Adaptive Robust Cubature Kalman Filter for Power System Dynamic State Estimation Against Outliers

Abstract

This paper develops an adaptive robust cubature Kalman filter (ARCKF) that is able to mitigate the adverse effects of the innovation and observation outliers while filtering out the system and measurement noises. To develop the ARCKF dynamic state estimator, a batch-mode regression form in the framework of cubature Kalman filter is first established by processing the predicted state and measurement data information simultaneously. Subsequently, based on the regression form, the outliers can be detected and downweighted by the robust projection statistics approach. Then, the adverse effects of innovation and observation outliers can be effectively suppressed by the generalized maximum likelihood (GM)-type estimator utilizing the iteratively reweighted least squares approach. Finally, an adaptive strategy is developed to adjust the state estimation error covariance matrix under different conditions. Extensive simulation results obtained from the IEEE New England 10-machine 39-bus test system under various operating conditions demonstrate the effectiveness and robustness of the proposed method, which is able to track the transients of power system in a more reliable way than the conventional cubature Kalman filter (CKF) and the unscented Kalman filter (UKF).