基于RNN的智能电网拓扑变异型FDI攻击检测方法  

Detection method based on RNN for topology variation FDI attacks in smart grid

作　　者：王海吉 胡健坤 田元 WANG Hai-ji;HU Jian-kun;TIAN Yuan(Guangdong Electric Power Design Institute Co.Ltd.,China Energy Engineering Group,Guangzhou 510660,China;Southern Energy Business Department,Guangzhou Haiyi Software Co.Ltd.,Guangzhou 510600,China)

机构地区：[1]中国能源建设集团广东省电力设计研究院有限公司,广州510660 [2]广州海颐软件有限公司南方能源事业部,广州510600

出　　处：《沈阳工业大学学报》2023年第2期139-144,共6页Journal of Shenyang University of Technology

基　　金：国家自然科学基金青年科学基金项目(61863113)。

摘　　要：针对近年来对智能电网运行状态构成严重安全威胁的虚假数据注入问题,提出一种基于循环神经网络的智能电网拓扑变异型虚假数据注入攻击检测方法.通过分析电力系统状态估计方法的不足和虚假数据注入攻击绕过系统监测与防御的入侵方式,引入循环神经网络分析连续数据序列的时序变化,并在IEEE-30节点系统上进行仿真验证.仿真结果表明,提出的方法能够高效、准确地检测智能电网中产生的虚假数据注入攻击行为,其检测准确率可达99.9%,相比于其他检测方法具有较大的优势.In order to solve the problem of false data injection causing serious security threat to the operation state of smart grid in recent years, a detection method based on recurrent neural network for topology variation false data injection(FDI) attacks in smart grid was proposed. By analyzing the deficiencies of power system state estimation methods and intrusion modes of FDI attacks by passing system monitoring and defensing, recurrent neural network(RNN) was introduced to analyze the time sequence changes of continuous data sequence, and simulation verification was carried out on IEEE-30 node system. The simulation results show that the as-proposed method can effectively and accurately detect the FDI attacks in smart grid with a detection accuracy of 99.9%, and it has greater advantages compared with other detection methods.

关 键 词：虚假数据注入 循环神经网络 智能电网 攻击检测方法 拓扑变异 时序变化 IEEE-30节点系统 潮流数据 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]