Graduation Year

2023

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Electrical Engineering

Major Professor

Mia Naeini, Ph.D.

Committee Member

Ismail Uysal, Ph.D.

Committee Member

Yasin Yilmaz, Ph.D.

Committee Member

Sriram Chellappan, Ph.D.

Committee Member

Kaiqi Xiong, Ph.D.

Keywords

Cyber Security, Cyber-physical Systems, Data Analysis, Power Grid, Situational Awareness

Abstract

Situational awareness in a large, dynamic, and complex cyber-physical critical infrastructure, such as a smart grid, is vital for ensuring its smooth and uninterrupted operation. With the evolving realities of the modern-day smart grids, new challenges associated with the situational awareness of these systems are emerging that demand intelligent and efficient solutions. This dissertation intends to address several problems for enhancing situational awareness by studying the dynamic interaction among the components of the smart grids through energy data analytics using various data-driven, machine learning, and graph signal processing (GSP) techniques. The presented work provides valuable insight into the data-driven analysis of the dynamics of cyber-physical power systems and contributes to the research regarding the security and reliability of smart grids.

Variations in load and generations as well as the operating states and conditions of the grid equipment, and the weather and environmental factors make the smart grid's dynamics stochastic with complex interactions among their components. This dissertation attempts to understand this dynamicity and interactions using numerical and analytical approaches by exploiting the measurement data captured by numerous sensors deployed throughout the system. The analysis of the correlation among the power system states is one of the energy data analytic tools used in this dissertation to study the behavior of the power system in normal operating conditions as well as under cyber and physical stresses. However, since the smart grid is a networked system, introducing the knowledge of topology and connectivity of its components in the analyses facilitates a better understanding of the system's behavior. GSP enables an explicit inclusion of the topological and connectivity information by extending the theories of classical signal processing to the irregular graph domain. Modeling the power grid as a graph by considering the buses as the nodes and the transmission lines as the edges, the power system measurements and states can be viewed as graph signals defined in the non-Euclidean vertex space of the graph. In this dissertation, both the correlation-based and the GSP-based study of the power system have been utilized in several applications related to the security and reliability of the smart grid.

Four specific applications for enhancing situational awareness in smart grids towards security and reliability stresses are studied. The first application is the data-driven detection and location identification of cyber attacks and physical stresses in the system. Timely detection and precise localization of stresses and anomalies are crucial for the quick restoration of the grid to its normal operating condition. As soon as a stress is detected and located in the system, the next task should be its proper classification and characterization for determining the best response, the root cause, and predicting similar scenarios in the future. The third application is the recovery of smart grid's states that are missing or corrupted due to cyber-attacks and physical damages to the measurement devices to ensure the observability of the system, which is crucial for monitoring and operation purposes. The final application is analyzing the nature of propagation of a single bus perturbation through the system. Moreover, in some of the aforementioned applications, machine learning and neural network models are used along with feature extraction techniques using GSP and correlation-based methods. In the majority of the studies, the problems are approached using analytical analyses, which are then verified through experiments by simulations. The results of the experiments have been presented, interpreted, and compared with the benchmark techniques. Future work directions are also discussed for each application.

Download

Share

COinS