Stealthy monitoring-control attacks to disrupt power system operations

Abstract

We propose in this paper a model to quantify the consequences of the vulnerabilities of the AC-based state estimation against data-integrity cyber attacks. We consider attacks whose goals are to corrupt the state estimator and/or the operation of the power grid. By considering AC-based models of the state estimator and the operation of the power system, the proposed model allows quantifying the reliability consequences (i.e., alteration in the active and reactive power), and the economic consequences (e.g., distorting locational marginal prices) of such attacks. We consider a resource-constrained omniscient attacker intending to alter the operation of the power system. We impose a stealthy condition on the attacker so that its malicious action is invisible to the bad-data-detection algorithm. The upper-level problem of the proposed bilevel model represents the attacker’s action and pursues disrupting the system operation and it is constrained by two lower-level problems. The first lower-level problem is the state-estimation that calculates the most likely state of the system considering the corrupted measurement. The second lower-level problem is a conic relaxation of the optimal power flow that uses the output of the state estimation problem to minimize operation costs. We illustrate the proposed bilevel model using two test systems.