Widely linear least mean kurtosis‐based frequency estimation of three‐phase power system

Abstract

We propose a widely linear (augmented) least mean kurtosis (WL-LMK) algorithm for robust frequency estimation of three-phase power system. The negated kurtosis-based algorithms are most celebrated for their computational efficiency and strong robustness against wide range of noise signals which can overcome the inherent performance degradation faced by the well-known minimum mean square error-based algorithms in noisy environments. The proposed widely linear LMK estimation technique utilises all second-order statistical information in the complex domain C for processing of non-circular complex-valued signals. The three-phase power system signal, modelled through Clarke's αβ transformation, is circular for balanced and non-circular for unbalanced systems, based on which, the proposed WL-LMK algorithm is able to achieve improved frequency estimation under unbalanced and other abnormal system conditions. Its estimation performance is evaluated for several cases that encounter in the day-to-day operation of power system. It is observed from simulation studies of synthetic and real-world power system data that the proposed WL-LMK algorithm exhibits superior estimation performance as compared to the standard linear complex LMK (CLMK) and the widely linear least mean square (WL-LMS) algorithms.