Research On Detection And Localization Of False Data Injection Attacks In Smart Grids

With the progress of information and automation technology,the physical system and information system of smart grids are deeply integrated,appearing intelligent cyber-physical systems(CPSs).The integrated power CPS has efficient economic dispatching,power trading and decision control,but it is easy to have information security problems.False data injection attack(FDIA)is a common network attack with strong destructive and concealment.It damages the security and stability of smart grid or illegally obtains benefits by tampering with the results of state estimation in smart grid.At present,the research on FDIA is in its infancy,and most of which only focuses on detecting the existence of FDIA,without realizing the localization.However,the localization of FDIA is as important as detection,because localization can find out the attacked area in power systems,and measures such as isolating buses can be taken to prevent the same area from being attacked again.Therefore,it is of great theoretical and practical significance to study the detection and localization of FDIA in smart grids.Aiming at the detection and localization of FDIA,this paper carries out the following work:1.The power system model is constructed.This paper expounds the basic principle and construction mode of FDIA in smart grids,and analyzes the principle of confidentiality.Considering that attackers are not able to obtain all the information of power grid in practice,the false data attack vectors are constructed based on phase Angle difference,which only needs partial information of the attack areas.The effectiveness of proposed method is verified in IEEE-14 standard power system.2.A detection method based on improved strong tracking unscented Kalman filter(ISTUKF)for FDIA is proposed.The improved strong tracking algorithm is used to correct the measurement noise error covariance matrix in untracked Kalman filter in real time,so as to reduce the influence of non-Gaussian measurement noise on dynamic state estimation and FDIA detection.Euclidean distance of deviations between dynamic state estimate result of ISTUKF and static estimate result of weighted least square method are calculated,which are compared with the improved thresholds to detect FDIA.Simulation results in IEEE-14 standard power system show that the proposed method can effectively detect FDIA and has advantages over other detection methods.3.Considering the complexity of FDIA localization method based on multiple filters which has great limitations,a localization method based on convolutional neural network(CNN)optimized by sparrow search algorithm(SSA)for FDIA is proposed.A CNN model for FDIA localization is designed,and SSA is used to optimize the parameters of CNN,which will avoid the difficulty of selecting parameters manually,thus improving the accuracy of the localization method.A joint strategy based on state estimation and SSA-CNN for detection and localization of FDIA is proposed.The detection method based on ISTUKF is used to remove abnormal measurement data and detect FDIA in whole system,and then the localization method based on SSA-CNN can accurately locate the area being attacked.Simulation results on IEEE-14 and IEEE-118 standard power systems show that the proposed method can effectively locate FDIA and has advantages over other localization methods.