Decentralized dynamic state estimation for multi-machine power systems with non-Gaussian noises: Outlier detection and localization.

In this paper, the decentralized dynamic state estimation (DSE) problem is investigated for a class of multi-machine power systems with non-Gaussian noises and measurement outliers. A model decoupling approach is adopted to facilitate the decentralized DSE for large-scale power systems. The particle filtering technique plays a key role in the developed DSE scheme with aim to tackle the nonlinearities and the non-Gaussian noises. To mitigate the negative impact from the measurement outliers on the DSE performance, a novel sliding-window-based online algorithm is proposed to detect and further locate the possible outliers based on the historical measurement data. Specifically, some criteria are constructed to (i) determine whether a newly arriving measurement vector is contaminated by measurement outliers and (ii) locate the abnormal components of such a vector. A conditional posterior Cramér–Rao lower bound is derived to evaluate the estimation performance of the proposed DSE algorithm. Finally, simulation experiments are carried out on the IEEE-39 bus system to verify the effectiveness of the proposed DSE algorithm under the non-Gaussian noises and the measurement outliers.