Research On Data Integrity Attack Detection Based On Node Voltage In Energy Internet

Data integrity attack in energy Internet is a new attack method derived from information physical fosion system vulnerabilities,the most typical of which is False Data Injection Attacks(FDIAs).FDIAs tamper with the measurements of power grid sensors in many ways to change the state estimation results and bypass the existing bad data identification mechanism.The voltage control system will issue error control instructions according to the collected false data to make the actual voltage offset exceed the limit and cause power system instability.Research on efficient detection methods is of great significance for building a safe and stable energy internet.Aiming at data integrity attacks in energy internet,this thesis focuses on new attack vector construction methods and corresponding detection methods for node voltage.By analyzing the principle of FDIAs and the research status at home and abroad,the existing attack methods and detection methods are compared and analyzed.In terms of attack methods,most of the existing studies attack global state variables,and few of them attack data integrity of node voltage.In the aspect of attack detection using deep learning or machine learning algorithm,the existing research mostly uses normal samples and attacked samples for model training,which can not overcome the shortcomings of small scale of attacked samples in actual power grid.In order to solve the above problems,this thesis proposes injection attack strategies for node voltage amplitude data and detection methods for time series reconstruction using cyclic neural networks from two perspectives of attack and detection.In the research of data integrity attack method based on node voltage,data injection is carried out for the voltage amplitude of the specified node,and the problem of superimposing false data in the real value of measurement is approximately transformed into a weighted linear programming.The fast regression algorithm is used to maximize the contribution of model candidate to model error reduction.The solution of this programming problem is aimed at node power.The optimal strategy for data integrity attacks.In the detection method of time series reconstruction based on cyclic neural network,the time series consisting of voltage amplitude and phase angle in power system is reconstructed,and the probability that the original time series belongs to the attacked sequence is estimated by the error of reconstructed output data and original input data.When the error exceeds a certain threshold,the data integrity attack exists in the system.Aiming at the proposed attack methods and detection methods,this thesis uses Matpower power power power simulation package to carry out simulation experiments in the standard test system of IEEE 14-bus and 39-bus,and verifies the feasibility and effectiveness of the method.The experimental results show that the proposed data integrity attack method based on node voltage has fewer measurement tampering and higher attack success rate than the traditional attack method.Compared with traditional machine learning classification algorithm,attack detection algorithm based on cyclic neural network has higher accuracy and can effectively detect data integrity attacks.