Permutation Entropy Thresholding: A Non-Linear Signal Processing Method for Islanding Detection

Abstract

ABSTRACT A novel passive method based on Permutation Entropy (PE) is proposed for fast, efficient, and differential detection of islanding from other disturbance events in power system. The extremely remarkable attributes of PE such as efficient and fast detection of change in dynamics as well as computational and conceptual simplicity can be appropriately utilized to devise a thresholding system that can differentiate the change in dynamics of islanding event from the non-islanding ones with negligible Non-Detection Zone (NDZ). As NDZ is the most critical issue of existing passive islanding detection methods, which overshadows its beneficial characteristics, this new approach can prove to be a suitable alternative with an added benefit of faster detection. The method is evaluated with grid connected single as well as multi-DG systems and its robustness against effects of load quality factors, active and reactive power mismatches is investigated. Results show that this method can detect the islanding disturbance and differentiate it from other disturbances within half a cycle. The performance of the proposed method is evaluated with photovoltaic (PV) sources and also validated on IEEE 14 bus model. Results indicate the suitability of the PE method for the common DG source of PV for a wide range of power mismatches including the worst cases and also its robustness against false tripping during DG switching. The results verify the credibility and advantage of the proposed PE thresholding method and its suitability for adopting into existing networks for passive islanding detection.