An Islanding Fault Detection Method with CFDF-SVM Based RPV Approach under Pseudo Islanding Phenomenon

Abstract

Islanding detection is an important issue in distributed generations (DGs) systems. Therefore, anti-islanding protection is a critical concern. A novel reactive power variation (RPV) islanding detection method based on correlation function data fusion and support vector machine (CFDF-SVM) approach is presented to reduce the non-detection zone (NDZ) and avoid the false detection caused by pseudo islanding phenomenon (PIP). The feature data sets (frequency, injected reactive power, fundamental and third harmonics of point of common coupling (PCC) voltage and current) are acquired by Fourier transform and fused by CFDF approach. After that, the fused data sets are classified into two categories by an SVM approach: islanding and non-islanding. In addition, the proposed approach is based on the active intermittent RPV method which reduced the NDZ fundamentally. Compared with the traditional RPV method, the proposed method could detect islanding event accurately, and the false detection rate caused by PIP is significantly reduced. Then a single-phase-inverter is built by Simulink, the detection results have proved the effectiveness of the proposed method.