Generalized attack separation scheme in cyber physical smart grid based on robust interval state estimation

Abstract

This paper proposes a new generalized attack separation scheme in energy internet to analyze how an attacker behaves. This scheme consists of a robust interval state estimation-based anomaly detection method and a coordinated bad data processor. We develop an anomaly detection method that takes system uncertainties, i.e., forecasting uncertainty and parameter perturbation, into account to approximate the largest possible deviations of state variables in smart grid, thus recognizing stealthy anomalies. In this method, we decouple the power balancing formulation into P-θ and Q-V sub-problems to improve the computational efficiency. Thereafter, a new coordinated bad data processor is proposed to automatically filter out the impact of malicious attacks on state estimation and to reconstruct the real states of a cyber physical smart grid. In this processor, three-stage state estimator is applied to rapidly assess the biased state from the current observations with the contamination of meters and communication channels. As a result, the implemented attack vector can be separated and isolated from the current observations online. Finally, the proposed scheme is validated via comprehensive tests on several IEEE systems. The obtained results show the strong robustness, high stability and promising performance for attack separation, indicating a high potential of the proposed scheme for practical implementations in deep-cyber penetrated smart grids.