基于数据驱动的电力系统虚假数据注入攻击检测

针对虚假数据注入攻击（FDIA）导致的电力系统安全问题，提出了一种数据驱动的二阶段联合方法。首先，构建由贝叶斯优化（BO）改进的极端梯度提升算法BO-XGBoost作为基分类器对正常数据和虚假数据进行初筛。其次，考虑到分类问题有一定误报率，构建最大互信息系数（MIC）校验器，计算由基分类器判断为正常状态下系统节点间的MIC，基于该系数的值进一步校验潜藏在系统里的FDIA。最后，通过IEEE 39节点测试系统的仿真分析，验证了所提方法的有效性。

False Data Injection Attacks Detection in Power System Based on Data-driven Algorithm

Targeting the problem of power system security due to false data injection attack （FDIA）, the paper proposes a data-driven two-stage joint method. Firstly， an extreme gradient boosting （XGBoost） algorithm modified by Bayesian optimization （BO） is established as a base classifier to perform the initial screening of normal data and false data. Then, by considering a certain false alarm rate in classification， a maximum mutual information coefficient （MIC） is used to construct a verifier to calculate the MIC between the system nodes under the normal state determined by the base classifier， and the FDIA hidden in the system is further verified based on the value of this coefficient. Finally， the validity of the method is verified through the simulation of an IEEE 39-node test system.