Research On Security Detection Strategy Of Smart Grid Under False Data Injection Attac

With the rapid development of the information society,the traditional power system has completed the deep interconnection with the control,monitoring,communication and other information systems,forming a smart grid with distinctive characteristics.Compared with the traditional power system,smart grid shows higher openness.While bringing great convenience to the society,it will also face more challenges from new network attacks.False Data Injection Attack is a new type of network attack on data integrity of power system in recent years.One of the most distinctive characteristics of False Data Injection attack is concealment,that is,it can tamper with the estimated results of the system while avoiding the bad data detection module to achieve covert attack.At present,the research on such attacks is in the ascension.Therefore,in-depth research on the attack mechanism of False Data Injection and countermeasures is of positive significance for the stable operation of smart grid.This paper carries out the following work on the detection strategy of False Data Injection Attacks:(1)The current widely used state estimation method is studied in detail,and the security vulnerability of the bad data detection module based on residual test in the weighted least square method is analyzed.The theoretical basis of False Data Injection Attack with complete topology information and partial topology information is studied in detail.Based on the view of attackers grasping part of the topology information,a false data modeling method with minimal attack vector is proposed,and simulation experiments are carried out to verify the damage effect of this method on state estimation.(2)Aiming at the false data modeled by minimizing attack vector,a fake data injection attack detection strategy based on untraced Kalman filter is proposed.By comparing the deviation with the weighted least square method and combining with the consistency test results,the method can determine whether the system suffers from false data injection attacks.Compared with the state estimation results of the weighted least square method and the extended Kalman filter,the average absolute error and the highest accuracy of the state estimation results of the Unscented Kalman filter before and after the attack are the smallest.Moreover,in the detection strategy studied in this paper,the more accurate the state estimation result is,the more stable it will be after being attacked,and the more advantageous it will be when detecting false data with stronger concealment.(3)The Sage-Husa noise estimator with time-varying properties and the fading memory index weighting method with forgetting factor are introduced on the basis of Unscented Kalman Filter.A false data injection attack detection strategy based on the improved adaptive Unscented Kalman Filter is proposed.Simulation results show that this method can effectively detect false data in the system,and the state estimation results are more accurate,the average absolute error before and after the attack is smaller,and the filtering performance is improved.Then it is verified that this method can detect more subtle fake data in the system by setting different attack intensity scenarios.Finally,by comparing with the support vector machine method,the effectiveness and accuracy of the proposed method in detecting covert attack vectors are verified.