LAUSR

The minimum error entropy (MEE) criterion has been receiving increasing attention over the minimum mean square error (MMSE) criterion in non-Gaussian noise distribution, because it accounts for all higher order moments. In this brief, a novel MEE algorithm was proposed by using information theoretical learning concepts and the widely linear (augmented) complex domain modelling approaches for enhanced power system frequency estimation. The proposed augmented complex minimum error entropy (ACMEE) utilizes the complex-valued voltage signal, modeled by the Clarke’s $\mathbf {\mathrm {\alpha }}\mathbf {\mathrm {\beta }}$ transformation, which used all second-order statistical information for processing of non-circular complex-valued voltage signals. Performance degradation of the MMSE criterion in impulsive noise environments can be overcome by MEE adaptation scheme due to the higher order moments imbedded in its cost function. Therefore, the proposed ACMEE algorithm is able to achieve robust frequency estimation for unbalanced conditions and under the interference of measurement noises. The effectiveness of the ACMEE frequency estimation technique is verified through simulation studies of synthetic signals and experimental data.