Improved extrema detection algorithm for the generalized empirical mode decomposition

Since the problem of monitoring the power system state based on the actual phasor measurements is relevant for multiple purposes, including identifying the low-frequency oscillations and taking the correspondent countermeasures, it is of a crucial importance to select the methods to be utilized on the basis of initial data features. Although modern wide-area monitoring systems provide information of high accuracy in terms of both systematic bias and random errors, numerous applications are subject to a negative influence of even minor measurement imperfections. One of the approaches being threatened by the random error component is the empirical mode decomposition, which was selected for the purpose of performing low-frequency oscillations monitoring for its evident advantages in dealing with the non-linear and non-stationary processes. In order to overcome potential drawbacks of applying the decomposition method to the initial data containing the random error component, a special extrema detection method is proposed based on the parabolas technique. Having been improved and optimized, it is characterized by relatively low computational burden and high quality results, which are demonstrated with the quantitative assessments.