A passive islanding detection method based on K-means clustering and EMD of reactive power signal

Abstract

Abstract Distributed Generation (DG), though offers many advantages in making modern power systems smart ones, unveils many technical problems in terms of its control and operation. Islanding is one such problem which represents the loss of grid condition during which DG continues to feed the load. Due to its harmful effects, islanding has to be differentially detected from other power system transients in minimum possible time. Since the islanding event immediately affects the reactive power level in the system, such a reflection in reactive power signal at Point of Common Coupling (PCC) is utilized to extract the significant information regarding its occurrence. Hence a novel islanding detection method based on the signatures extracted from reactive power using Empirical Mode Decomposition (EMD) is proposed. The effectiveness of the proposed method is tested in different power system models developed in MATLAB/SIMULINK and the simulation results demonstrate the efficiency of the method in islanding detection within two cycles time Also, the results of various test cases of load parameter combinations evidence that this method has very near zero Non-Detection Zone (NDZ) with no effect on power quality. Further, the proposed method is validated through extensive simulations on IEEE 14 bus and applicability on PV systems are tested using modified IEEE 14 bus model. By suitably setting the detection threshold, the proposed method can distinguish islanding event from the non-islanding ones. For making the system autonomous, K means clustering is combined with EMD for the detection and classification of islanding from other power system transients.