Full Program »
A Bad Data Resilient Multisensor Fusion Framework For Hybrid State Estimation
This paper proposes a novel estimation fusion approach for hybrid multi-stage power system state estimation. Hybrid estimation fusion architectures allow the incorporation of new classes of sensors without doing away with pre-existing estimators. The fusion module is responsible to produce final estimates based on the optimal merging of the individual estimators' outcomes. The proposed fusion formulation is based on the Maximum Correntropy Criterion (MCC), an optimization criterion borrowed from the information theoretic area that has been recently applied to conventional power system state estimation and enables the automatic suppression of bad data effects. In this paper, the correntropy-based formulation is used to replace the traditional minimum variance fusion methodology in the second stage of the hybrid estimation algorithm. To assess the applicability of the novel method, a three-sensor, MCC-based fusion architecture is evaluated through simulations conducted on the IEEE 14- bus, 30-bus, 57-bus, and 118-bus test systems. In the absence of bad data, the results confirm the expected compatibility between the proposed MCC-based fusion and the conventional minimum variance formulation. In addition, extensive experiments considering distinct bad data scenarios under several operational conditions have been performed. They illustrate a novel functionality imparted by the proposed approach to the fusion module, namely, the ability to automatically reject discrepant data. Therefore, MCC fusion provides an extra layer of protection against the harming effects of individual estimates contaminated with gross errors.