False Data Injection Attacks Detection with Deep Belief Networks in Smart Grid

Abstract

As a typical information physics system, the smart grid can realize the two-way transmission of information. The deep integration of the information physics system brings more risks to the system while improving the automation management level of the power system. The false data injection attack is a new type of attack method for power system state estimation. The attacker can bypass the traditional detection method to change the state estimation result, so that the control center makes wrong decisions and threatens the safe operation of the power grid. In this paper, we focus on the detection of false data injection attacks in smart grids. A DBN-based attack detection method is proposed. Unsupervised learning is performed from the bottom of the restricted Boltzmann machine to provide initial weight for the network. The backpropagation algorithm propagates the error from top to bottom and fine-tunes the model parameters. To evaluate the effectiveness of the proposed detection method, simulation experiment is performed in the IEEE standard test system. We set different contrast scenarios to verify the feasibility and effectiveness of the detection model. The results demonstrate that the proposed approach achieves better performance with comparison to the SVM based detection approach.