Detection of False Data Injection Attacks in Automatic Generation Control Systems Considering System Nonlinearities

Abstract

Maintaining the power system frequency around its nominal value is a very critical issue for the system stability. This operation is performed by the Automatic Generation Control (AGC) system. A cyber attack on the AGC system may affect the whole stability and economic operation of the power system. This paper proposes a method using Recurrent Neural Networks to detect False Data Injection (FDI) attacks in AGC systems. The novelty of this work over other approaches is that the nonlinearities of the AGC system are considered, which make it difficult to use the conventional approaches to detect FDI in case of considering the nonlinearities. The AGC of a two-area power system is used and the results show that the proposed approach succeeded to detect FDI in AGC system with an accuracy of 94%.