摘要
针对虚假数据定位检测适应性低、篡改量测影响系统状态精确感知的问题,提出一种基于WGAN (Wasserstein generative adversarial networks, WGAN)状态重构的智能电网虚假数据检测与修正模型。首先,根据历史状态变量的概率分布,初步锁定并剔除具有潜在攻击风险的状态变量。然后,采用Wasserstein生成对抗网络重构缺失变量,WGAN通过Wasserstein距离衡量生成分布与真实分布之间的差异,能够生成有意义的梯度以优化网络模型参数。最后,以重构状态作为一种状态参考,精确定位攻击节点,并结合网络拓扑参数修正篡改量测值。将纽约州数据用在IEEE-14节点测试系统,进一步验证所提方法的可行性与有效性。
Aiming at the problems of low adaptability of false data location detection and the influence of tamper measurement on the accurate state awareness for the power system, a false data detection and correction model of smart grid based on state reconstruction of Wasserstein Generative Adver-sarial Networks (WGAN) was proposed. Firstly, the state variables with potential attack risk are ini-tially locked and eliminated according to the probability distribution of historical state variables. Then, the missing state variables are reconstructed by Wasserstein generative adversarial net-works. WGAN measures the difference between the generated distribution and the real distribution through Wasserstein distance, which can generate meaningful gradients to optimize the parame-ters of the network model. Finally, the reconstructed variables are used as a state reference to lo-cate the attacked bus, and to correct the measurement data combined with the network topology parameters. The feasibility and validity of the proposed method are further verified in IEEE-14 bus test system with the New York data.
出处
《建模与仿真》
2023年第3期2182-2196,共15页
Modeling and Simulation
