Data-Driven Detection of False Data Injection Attacks in Smart Grids

Abstract

Cyber-physical attacks are the most significant threat facing the utilisation and development of the various smart grid technologies. Among these attacks, false data injection attacks (FDIAs) represent a major category, with a wide variety of types and effects. There has been extensive reporting on FDIAs recently. Several detection algorithms have been developed over the past few years to address this threat. In Chapter 2, this thesis starts by providing a deep analysis of the literature on these algorithms. The concluding remarks of this chapter present the main criteria that should be considered in developing future detection algorithms for FDIAs in different systems of smart grids. Following that, this dissertation proposes FDIA detection algorithms in the major systems in smart grids that are the most susceptible and vulnerable towards FDIAs. In wide-area monitoring systems, being able to promptly differentiate FDIA from normal grid contingencies is crucial for a grid operator to decide the correct response and reduce FDIA false alarms. In Chapter 3, two FDIA characterisation algorithms are developed to address this issue. The automatic generation control (AGC) is paramount in maintaining the stability and operation of power grids. FDIAs are particularly difficult to detect and represent a major threat to AGC systems. Chapter 4 proposes a novel spatio-temporal learning algorithm that can learn the normal dynamics of the power grid with AGC systems. It then utilises this unsupervised learned model in detecting FDIA affecting the AGC system. The utilisation of distributed generation units in power distribution systems has increased the complexity of system monitoring and operation. Numerous information and communication technologies have been adopted recently to overcome the associated challenges, but they have created wide opportunities for energy theft and other types of cyber-physical attacks. Chapter 5 utilises the developed spatio-temporal learning algorithm in Chapter 4 in detecting the various possibilities of FDIA affecting the distribution systems by evaluating the reconstruction error of each measurement sample. The proposed algorithm is data-driven, which makes it resilient against distribution systems’ uncertainties and nonlinearities. The collected results indicate a superior detection performance of the proposed detection algorithms compared to those in the literature.