Principal Component Analysis of Wide-Area Phasor Measurements for Islanding Detection—A Geometric View

This paper presents a new technique for the detection of islanding conditions in electrical power systems. This problem is especially prevalent in systems with significant penetrations of distributed renewable generation. The proposed technique is based on the application of principal component analysis (PCA) to data sets of wide-area frequency measurements, recorded by phasor measurement units. The PCA approach was able to detect islanding accurately and quickly when compared with conventional RoCoF techniques, as well as with the frequency difference and change-of-angle difference methods recently proposed in the literature. The reliability and accuracy of the proposed PCA approach is demonstrated by using a number of test cases, which consider islanding and nonislanding events. The test cases are based on real data, recorded from several phasor measurement units located in the U.K. power system.