Research On Detection Method Of Power Cyber-physical System Considering False Data Attacks

The power system provides electricity for all walks of life in society,and its safe and stable operation is related to people's livelihood and social stability.With the construction of the ubiquitous electric power Internet of Things,China's electric power is further enhanced.A large number of various devices are interconnected through the communication network and electric power network to form a complex multi-dimensional heterogeneous system,a cyber-physical system(CPS).Due to the randomness of sensor nodes in the power CPS sensing layer and the openness of data interaction communication channels,the system is vulnerable to network attacks,such as false data injection attacks(FDIAs).It is a typical new attack method for data integrity of state estimation in power CPS.FDIAs can effectively avoid traditional bad data detection,tamper with the state value of the power grid,and affect the safe and stable operation of the power system.Therefore,researching efficient and feasible FDIAs detection methods is of great significance for building a safe and stable running power cyber-physical system.This paper conducts in-depth research on detection methods of power cyber-physical system considering false data injection attacks.The specific content is as follows:First of all,the structure of the power cyber-physical system is studied.It is a typical CPS for power systems to connect power plants and users through the transmission and distribution network.The information interaction between power cyber physical systems is becoming more and more complex while facing more network security issues.There are two types of cyber attacks on the power cyber physical system,which are divided into two categories according to network coverage and attack purpose.At the same time,the weighted least squares method and residual data detection based on residuals are often used in power system state estimation.Secondly,a false data attack detection method based on power system dynamic estimation is proposed.The basic principle of the false data injection attack is studied.The false data is constructed by constructing a false data attack vector model to find the smallest false data attack vector to launch an attack on the system.At the same time,it is proposed to use the detection methods based on extended Kalman filtering and based on unscented Kalman filtering to detect false data attacks.By performing a consistency check on the state estimation results,false data can be successfully detected,and the validity of the proposed method is verified.Feasibility of testing methods.Finally,An adaptive unscented Kalman filter based on a variable noise statistical estimator is proposed to detect false data attacks.Improve the unscented Kalman filter,adjust the weighting value according to the evanescent memory index,and introduce a timevarying noise estimator to make the unscented Kalman filter more adaptive to system changes and reduce estimation errors.Simulation results show that the improved unscented Kalman filter can effectively identify false data injection attacks,and the deviation of state estimation results based on extended Kalman filtering and unscented Kalman filter detection is smaller,and the filtering performance is greatly improved.