抵御多点虚假数据攻击的主动配电网状态估计方法

引入同步相量量测装置（PMU）识别虚假数据注入攻击（FDIA）的动态状态估计是实现主动配电网安全准确决策的有效途径。提出一种融合贝叶斯定理的深度森林（BA-DF）FDIA检测机制的混合量测加权平方根容积卡尔曼滤波（WASRCKF）动态状态估计方法。首先，通过图卷积神经网络预测PMU和SCADA混合量测融合，提高数据冗余度；其次，利用WASRCKF估计状态量和混合量测预测量进行加权估计，降低FDIA对状态估计更新层的影响；然后，采用BA-DF进行FDIA检测，判断虚假数据攻击位置，使用混合量测预测值进行修正，形成BA-DF-WASRCKF组合方法。最后，采用PG&E69配电网进行验证，结果表明该方法在不同PMU配置下均可获得更高精度状态估计结果，配置24台PMU的FDIA识别率为95%，较传统方法状态估计精度提高了77.8%。

Active Distribution Network State Estimation Method Against Multi-Point False Data Injection Attacks

The dynamic state estimation with synchronous phasor measurement unit （PMU） identifying false data injection attacks （FDIA） is an effective way to make safe and accurate decisions for active distribution network. A dynamic state estimation method of Weighting Adaptive Square-root Cubature Kalman Filter （WASRCKF） with Bayesian algorithm and deep forest （BA-DF） FDIA detection is proposed. Firstly，the graph convolutional network is used to predict the fusion of PMU and SCADA to improve the data redundancy. Secondly， WASRCKF estimation and mixed measurement prediction are used to reduce the influence of FDIA on the update layer. Then，BA-DF detection finds the location of FDIA， corrected by prediction， forming BA-DF-WASRCKF. Finally，experimental results show the proposed method can improve the accuracy of distribution network state estimation with PG&E69 distribution network under different measurement configurations. The FDIA recognition rate with 24 PMUs is 95%， and estimation accuracy of BA-DF-WASRCKF is 77.8% higher than that of the traditional method.