FDIA Detection And Recovery Research Based On Edge Computing

False data injection attack(FDIA),as one of the power system cyber attacks,poses a great threat and challenge to the safe and stable operation of smart grid.Attackers construct stealthy attack vectors through power system parameters and topology information to tamper with the measurement data of smart measurement devices and attack the edge computing centers of smart grids,causing grid operation accidents and economic losses.However,current FDIA detection usually requires a priori knowledge of system models and parameters,as well as the need to determine judgment thresholds,which can bring high detection computational complexity and latency.In this paper,we take full advantage of the low latency,close proximity to terminals and intensive data collection of edge computing to study the defense of false data injection attacks in smart grid under edge computing.Data prediction is performed by historical data densely collected from edge devices,FDIA detection is performed based on a prediction +classification model,and data recovery is performed by generating an adversarial network to achieve an integrated FDIA defense scheme under edge computing with detection and recovery.The main research elements of this paper are as follows.(1)Three deep learning algorithms,recurrent neural network,long and short-term memory network,and gated neural unit,are used to implement prediction of power system measurement data,and the prediction performance of the three different built forms of neural networks is compared,and the optimal data prediction algorithm and network building form are selected.(2)A prediction+classification based FDIA detection scheme is proposed,a classification method based on the combination of Kalman filter and convolutional neural network is designed,and the detection performance of the classification network before and after the introduction of Kalman filter is compared under several performance indexes.Simulation experiments are completed under IEEE 14-node system in combination with open-source datasets to demonstrate the feasibility of the proposed detection scheme.Two existing detection algorithms are compared and the performance of the proposed detection scheme is demonstrated with numerical results under the same dataset and performance metrics.(3)A data recovery scheme based on generative adversarial networks is proposed,and the measurement data generation network model under the power system is elaborated based on the study of the picture generative adversarial network DCGAN.Combined with the previous discriminative network,the generative adversarial network for data recovery under power system is built.The data recovery performance of the proposed scheme is verified under several performance metrics by comparing the direct use of predicted data as real data.The feasibility of using the restored measurement data from the generative network for system state estimation is verified.(4)An experimental scenario of microgrid under edge computing is built,and the real-time data acquisition at the terminal,FDIA detection algorithm implementation,and other functions are completed under the edge computing device to realize the development of a real-time online FDIA detection system for microgrid and verify the feasibility of the proposed detection scheme and recovery scheme.