Smart Grid System Security Protection by Deep Neural Network
Fan Li, Yi Chao Wu, Yihan Liu, Yingjie Tian, Hongshan Yang · 2021
The increasing number and types of attacks in the smart grid lead to the decline of its security. Therefore, in order to improve the security performance of the power grid, this paper first proposes an attack detection architecture to detect attack types. Then the representative features of different types of attacks are extracted on the basis of deep belief network and support vector machine. Finally, the attack behavior is identified by deep neural network. The main contributions of this paper are as follows: 1) in view of the low accuracy of behavior detection caused by a large number of attacks and long data segments, we enhance the data information of the original model by generating an improved confrontation network. It improves the stability of attack behavior recognition. 2) in the known attack behavior experiments with complex types and high concealment, we use the nonlinear iterative algorithm based on deep belief network to select features automatically to improve the recognition accuracy of covert behavior. 3) for the detection task of unknown attack behavior, we propose a method of extracting behavior features based on support vector machine, which significantly improves the detection accuracy and proves the effectiveness of our model for representative behavior feature extraction.