基于LSTM的核电站除氧器水位控制系统隐蔽攻击方法研究  

Research on Covert Attack Method of Deaerator Water Level Control System in Nuclear Power Plants Based on LSTM

作　　者：王东风[1] 邓鉴湧 黄宇[1] 高鹏 WANG Dongfeng;DENG Jianyong;HUANG Yu;GAO Peng(Department of Automation,North China Electric Power University,Baoding 071003,Hebei Province,China)

机构地区：[1]华北电力大学自动化系,河北保定071003

出　　处：《动力工程学报》2023年第5期590-597,605,共9页Journal of Chinese Society of Power Engineering

基　　金：中央高校基本科研业务费专项资金资助项目(2021MS089)。

摘　　要：针对实现核电站除氧器水位控制系统的隐蔽攻击需要受攻击对象具备高精度估计模型的问题,提出了一种基于长短期记忆(LSTM)神经网络的隐蔽攻击方法。该方法将核电站除氧器水位控制系统中获得的反馈控制器输出信号与输入信号作为LSTM的数据集,通过训练得到高精度的估计模型,此估计模型被用于设计隐蔽攻击器,向受攻击对象施加攻击信号,从而实现隐蔽攻击。最后,通过仿真实验验证了该方法的有效性和隐蔽性。结果表明:该方法对核电站除氧器的水位造成一定影响的同时具有较高的隐蔽性。Aiming at the problem that realizing covert attack of deaerator water level control system in nuclear power plants would have a high-precision estimation model of the attacked objects,a covert attack method based on long-short term memory(LSTM)neural network was proposed.In this method,output and input signals of feedback controllers obtained from deaerator water level control system in nuclear power plants were utilized for the data set of LSTM,and a high-precision estimation model was obtained through training.This estimation model was used to design a covert attacker to impose attack signals on an attacked object,so as to realize covert attack.Finally,the effectiveness and concealment of the proposed method were verified by simulation experiments.Results show that the method has a certain influence on deaerator water level in nuclear power plants and higher concealment.

关 键 词：核电站 除氧器水位控制系统 隐蔽攻击 长短期记忆神经网络 模型辨识 

分 类 号：TM623[电气工程—电力系统及自动化]