Research On Defense Of Smart Grid Data Integrity Attacks

With the improvement of the intelligence of my country’s power system,the integration of smart grids and communication networks continues to deepen,the power system is facing serious network security threats,and information security has become an important factor affecting the stable operation of the power system.Existing research shows that data integrity attacks can bypass the bad data detection mechanism and seriously interfere with the normal operation of the power grid.However,traditional detection methods cannot detect this type of attack,so in order to ensure the security of the smart grid,it is of great theoretical and practical significance to study the corresponding detection methods.The main contributions of this article are as follows:1.This article builds a data integrity attack detection model based on deep learning,so the power measurement data sample set required for model training and testing is the basis of this article.First of all,on the basis of in-depth research on the principle of data integrity attacks,we successfully constructed a data integrity attack vector that can only know the topological structure of the power grid,and conducted experiments on Mat Power to build IEEE 9,14 and 30 standard bus systems The network topology of DDQN,and obtained a large amount of measurement data including normal measurement data of the power system and the measurement data after the power system is attacked,which provides a large amount of data for the construction of the DDQN detection model.2.Electric power measurement data has the characteristics of high dimensionality and strong noise,and cannot be directly used for the training and detection of detection models.Therefore,this article describes the defense process against data integrity attacks as the Markov Decision Process(MDP).In this process,we proposed expressions for state space,action space,reward function and observation space,which eliminated the influence of noise and improved the accuracy and speed of the detection strategy.3.For smart grids,high-dimensional state-actions will bring dimensionality disasters,so in order to solve the dimensionality disasters,this paper proposes a Double-Deep-Q-Network(DDQN)detection scheme to defend against electricity Data integrity attacks in the system.DDQN is a deep reinforcement learning solution,which avoids the curse of dimensionality that traditional reinforcement learning solutions have.DDQN strategy Apply the main network and target network to learn the best defense strategy.In order to improve the learning efficiency,a quantification method of the observation space is proposed,and a sliding window is also adopted.Experimental evaluation results show that DDQN is superior to the existing detection schemes based on deep reinforcement learning in terms of detection accuracy and speed in IEEE 9,14 and 30 bus systems.