Detection of False Data Injection Cyberattacks: Experimental Validation on a Lab-scale Microgrid

Abstract

Instantaneous monitoring of smart microgrids is a crucial task of system operators to ensure an acceptable level of security and reliability in the system. To this end, modern communication platforms are being introduced to facilitate the data exchange between field devices and microgrid control center (i.e., a cyber-physical power system). This is where an adversary can take advantage of the smart system to stealthy compromise the sensors’ readings in the cyber layer resulting in different operational issues, which can lead to cascading failures or blackout in extreme cases. As a step toward protecting modern power networks, this article proposes a detection approach, which is oriented toward recurrent neural networks (RNNs), against false data injection (FDI) cyberattacks targeting a lab-scale microgrid developed in Southern Illinois University, Carbondale, IL, USA. To significantly enhance the quality of obtained results in order to be adapted with realistic systems, the proposed framework (i.e., the FDI cyberattack and the RNN-based detection approach) is set up as a real-time digital simulator in the form of hardware-in-the-loop (HIL) testbed, which utilizes the physical components of the developed lab-scale microgrid. In this regard, the OP4510 HIL simulator is upgraded by a 16-channel Imperix Power Interface and a dual-port PCI-E X1 Gigabit Ethernet to perform the FDI attacks on the sensors’ readings. Moreover, the proposed detection framework is able to distinguish FDI attacks from other transient events (e.g., alteration in demand level). The experimental results demonstrate the effectiveness of the developed false data detection approach in different scenarios.