摘要
状态估计是现代电力系统运行与控制的关键环节,假数据注入攻击可以绕过基于残差的传统错误数据检测方法,威胁电网正常运行。针对交流系统状态估计,提出了基于DWT(离散小波变换)和DNN(深度神经网络)的电网假数据注入攻击检测方法。利用小波变换提取多时段状态估计结果中的时-频特征,并将时-频特征进一步量化为所有时刻小波变换系数的期望和方差;再将状态估计的时-频特征作为输入,通过经离线训练的DNN在线输出攻击检测结果。算例证明所提方法能够有效提高电网假数据注入攻击检测的准确性。
State estimation is critical to modern power system operation and control.The false data injection attack(FDIA)can bypass the traditional residual-based error data detection and threaten the normal operation of the power grid.For the state estimation of the AC system,this paper proposes a detection method of FDIA based on DWT(discrete wavelet transform)and DNN(deep neural network).Wavelet transform is applied to extract the time-frequency feature of multi-period state estimation results,and the time-frequency features are quantized as the expectation and variance of wavelet transform coefficients;then taking the time-frequency feature of state estimation as the input,the detection results of false data attack are output online through the DNN trained offline.An example shows that the proposed method can effectively improve the accuracy of FDIA detection.
作者
蒋正威
张超
孙伟乐
赵友国
隋向阳
JIANG Zhengwei;ZHANG Chao;SUN Weile;ZHAO Youguo;SUI Xiangyang(State Grid Zhejiang Electric Power Co.,Ltd.,Hangzhou 310007,China;State Grid Hangzhou Power Supply Company,Hangzhou 310009,China;Dongfang Electronics Co.,Ltd.,Yantai Shandong 264000,China)
出处
《浙江电力》
2020年第11期45-50,共6页
Zhejiang Electric Power
基金
国网浙江省电力有限公司科技项目(5211HZ1800P3)。
