False Data Injection Attack Detection for Smart Grid State Reconstruction Based on Adaptive Generative Adversarial Networks
Considering the deep coupling between power system and energy Internet,the threat of false data injection attack to power system is increasing.This paper designs an adaptive generative adversarial networks(AGAN)state reconstruction method for false data injection attack(FDIA)detection.AGAN state reconstruction method for false data injection attack detection.The method is based on generative adversarial networks(GAN),incorporates convolutional neural network(CNN)and self-attention(SA)mechanism under adaptive constraints to achieve global referentiality among nodes,so as to realize effective reconstruction of states and accurate prediction of abnormal states.The detection logic combining network discriminant values is proposed based on the abnormal data prediction results of AGAN.Finally,the effectiveness of the proposed method is verified on the power system of IEEE14 nodes,and the average absolute percentage error of AGAN reconstruction is 0.0001%compared with GAN and CNN,and the detection accuracy can reach 98%.
generative adversarial networksself-attention mechanismsadaptive constraintspower systemsfalse data injection attacks
万方数据
维普
