Research On Detection Method Of False Data Injection Attack In Smart Grid

With the development of information and communication technology,the traditional power system is also developing towards the integration of information and things,gradually evolving into today’s smart grid.However,smart grids are very vulnerable to network attacks,such as denial of service attacks and false data injection attacks.Among the many types of network attacks,false data injection attacks(FDIAs)are the most threatening attack method to the state estimation of the power grid.In false data injection attacks,cyber attackers obtain and tamper with measurements of the power grid by attacking measurement devices or communication equipment in the power grid.The resulting erroneous system status will affect the operation of the power grid and cause physical or economic loss.Unlike other types of network attacks,successful FDIAs can bypass the traditional bad data detection mechanism based on residuals.The current state estimation of the power system generally uses a DC state estimation model for state estimation,and the largest normalized residual detection(LNR)is performed during the estimation process to eliminate bad data to ensure the accuracy of the state estimation.However,when an attacker constructs a specific attack vector,it will cause a specific deviation in the state estimation value,and the bad data detection module in the state estimation may lose its function.Therefore,in order to ensure the stable operation of the power system,it is necessary to establish an efficient false data detection method.An in-depth study has done on false data injection attacks,the main research contents include:Firstly,starting from the principle of power system state estimation,it introduces the bad data detection function based on residual detection in the state estimation process,analyzes its shortcomings,and leads to the attack principle and construction method of false data injection attack.Then,it simulates the determination of the attack area and the construction of the attack vector on the IEEE-14 node system when the attacker grasps part of the power grid information,and expounds the possible impact of the false data injection attack.Secondly,by comparing the state estimation effects of the extended Kalman filter,the unscented Kalman filter and the adaptive unscented Kalman filter,the adaptive Kalman filter with the best estimation effect is selected as the state estimator,and a Kalman filter-based method is proposed to detect FDIAs.Perform adaptive Kalman filtering to estimate the state value of the system node at each moment,and calculate the root mean square error between it and the estimated value of the system weighted least squares state estimation at that moment.The cumulative distribution function of the root mean square error of all historical moments is used to determine the detection threshold of the proposed method,and the simulation experiment is performed on the IEEE-14 node system.Finally,considering the characteristics of the front and back time correlation of the state values of the grid nodes and the spatial correlation between the nodes,the LSTM neural network suitable for training time series data is selected to predict the state value of the nodes,and the FDIAs detection method based on the LSTM neural network is proposed.Calculate the root mean square error between the predicted value of each historical moment and the state estimate of the system WLS method at that moment,and then use the cumulative distribution function of the root mean square error to determine the detection threshold of the proposed scheme,and the simulation experiment is performed on the IEEE-14 node system.