Adaptive Cubature Kalman Filter for Power Systems Dynamic State Estimation in Face of Unknown Non-Stationary Noise Statistics

An adaptive Gaussian filter with deterministically chosen cubature points using spherical-radial cubature rule has been proposed in this paper for dynamic state estimation (DSE) of power systems where the process noise statistics remains unknown and time varying. This paper includes the online adaptation algorithm for process noise covariance (Q) based on the Maximum-likelihood (MLE) estimation method and its competing algorithm of Q-scaling. The estimation is carried out using the measurements from remote terminal units (RTU) as well as the phasor measurement units (PMU). Apart from this a state forecasting approach known as Holt's two-parameter technique is also employed to forecast the system states at the prediction step. The simulation is performed on standard IEEE 30 bus test system. The superiority of the Adaptive Cubature Kalman filter (ACKF) over its Non-adaptive version has been illustrated here using 200 Monte Carlo simulations. From their relative performance comparison, it is observed that the Adaptive cubature Kalman filter is recommended over Non-adaptive CKF (NACKF) because of its performance accuracy and numerical stability.