Research On Detection And Identification Of False Data Injection Attacks In Power Cyber Physical System

With the continuous update of information technology,smart meters and other equipment,the coupling degree between the grid side and the information side of the power system is deepening,forming the Cyber physical system(CPS)of the power grid.The emergence of power CPS further improves the observability and controllability of the power system,but also introduces the network attack threats encountered in the traditional Internet industry into the power system,among which False data injection attack(FDIA)is the most typical and threatening.Therefore,this paper studies and analyzes FDIA in the grid CPS,and the specific contents are as follows:(1)Establishment of power CPS state estimation and construction of FDIA.Firstly,weighted least square state estimation(WLS-SE)is introduced.During the process of Belief propagation based state estimation(BP-SE),the transformation of the power system maps and factor maps is completed and the correlation convergence is proved.The accuracy of the two state estimates is verified through simulation.According to different attack characteristics,two attack models of specific FDIA and random FDIA are constructed and simulated.(2)FDIA was detected based on the optimal scheduling strategy of measuring equipment.Firstly,considering the construction principle and attack characteristics of FDIA,the attack logic that the attacker tampered with measurement information and avoided residual detection,the hat matrix was introduced to determine the optimal scheduling matrix,and the optimal scheduling matrix of measuring equipment was realized.Then the optimal scheduling matrix is used to improve the detection parameters,destroy the attacker’s attack logic,and realize the accurate detection of FDIA.Finally,the superiority of the detection model in detection efficiency,detection time and training cost is verified by simulation and comparative analysis.(3)FDIA was identified based on the improved BP-SE and WLS-SE parallel calculation and pseudo-measurement model.Firstly,the random damping coefficient is used to improve the BP-SE algorithm and increase the pseudo-measurement model,which reduces the requirement of the observability of the system and improves the accuracy of state estimation.Then,based on the different characteristics of WLS-SE and BP-SE at the time of FDIA,Euclidean distance and cosine similarity indexes were introduced for FDIA identification.Finally,simulation verifies that the identification method can identify FDIA through single node variable index,and determines the identification limit of the identification method,and can also detect FDIA through global variable index to a certain extent.