Robust unscented Kalman filter based on minimum error entropy with fiducial points utilizing generalized Versoria-Gaussian kernel to forecasting-aided state estimation for power systems

As an outstanding forecasting-aided state estimation method for power systems, unscented Kalman filters (UKF) based on information theoretic criteria have been widely applied in recent years. In this paper, a robust UKF based on minimum error entropy with fiducial points utilizing generalized Versoria-Gaussian kernel (R-GVG-MEEF-UKF) is proposed to overcome non-Gaussian noise and outliers, sudden load changes, and bad measurement data. Specifically, the statistical linearization technique is applied to merge the measurement and state errors in the cost function and through fixed-point iteration to obtain the state estimate value. At the same time, to solve the problem of the influence of kernel shape coefficients, a framework for automatically searching for the optimal value ​​of these coefficients is developed. In addition, the QR decomposition method is utilized to ensure the condition of the Cholesky decomposition. Finally, through IEEE-14,30,57 bus test systems, the numerical results have confirmed the high accuracy of the proposed algorithm compared with the existing algorithms.