摘要
针对当前基于循环神经网络的智能电网虚假数据注入攻击(False Data Injection Attacks,FDIA)检测方法无法提取FDIA数据深层特征的问题,提出一种基于双向门控循环单元(Bidirectional Gated Recurrent Unit,Bi-GRU)和自注意力的检测方法。采用Bi-GRU学习量测序列,输出各时间步的隐状态;引入自注意力机制计算各时间步隐状态的线性加权和作为量测序列的深层特征;通过全连接神经网络层和Softmax层输出预测概率。在IEEE 30-bus和IEEE 14-bus中的仿真实验结果表明该方法切实可行,相较于次优结果,准确率平均提高7.1%,正报率平均提高3.95%,误报率平均降低38.85%。
Aiming at the problem that the current detection methods of false data injection attacks(FDIA)in smart grid based on recurrent neural network cannot extract the deep features of FDIA data,a detection method based on bidirectional gated recurrent unit(Bi-GRU)and self-attention is proposed.The Bi-GRU model was used to learn the sequence of measurement and output the hidden state of each time step;the self-attention mechanism was introduced to calculate the linear weighted sum of the hidden states of each time step as a deep feature of the sequence;the predicted probabilities of the deep features were output by fully connecting layer and the Softmax layer.The simulation results on IEEE 30-bus and IEEE 14-bus show that the proposed method is feasible.Compared with the sub-optimal results,the accuracy rate is increased by 7.1%on average,the positive rate is increased by 3.95%on average,and the false alarm rate is reduced by 38.85%on average.
作者
陈冰
唐永旺
Chen Bing;Tang Yongwang(Luohe Institute of Technology,Henan University of Technology,Luohe 462000,Henan,China;Institute of Information and System Engineering,PLA Information Engineering University,Zhengzhou 450001,Henan,China)
出处
《计算机应用与软件》
北大核心
2021年第7期339-344,349,共7页
Computer Applications and Software
关键词
双向门控循环单元
自注意力
智能电网
虚假数据注入攻击
隐状态
Bidirectional gated recurrent unit
Self-attention
Smart grid
False data injection attacks
Hidden state
作者简介
陈冰,讲师,主研领域:人工智能,控制工程;唐永旺,讲师。
