to impart a fair classification result. This scheme is found to be effective for the detection of I and NI conditions, satisfying the IEEE islanding detection standard. Additionally, the obtained simulation results are compared with other existing works to show its usefulness.