基于最大信息系数-双层置信极端梯度提升树的电网虚假数据注入攻击定位检测  

Grid False Data Injection Attack Localization Detection Based on MIC-double-deck Confidence XGBoost Tree

作　　者：席磊 王文卓[1] 白芳岩 陈洪军 彭典名 李宗泽 XI Lei;WANG Wenzhuo;BAI Fangyan;CHEN Hongjun;PENG Dianming;LI Zongze(College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443002,Hubei Province,China;Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station(China Three Gorges University),Yichang 443002,Hubei Province,China)

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

出　　处：《电网技术》2025年第2期824-833,I0112-I0114,共13页Power System Technology

基　　金：国家自然科学基金项目(52277108)。

摘　　要：面向高维复杂的电力量测数据,现有攻击定位检测方法存在定位精度差的问题。为此该文提出一种基于最大信息系数-双层置信极端梯度提升树的电网虚假数据注入攻击定位检测方法。所提方法引入最大信息系数对量测数据进行特征选择,能够非线性地衡量数据特征之间的关联性,且公平地根据一个特征变量中包含另一个特征变量的信息量来去除冗余特征,有效解决虚假数据注入攻击定位检测方法普遍面临的量测数据高维冗余问题;同时提出一种具有正反馈信息传递作用的双层置信极端梯度提升树来对各节点状态进行分类,通过结合电网拓扑关系学习标签相关性,从而有选择性地利用前序标签有效预测信息,来减少后续分类器学习到的前序标签预测信息中包含的错误,最终实现对受攻击位置的精确定位。在IEEE-14、IEEE-57节点系统上进行大量仿真,算例结果验证了所提方法的有效性,且相较于其他方法具有更高的准确率、精度、召回率、F1值和AUC(area under curve)值。As far as the high-dimensional and complex electrical force measurement data has been concerned,the existing attack-location detection methods unveil the problem of poor location accuracy.Therefore,this paper proposes a new method of location and detection by False Data Injection Attack in power grids on the foundation of Maximum Information Coefficient-Double-Deck Confidence Extreme Gradient Boosting Tree.The proposed method introduces the Maximum Information Coefficient for feature selections of measured data,which benefits the correlation measurement between data features nonlinearly,and the efficient movement of redundant features fairly according to the information of one feature variable containing another feature variable,further for an effective solution for the problem of high-dimensional redundancy of measurement data commonly faced by the method of location and detection with False Data Injection Attack.At the same time,the current research puts forward a Double-Deck Confidence Extreme Gradient Boosting Tree with the effect of positive feedback information transmission to classify the different states of nodes.On the other,the label correlations are learned by combining the power grid topological relationship.Therefore,the preorder label is selectively used to effectively predict information,reducing the errors contained in the prediction information of the preorder label learned by subsequent classifiers with the eventual objective that a precise location of the attacked positions are satisfactorily realized.More importantly,many simulations on IEEE-14 and IEEE-57 node systems validate the effectiveness of the aforementioned method,which exhibits much higher accuracy,precision,recall rate,F1 Score,and area under curve(AUC)score when compared with other methods.

关 键 词：虚假数据注入攻击 最大信息系数 双层置信 极端梯度提升树 标签相关性 

分 类 号：TM721[电气工程—电力系统及自动化]