Design of Linear Deception Attacks Against State Estimation in Multiplant Power Systems

In this article, we address the problem of designing linear deception attacks against multiplant power systems. In the existing deception attack schemes, only sensor measurement channels are regarded as targets of attackers. Here, attack strategies against the edge measurements in power systems are provided as well to achieve a larger estimation error. Moreover, it is found that attack signals with a dependent nature perform better than the independent ones under the same stealthiness constraints. For dependent attacks, a necessary and sufficient condition to guarantee the unboundedness of the estimation error was obtained. Then, the problem of designing dependent attacks is further analyzed. Different suboptimal attack strategies are provided by solving nonconvex problem or transforming the nonconvex problem into a convex one. It is worth noting that all the attack signals concerned in this article are independent of real-time data of system operation, thus can be designed offline and be easier for application. Finally, simulation results are provided based on the IEEE 4-bus system to demonstrate the feasibility and the efficiency of theoretical results. Comparisons of different attack schemes are provided as well.