目标冲突下电力信息物理协同攻击分析

电力信息物理协同攻击(coordinated cyber physicalattack,CCPA)是智能电网面临的新型网络攻击之一,攻击者和调度中心目标分别为最大化和最小化攻击效果,两者目标相互冲突。以直流潮流模型为基础,首先建立基于虚假数据攻击的电力信息物理协同攻击数学表达式,用于表征量测单元中注入的虚假数据、攻击前后电力系统拓扑和电气参数之间的关系。然后,考虑攻击者的目标在于最大化攻击破坏效果,基于双层规划理论建立考虑攻击者和调度中心交互关系的电力信息物理协同攻击分析模型。最后,采用KKT(Karush-Kuhn-Tucker)条件和Fortuny-Amat-McCarl方法,将双层规划模型转化为单层混合整数线性规划模型。以IEEE14节点测试系统为例,仿真结果表明,电力信息物理协同攻击通过在量测单元中注入虚假数据重新分配母线负荷,并倾向于增加负荷需求较大的母线负荷,造成切负荷,最终影响电力系统的运行状态。与负荷重分配攻击和物理攻击相比,电力信息物理协同攻击不仅可以增加发电机出力成本和切负荷损失,而且可能导致更多线路过载。同时,量测单元在双层规划模型解中出现的次数可以表征该量测单元的脆弱程度,出现次数越多,则该节点越脆弱。

Analysis of Electrical Coordinated Cyber Physical Attacks Under Goal Conflict

Electrical coordinated cyber physical attack (CCPA) is one of the emerging cyber attacks in smart grid. The goals of attackers and dispatch centers are to maximize and minimize destructive consequences, respectively, conflicting with each other. Based on DC power flow model, a mathematical formulation of false data injection attack-based CCPAs is built to express the relationship among false data injected to measurement unit, topology and electric parameters before and after attacks. Furthermore, because the attackers’ objective is to maximize the damages of power system, a bi-level programming model of CCPAs is proposed to express the interaction between attackers and control center. Finally, the bi-level model is transformed to a single-level mixed integer linear programming problem using KKT (Karush-Kuhn- Tucker) conditions and Fortuny-Amat-McCarl method. The results of the bi-level model are demonstrated in IEEE 14-bus system, showing that the CCPAs redistribute loads in some buses by injecting false data to measurement unit and tend to increase loads in the buses with larger load demands, causing load shedding and affecting power system operation. Compared with load redistribution attacks and physical attacks, CCPAs increase generation cost and load shedding losses, and possibly cause more lines overloaded. Moreover, it is found out that the number of bi-level programming models containing specific measurement unit can reflect vulnerability of the unit, namely, the larger the number is, the more vulnerable the specific measurement unit will be.