A Deep Learning-Based Classification Scheme for False Data Injection Attack Detection in Power System
Yucheng Ding,Kang Ma,Tianjiao Pu,Xinying Wang,Ran Li,Dongxia Zhang
DOI: https://doi.org/10.3390/electronics10121459
IF: 2.9
2021-06-18
Electronics
Abstract:A smart grid improves power grid efficiency by using modern information and communication technologies. However, at the same time, due to the dependence on information technology and the deep integration of electrical components and computing information in cyber space, the system might become increasingly vulnerable to cyber-attacks. Among various emerging security problems, a false data injection attack (FDIA) is a new type of attack against the state estimation. In this article, a deep learning-based identification scheme is developed to detect and mitigate information corruption. The scheme implements a conditional deep belief network (CDBN) to analyze time-series input data and leverages captured features to detect the FDIA. The performance of our detection mechanism is validated by using the IEEE 14-bus test system for simulation. Different attack scenarios and parameters are set to demonstrate the feasibility and effectiveness of the developed scheme. Compared with the artificial neural network (ANN) and the support vector machine (SVM), the experimental analyses indicate that the results of our detection mechanism are better than those of the other two in terms of FDIA detection accuracy and robustness.
engineering, electrical & electronic,computer science, information systems,physics, applied