Dynamic State Estimation of Power System Utilizing Cauchy Kernel-Based Maximum Mixture Correntropy UKF Over Beluga Whale-Bat Optimization

Non-Gaussian noise, outliers, sudden load changes, and bad measurement data are key factors that diminish the accuracy of dynamic state estimation in power systems. Additionally, unscented Kalman filters (UKFs) based on correntropy criteria utilize bandwidth-sensitive Gaussian kernels, which may lead to singular matrices in the Cholesky decomposition. To overcome all the above problems, in this article, a robust UKF based on Cauchy kernel maximum mixture correntropy (CKMMC) criteria over hybrid Beluga Whale-Bat (BWB) optimization (BWB-CKMMC-UKF) is proposed, in which the kernel is merged of two Cauchy functions. Specifically, the measurement error and state error are unified in the cost function by the statistical linearization technique, and the optimal value of state estimation is obtained by fixed-point iteration. Because of its insensitive feature to kernel bandwidth and notable thick-tailed feature, the Cauchy kernel function is utilized instead of the Gaussian kernel in the optimization criteria. Additionally, to fit the power system model, the shape coefficients of the kernel in the CKMMC criterion and scale coefficients that influence the selection of sigma points in the unscented transform are determined based on the BWB algorithm. Simulation results on IEEE 14, 30, and 57-bus test systems validated the performance of the proposed algorithm.