| Popis: |
The occurrence of unexpected islanding is one of the major issues in the microgrid integrated distributed generation units. The islanding issue has to be immediately solved to protect the device from faults and power quality issues. Though several techniques are established in the recent days for the detection of islanding, the ultimate aim of detecting the fault is not achieved as these approaches generate a high risk of false detection. Hence, an effective and simple signal processing approach is proposed in this article. Initially, the grid signal is preprocessed with the implementation of Wiener filter, which performs efficient restoration of desired signal. The preprocessed signal is segmented with the assistance of DCT-DOST approach, which minimizes the time-locality. After the segmentation, the extraction of features is carried out by SIFT method, which estimates the feature descriptors by extracting single or numerous dominant orientations in every key point. Finally, the supervised classification is performed by PNN, which offers rapid training process in the absence of local minima. The proposed methodology is simulated and compared with other existing approaches. It delivers less training time of and testing time of . The obtained accuracy is for islanding conditions and for non-islanding conditions. |
