Islanding detection based on nonparametric data reduction techniques and statistical quality control

In this paper, a wide-area, measurement-based islanding detection technique based on nonlinear, nonparametric data reduction methods and statistical quality control charts is presented. The method exploits the spectral properties of a probability transition matrix of similarities measures on updated observation window, and accounts for nonlinearities in the measured signals. When combined with a spectral change detection algorithm, changes in the measured data leading to island formation and deviations from data-driven (upper and lower) control limits are computed in an automated manner. Detailed time-domain simulations are conducted to test the ability of the method to detect islanding conditions.