Date of Award
Doctor of Philosophy (PhD)
Electrical and Computer Engineering
Hayat, Majeed M.
Critical infrastructures such as smart grids rely heavily on the seamless interaction between the grid subcomponents, i.e., the communication networks which transfers information from and to the grid, and the human operators/AI agents for taking necessary control actions. Smart grids are prone to cascading failures, which trigger from a few initial the tripping of a few transmission lines or generators, creating a ripple effect in the entire network, which may, in turn, lead to a total blackout. Having additional information through the communication network increases the probability of taking better control actions (e.g., effective load shedding and other protection mechanisms), which increases the reliability of the grid. On the other hand, enhancing the smart grid's communication capability increases the risk of harm through cyberattack and other faults in the communication network. A fundamental question is how can we balance the trade-off between grid’s performance enhancement and robustness to information infidelity? In this dissertation, we develop a predictive analytic, scalable and tractable Markov-chain model for cascading failures in smart grids including the role of the human operators, while taking into account the benefits and harm of the communication network (e.g., supervisory control and data acquisition). The state transition probabilities of the Markov chain captures the benefits and added vulnerabilities resulting from the communication network. A detailed mapping between power-grid states and the operators’ response has been established that allows capturing a wide range of operator behavior and their probabilities into in the dynamics of the Markov chain. The model shows the existence of a point of diminishing returns beyond which the harm of cyber threat and human errors outweighs the benefits of having information. An optimal level of inter-connectivity is achieved between the power grid and the communication network minimizing the expected value of the transmission-line failures.
Available for download on Friday, May 06, 2022