Detection Method Based On Deep Learning Of Malicious Data Attack In Smart Grid

According to the various measurement information of the power system,the smart grid uses the state estimation to obtain the current operating state of the power grid,so accurate state estimation is critical to maintaining the legitimate operation of the smart grid.The Malicious data injection attack can tamper with the collected measurement information from data acquisition and monitoring system to threat security of the power grid.Therefore,it is of great theoretical and practical significance to analyze the mechanisms of malicious data injection attack and to research the detection model to ensure the safe operation of smart grid.In this paper,kernel principal component analysis(KPCA)is used to derive the network topological structure matrix and construct blind false data injection attacks further,use invertible autoencoder to reduce the dimensionality of the original data,and use Recurrent Neural Network(RNN)base on Long and Short Time Memory(LSTM)algorithm for malicious data injection attack detection.The main work is as follows:1.This paper analyzed the formation mechanism of malicious data attacks.According to the fact that it is difficult to obtain the information of power system topology in practical situations,KPCA is used to derive the topology matrix of power system and the blind malicious data injection attacks is constructed according to the calculated topology matrix.2.In view of the characteristics that the power system original data such as high dimension and large amount of data,invertible autoencoder is used to reduce the dimensionality of the original data.The invertible autoencoder is built on the multi-layer model,the original data set is mapped into a random feature space and the coding layer is continuously updated through the learning model.After the weight value is updated in the coding layer,the generated low-dimensional data can represent the original input data.3.Considering that the traditional bad data detection method cannot detect the malicious data injection attacks,the recurrent neural network is used for malicious data injection attack detection,and the long and short time memory model is adopted to solve the data long-distance dependence.A large number of simulation experiments are carried out in the IEEE-14 and IEEE-118 bus test systems.Compared with other algorithms under different attack conditions,the proposed method is validated by experimental results.