Hassan Ghadimi - Department of Engineering, Ardabil Branch, Islamic Azad University, Ardabil, Iran
Homayoun Ebrahimian - Department of Engineering, Ardabil Branch, Islamic Azad University, Ardabil, Iran

Due to increase distributed generations, Islanding is an important concern for these resources. Personneland equipment safety issues are main reasons to detection of islanding. Several techniques based onpassive and active detection schemes have been proposed previously. Although passive schemes have alarge non-detection zone (NDZ), concerns have been raised about active methods because of theirdegrading effect on power quality. Reliably detecting this condition is regarded by many as an ongoingchallenge because existing methods are not entirely satisfactory. This paper proposes a histogramanalysis method using a Multi Layer Perceptron (MLP) Neural Network approach for islanding detectionin grid-connected wind turbines. The main objective of the proposed approach is to reduce the NDZ andto maintain the output power quality unchanged. In addition, this technique can also overcome theproblem of setting detection thresholds which is inherent in existing techniques. The method proposed inthis study has a small non-detection zone and is capable of detecting islanding accurately within theminimum standard time. Moreover, for those regions which require better visualization, the proposedapproach can serve as anefficient aid for better detecting grid-power disconnection.

Signal histogram, MLP,distributed generation,islanding detection, nondetectionzone