基于改进多隐层极限学习机的电网虚假数据注入攻击检测  

Research on False Data Injection Attack Detection in Power System Based on Improved Multi Layer Extreme Learning Machine

作　　者：席磊 何苗[1] 周博奇 李彦营 XI Lei;HE Miao;ZHOU Bo-Qi;LI Yan-Ying(College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443002;Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station,China Three Gorges University,Yichang 443002)

机构地区：[1]三峡大学电气与新能源学院,宜昌443002 [2]梯级水电站运行与控制湖北省重点实验室(三峡大学),宜昌443002

出　　处：《自动化学报》2023年第4期881-890,共10页Acta Automatica Sinica

基　　金：国家自然科学基金(51707102);信息物理融合防御与控制系统宜昌市重点实验室(三峡大学)开放基金(2020XXRH04)资助。

摘　　要：虚假数据注入攻击(False data injection attacks,FDIA)严重威胁了电力信息物理系统(Cyber-physical system,CPS)的状态估计,而目前大多数检测方法侧重于攻击存在性检测,无法获取准确的受攻击位置.故本文提出了一种基于灰狼优化(Gray wolf optimization,GWO)多隐层极限学习机(Multi layer extreme learning machine,ML-ELM)的电力信息物理系统虚假数据注入攻击检测方法.所提方法将攻击检测看作是一个多标签二分类问题,不仅将用于特征提取与分类训练的极限学习机由单隐层变为多隐层,以解决极限学习机特征表达能力有限的问题,且融入了具有强全局搜索能力的灰狼优化算法以提高多隐层极限学习机分类精度和泛化性能.进而自动识别系统各个节点状态量的异常,获取受攻击的精确位置.通过在不同场景下对IEEE-14和57节点测试系统上进行大量实验,验证了所提方法的有效性,且分别与极限学习机、未融入灰狼优化的多隐层极限学习机以及支持向量机(Support vector machine,SVM)相比,所提方法具有更精确的定位检测性能.False data injection attacks(FDIA)seriously threaten the state estimation of power cyber-physical system(CPS).At present,most detection methods focus on the detection of attack existence,and can not obtain the accurate attacked location.Therefore,this paper proposes a false data injection attack detection method for power cyber-physical system based on gray wolf optimized multi layer extreme learning machine(ML-ELM).The proposed method regards attack detection as a multilabel binary classification problem.It not only changes the extreme learning machine used for feature extraction and classification training from single hidden layer to multi hidden layer to solve the problem of limited feature expression ability of extreme learning machine,but also integrates the gray wolf optimization(GWO)algorithm with strong global search ability to improve the classification accuracy and generalization performance of multi layer extreme learning machine.Furthermore,it automatically identify the abnormal state variables of each node of the system,and obtain the accurate attacked location.Through a large number of experiments on IEEE-14 and 57 node test systems in different scenarios,the effectiveness of the proposed method is verified,and the proposed method has more accurate positioning and detection performance than the extreme learning machine,the multi layer extreme learning machine without gray wolf optimization and the support vector machine(SVM),respectively.

关 键 词：电力信息物理系统 虚假数据注入攻击 状态估计 灰狼优化 多隐层极限学习机 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TM73[自动化与计算机技术—控制科学与工程] TP393.08[电气工程—电力系统及自动化]