Research On False Data Detection And Defense Of Smart Grid Based On PMU/SCADA Measurement

With the development of information physics system and the high integration of information system and physical system,the single physical power grid is gradually replaced by a new generation of typical CPS,such as smart grid,energy Internet and so on,and the dependence of power grid on information system is increasing.It not only brings convenience to life,but also brings hidden dangers to the security problems of power system.There is a risk of network attack in the introduction of information into the power grid.At present,the control center has the ability to detect most of the bad data,but there is no effective detection method for this concealment to avoid the attacks of the current monitoring mechanism.Attackers carefully plan to inject false data into the measurement system to avoid the traditional bad data detection mechanism.This kind of attack is called false data injection attack.In this thesis,the online detection and offline defense model of false data injection attack based on PMU/SCADA measurement are deeply studied.In the context of power grid CPS,it is of great theoretical and practical value to study the integrated online and offline defense methods of malicious data injection attack for the safe and reliable operation of power grid CPS.(1)This thesis introduces the loophole of the existing power system control center and the bad data detection scheme of state estimation,and models the false data attack.Then a method of reasonably configuring PMU offline is proposed to weaken the influence of false data attack on state estimation.The PMU configured should satisfy global observability,and consider the situation of zero injection node,so as to reduce the number of PMU as far as possible to complete the optimization configuration of PMU.(2)Because optimized PMU still have the possibility of attack,and in large-scale power grid close to optimized PMU deployment and protection measurement will greatly increase the additional human and financial disadvantages.Based on the characteristics of malign data injection attack tampering measurement,a detection method is proposed and classified for the false data attack of PMU.The proposed method is based on continuous monitoring of the equivalent impedance of the transmission line and classification of the observed anomalies to detect the presence and location of the attack.With this approach,data integrity can be tested in a distributed and non-iterative manner,thereby reducing memory and processing.(3)Due to the price of PMU itself and other equipment required by PMU installation(such as communication system)are very expensive,so it is impossible to fully configure PMU and completely replace SCADA in the whole network at the present stage.Therefore,PMU and SCADA measurement system will coexist for a long time.The measurement data of PMU and SCADA are firstly fused and processed,and then a kind of false data detection based on improved adaptive Untracked Kalman Filter is proposed.Untracked Kalman Filter state estimation algorithm belongs to dynamic state estimation,which is based on measurement data of multiple time sections,and USES state equation to estimate the state of the system at the current moment and predict the state of the system at the next moment.It has high theoretical and practical research value.This thesis improved the traditional no trace kalman filter,the introduction of fading factor and the adaptive factor,dynamic adjustment process noise and measurement noise covariance matrix,makes the forecast more stable and accurate,and recycle using the least square method and kalman filter for state estimation of residual error is analyzed,so as to achieve the purpose of detecting false data.(4)From the point of view of physical defense against false data injection attack,a multi-stage dynamic defense method is proposed to analyze the complex interaction between attackers and defenders in the power grid.A two-player zero-sum random game is established between the attacker and the defender.In this game,each player seeks to maximize the minimum return under the optimal strategy of the other party,and measures the effectiveness of attack and defense by reducing the amount of load.Using these quantified rewards as input,the algorithm is used to derive Nash equilibrium strategies for attackers and defenders,and the convergence and properties of the algorithm are studied.