摘要
针对智能电网中虚假数据注入攻击(false data injection attack, FDIA)能够躲避传统坏数据检测的问题,本文研究了交流模型下研究FDI攻击方式,以历史状态数据间的内在关联为基础,构建了基于改进极限学习机(extreme learning machine, ELM)的预测模型。结合电网拓扑特定的电气关系,对比量测设备传输数据,对离群值进行类别标签。根据异常分布结合图论计算节点强度判定异常并检测攻击。在IEEE-14节点系统进行仿真,所提方法能快速检测并定位到异常节点,验证了所提方法的可行性和有效性。
To address the problem that false data injection attack (FDIA) can evade traditional bad data detec-tion in smart grid, this paper investigates the way to study FDI attack under AC model and con-structs an improved extreme learning machine (ELM) based on the intrinsic correlation between historical state data. The outliers are labeled with categories by comparing the transmission data of the measurement devices in combination with the electrical relations specific to the grid topology. Node strength is calculated based on the anomaly distribution combined with graph theory to de-termine anomalies and detect attacks. Simulations are performed on the IEEE-14 node system, and the proposed method can quickly detect and locate the anomalous nodes, which verifies the feasi-bility and effectiveness of the proposed method.
出处
《建模与仿真》
2023年第2期799-812,共14页
Modeling and Simulation
