Variational Bayesian Unscented Kalman Filter for Active Distribution System State Estimation

Real-time monitoring and control of distribution networks relies on a robust distribution system state estimation (DSSE). The use of pseudo measurements, typical for DSSE, may negatively affect estimation accuracy as their uncertainties are high. Increased integration of intermittent renewable generation makes active distribution networks more prone to sudden state changes. To overcome these challenges, this paper proposes a variational Bayesian Unscented Kalman Filter (VBUKF). By efficiently adapting the prediction error covariance matrix and measurement noise covariance matrix, VBUKF copes with unpredictable sudden state changes and bad data, as well as unknown measurement noise. The proposed VBUKF makes use of a vector autoregressive process to capture temporal and spatial correlations in system states and improve prediction accuracy. Extensive simulations are conducted on three IEEE test systems with PV generations to demonstrate the performance of the proposed VBUKF in terms of estimation accuracy, convergence speed, numerical stability and scalability. Results obtained are compared with state-of-the-art state estimation algorithms to highlight the advantages of the proposed approach.