基于递归神经网络的智能电网虚假数据检测  

False Data Detection of Smart Grid Based on Recurrent Neural Network

作　　者：邱荣福 谢型浪 谢虎 杨清浩 QIU Rong-fu;XIE Xing-lang;XIE Hu;YANG Qing-hao(Digital Grid Research Institute,China Southern Power Grid.,Guangzhou 510770 China;Guangzhou Zhixun Information Technology Co.,Guangzhou 510635 China)

机构地区：[1]南方电网数字电网研究院有限公司,广东广州510770 [2]广州致讯信息科技有限责任公司,广东广州510635

出　　处：《自动化技术与应用》2021年第1期105-109,共5页Techniques of Automation and Applications

摘　　要：虚假数据注入(FDI)攻击为智能电网的稳定运行带来了挑战,特别是虚假数据经过人为精心构造后,能够绕过电力系统中的常规错误数据检测技术,为电网安全运行带来了安全隐患。为此本文应用递归神经网络(RNN)来检测智能电网中的虚假数据攻击(FDI攻击)。通过编写检测算法,基于IEEE-30总线系统以及从实际电网中获得的为期五年的潮流数据,对FDI进行了仿真模拟。仿真结果表明,本文提出的基于递归神经网络的算法通过观察连续数据序列的时间变化,有效检测出了混入正常潮流数据中的虚假攻击数据。The attack of false data injection(FDI)brings challenges to the stable operation of smart grid,especially after the artificial and careful construction of false data,it can bypass the conventional error data detection technology in the power system,which brings security risks for the safe operation of power grid.In this paper,recurrent neural network(RNN)is used to detect the false data attack(FDI attack)in smart grid.Based on the IEEE-30 bus system and the five-year power flow data obtained from the actual power grid,the simulation of FDI is carried out.The simulation results show that the algorithm based on recurrent neural network can effectively detect the false attack data mixed with normal power flow data by observing the time change of continuous data sequence.

关 键 词：智能电网 网络安全 错误数据注入 递归神经网络 状态估计 不良数据检测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]