Research On Defense Methods Of Data Integrity Attack In Smart Grid

With the further development of the smart grid and the concept of the energy Internet,the combination of power systems and communication networks is becoming more and more tight,and the types of network attacks that may be encountered are increasing.The data integrity attack is developed from the initial false data attack,which can tamper with the measurement information collected by the data terminal,destroy the integrity of the measurement data,and have a significant impact on the safe operation of the smart grid.In recent years,many new data integrity attacks have emerged.The new types of attacks require fewer resources,are more concealed,and are more difficult to guard against.Therefore,it is very important to study the detection and defense methods of data integrity attacks to ensure the safe and stable operation of smart grids.The thesis considers the data integrity attack of nonlinear dynamic state estimation,and uses the deep learning method of hybrid neural network based on convolutional neural network combined with gated recurrent unit(CNN-GRU)to detect data integrity attacks and improve the generation of anti-network.The method is used to defend against data integrity attacks,and a smart grid data integrity attack defense model is designed.The specific work is as follows:1.The research status of data integrity attack structure in smart grid,the research status of data integrity attack detection and defense are summarized,and based on the principle of power system state estimation and traditional data integrity attack,the nonlinear state estimation is proposed.And two data integrity attack construction methods based on dynamic state estimation.2.Aiming at the problem of the sharp increase of data volume of power system,the traditional shallow machine learning algorithm has a significant decline in detection performance.This thesis proposes a data integrity attack detection model based on CNN-GRU hybrid neural network,considering the space of data and Time characteristics feature extraction to improve the accuracy of smart grid data integrity attack detection.3.A defense method for data integrity attacks is proposed.The WGAN method is used to defend against data integrity attacks,and the problem of unstable network training is solved.It is no longer necessary to carefully balance the training level of generators and discriminators.Applying the proposed method,a smart grid data integrity attack defense model is proposed.A large number of experiments were carried out in the IEEE-14 and 118 standard bus test systems.The deviations between the complementary data and the normal data after different iterations were compared and analyzed,and the performance and effectiveness of the defense mechanism were verified.