Fair Mechanisms for Smart Grid Congestion Management

Abstract

We consider energy systems in the built environment. With the transition to a more sustainable, distributed, and 'smart' energy system, such local grids are undergoing significant changes. Among other developments, the new role of end-users as 'prosumers' - users that can either produce or consume power depending on the situation - is turning energy systems in the built environment into autonomous microgrids with complex internal interactions.

One of the primary challenges for these local grids is maintaining grid stability, which requires constant balancing of supply and demand. Because local grids were not designed for distributed energy generation and large loads such as electric vehicle charging, their limited capacity is now leading to congestion. Since the responsibility for resolving congestion falls increasingly on the individual prosumers and their flexibility, the concept of fairness must take a central role in congestion management.

In this dissertation we present our research on supply-demand matching mechanisms for fair congestion management. The local networks populated by users can be represented by radial multi-agent commodity flow systems. For the resource allocation problems in this setting we draw on the fields of mechanism design and fair division to design provably fair congestion management mechanisms. We evaluate the merit of different notions of fairness and present algorithmic mechanisms that align agent incentives with fair allocations.

We find that notions of fairness regarding congested commodity flow networks can either focus on local or global fairness. Agents can have differing opinions on the two, depending on how wide they draw the circle of peers that they compare themselves to. We find that the mix of producers and consumers requires slight adaptation of notions of fairness, with agents envying one group while welcoming the other. Furthermore, we find that it is possible to combine notions of fairness with welfare optimization by letting individual agents decide which of the two is more important, and protecting their fair shares.

We are able to use the radial structure prevalent in energy systems in the built environment to design algorithmic mechanisms of consistently low computational complexity. The congestion solutions of these mechanisms satisfy different local and global fairness criteria, for which we provide rigorous proofs. We prove that our mechanisms are individually rational and, for variations of egalitarian fairness, also incentive compatible. Finally, we introduce a congestion aftermarket where agents compensate their peers for flexibility.