Research On Cyber-physical Attack Detection And Isolation In Smart Grid Via State Observer

In recent years,the traditional power system is gradually expanding from a single transmission and distribution system to a modern large-scale smart grid system that exchanges,shares,and interacts with energy flows and information flows.However,the new features of smart grid systems,such as complexity,intelligence,and openness,make them vulnerable to various malicious cyber physical attacks.Compared with traditional cyber attacks,the characteristics of cyber physical attacks are reflected that an attacker can inject false data to cause the detection residual to remain unchanged,thereby avoiding detection methods from the signal perspective.Based on this deceptive characteristic,cyber-physical attacks pose significant challenges to current signal processing-based detection methods,such as the accuracy of physical dynamic estimation,the design of detection thresholds,and undetectable attacks under structural vulnerability.For this reason,esigning the attack detection mechanism for different cyber-physical attacks is critical to the safe operation of the smart grid for different cyber-physical attacks.Based on this,combined with the theory of cybernetics,this paper conducts research on the detection and separation of cyber physical attacks based on state observers.The main research work is as follows:(1)Aiming at the problem of bias load attack on the motor side in a centralized smart grid,a robust adaptive observer-based attack detection and isoaltion algorithm is proposed.Firstly,a topology-based division algorithm is proposed to reduce the complexity of attack detection and isolation;Then,considering the effect of attack on the angular frequency of the generator,a physical dynamic power grid model is established,and then a robust adaptive observer-based attack detection algorithm is designed;Adaptive thresholds are computed to replace traditional precomputed thresholds;Based on the combination of sensors,a logical judgment matrix is further designed to solve the problem of separation of attacks in the attack area.Finally,simulation research proves the effectiveness of the proposed attack detection and isolation algorithm.(2)Aiming at the problem of bias load attack on the motor side in a distributed smart grid,a distributed unknown input observer-based attack detection and isoaltion algorithm is proposed.Based on the proposed subarea division algorithm,a distributed power grid model is established;The distributed unknown input observer is further designed to deal with model uncertainties and the impact of interconnected relative physical dynamic estimation;Considering model uncertainty and external interference,the computation of distributed adaptive threshold is further given.Considering the influence of interconnection information between neighborhoods on the isolation of attacks in the area,a isolation algorithm based on attack feature matrix is proposed.Finally,the effectiveness of the above detection and isolation algorithm is verified by simulation.(3)Aiming at the problem of stealthy false data injection attacks on the motor load bus in a centralized smart grid,a linear singular interval observer-based attack detection and isoaltion algorithm is proposed.Considering the upper and lower bounds of disturbances in the grid model,the process of interval observer design is given;Based on the relationship between zero and interval residuals,an interval observers-based attack detection algorithm is designed;The proposed detection algorithm can solve the limitation of the residual evaluation function and threshold design in current detection methods;Considering the impact of the transmission of stealthy false data injection attacks on the sensors in the attack area,an attack isolation matrix-based isoaltion algorithm is proposed.Finally,the effectiveness of the above detection and isolation algorithm is verified by simulation.(4)Aiming at the problem of stealthy false data injection attack on the motor side in a centralized smart grid under structural vulnerability,a nonlinear unknown input interval observer-based attack detection and isolation algorithm is proposed.In order to improve the accuracy of interval state estimation,an unknown input interval observer is designed;By designing the parameters of the unknown input interval observer,the influence of unknown disturbances on the interval state estimation can be eliminated to the greatest extent;Furthermore,based on the combination of observable sensors in the attack area,a logical isolation matrix-based isolation algorithm is proposed to detect and isolate the stealthy false data injection attacks under structural vulnerability.Finally,simulation research proves the effectiveness of the proposed attack detection and isolation algorithm.(5)Aiming at the problem of stealthy false data injection attacks on the motor side in a distributed smart grid,a distributed linear interval observer-based attack detection and isolation algorithm is proposed.Considering the characteristics of the stealthy false data injection attacks in distributed smart grids,a distributed detection algorithm based on non-linear interval observers is proposed;Taking the structural characteristics of distributed smart grids into account,the local and global-based attacks isolation matrices are constructed;Considers all possible effects of neighborhood attacks on sensor attacks in the entire area,the local and global attack signature isolation matrices are established;Then,an attack signature isolation matrix-based detection and isolation algorithm for false data injection attacks in distributed smart grids is given.Finally,the effectiveness of the above detection and isolation algorithm is verified by simulation.