| The primary goal of this research is to develop new energy management mechanisms that properly account for the decisions and control processes at the side of both customers and utility companies in the smart grid. While there has been considerable recent works on demand-side management, storage units integration, and related ideas, most existing works are based on classical game-theoretic concepts such as static noncooperative games and do not account for the real-world user behavior and its impact on the operation of the smart grid and of energy markets. Indeed, most of the existing models proposed so far are based on the ideal assumption that grid customers can make rational and perfect decisions. For instance, even if it is technologically beneficial for power companies to offer dynamic pricing mechanisms aimed at reducing peak-hour demand, customers may not subscribe to such features. For example, on a hot summer day, customers may not allow the power company to reduce their air conditioning usage during peak hours, despite the associated benefits to the grid and/or the lower offered prices. Accordingly, there is a need to develop a novel framework that captures realistic user behavior during energy management while taking into account the associated benefits and costs on both customers and the grid. Studying realistic user behavior in energy management faces numerous challenges. First, introducing storage units for energy trading between customers and the grid can lead to both competition and cooperation at different levels that can involve the complex interactions between customers, power companies, and energy providers. Second, due to the large-scale and heterogeneous nature of the smart grid, it is necessary to deploy distributed energy management protocols, in which the control center only provides a small amount of control information without operating at customers' sides. Finally, the involvement of customer-owned devices leads to uncertainties that stem from the unpredictable behavior of end-users.;In this thesis, we propose a novel approach based on behavioral game theory that can serve as a framework for capturing realistic user behavior in smart grid energy management. Taking such realistic decision-making settings into account allows us to go beyond classical game-theoretic concepts in order to explore how a user perceives the actions of its opponents and how this user evaluates its own utility function. In particular, we adopt the mathematical tools expounded by the Nobel-prize winning framework of prospect theory (PT), to study the decisions made by grid users and their impact on energy management.;Our primary results include the development of four game-theoretic models for studying user behavior in energy management. In the first model, we introduce a two-level approach that combines both auction theory and game theory to model and analyze energy trading markets between customers and the grid. For this first game, we study the various properties of the equilibrium and assess the performance of adopting a game-theoretic approach as opposed to conventional heuristic schemes. Then, in the second scenario, we study the use of PT as a tool for understanding the charging and discharging behavior of customer-owned storage units. In particular, we respectively develop two frameworks that enable us to consider the case in which a customer can have weighted observation on its opponents' operation as well as the case in which a customer can have its own, subjective evaluation on the gains and losses achieved from utilizing the storage unit. Our results show that ignoring user behavior during charging or discharging of storage units can lead to undesirable or unexpected performance in terms of both the load of the grid and the power company's revenues. Third, we study a demand-side management scenario in which customers can decide on whether or not to subscribe to demand-side management. For this application, we show that PT can provide a novel insight on the load reduction over hours, and due to different rationality settings, customers might take a more positive/negative participation in practice. Finally, we study how the use of PT can provide interesting insights on security problems such as the problem of hardware trojan detection. |
