Optimal parameterization of Kalman filter based three-phase dynamic state estimator for active distribution networks

This paper presents a new method for the assessment of the process noise covariance matrix for three-phase dynamic state estimation in unbalanced active distribution networks which operate under normal conditions. The assessment is done in order to minimize the estimation error. The proposed assessment method, based on minimization of a particular cost function, enables the a priori assessment of covariance matrix by extracting information from previously observed measurements, without the need to simulate the true state of the system. The method was applied on two commonly used Kalman filter based estimation algorithms in nonlinear systems: Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF). A comparative analysis was performed between two different cost functions based on average root mean square of innovations and maximum likelihood technique. Also, the importance of determining initial state vector and its error covariance matrix needed for the initialization of dynamic state estimation was examined, as well as the ability of UKF and EKF to handle measurement nonlinearities. The analysis was carried out and the proposed method was verified on modified IEEE 13- and 37-bus distribution test systems.