Electric power systems have evolved over the past century to the largest and the most complex cyber-physical systems. With the development of... Show moreElectric power systems have evolved over the past century to the largest and the most complex cyber-physical systems. With the development of Smart Grid, cyber security has become an area of growing concern. False data injection attack has recently attracted wide research interest. This thesis develops the concept of load redistribution attack, a special type of false data injection attack. The physical and economic impact of load redistribution (LR) attacks is quantitatively analyzed in this thesis. Since LR attacks can successfully bypass bad data detection and manipulate the state estimation outcome, security constrained economic dispatch (SCED) based on the false state estimation would lead the system into a non-optimal or insecure operation state. Based on the consequence analysis, two different attacking goals are differentiated from the adversary’s perspective, i.e., immediate attacking goal and delayed attacking goal. For the immediate attacking goal, a max-min attacker-defender model is proposed to identify the most damaging immediate LR attack. Two different algorithms are used to solve this bi-level optimization problem. For the delayed attacking goal, a tri-level model is proposed to identity the most damaging delayed LR attack. This thesis studies the economic impact of LR attacks on power market operation. A convex model is proposed under the mechanism of virtual bidding to compute the optimal injection of LR attack, which gains the most profit from the attackers’ perspective. The quantitative analysis of LR attacks provides an in-depth insight on effective attack prevention with limited protection resource budget. This thesis proposes a new protection strategy, which intends to mitigate the most damaging effect of LR attacks on power system operation. The criterion of determining effective protections against the most damaging LR attack, considering the existence of stochastic measurement errors, is deduced. Ph.D. in Electrical Engineering, May 2012 Show less
