Analysis of Targeted Coordinated Attacks on Decomposition-Based Robust State Estimation

Abstract

The impact of false data injection (FDI) attacks on static state estimation of power systems has been actively studied in the past decade. In this paper, we consider an estimation method that first decomposes the system into islands and then implements robust regression estimators at the island level as well as the system level. We carry out an analysis to establish its advantages in terms of state estimation accuracy and attack detections. In particular, we focus on highly adversarial cases where the attacker can attack both the measurement vector and the regressor matrix and attempts to manipulate the states to targeted values. Our estimation approach employs a system decomposition method capable to generate islands small in their sizes and applies the robust estimation method of least trimmed squares. We make comparisons with methods using other decompositions and other robust estimators. To this end, we analyze the structure of the system topology and measurements and perform extensive simulations using the IEEE 14- and 118-bus systems. Furthermore, we investigate robustness improvement when phasor measurement units (PMUs) are available and hybrid state estimation can be employed.