Coordination strategies for plug-in hybrid vehicles and decentralized generation in a Smart Grid.

Electric vehicles will play a key role in the electricity grid of the future. In recent years, the increase in the amount of electric vehicles is gaining momentum, as global environmental concerns are getting stronger, and automotive OEMs (Original Equipment Manufacturers) are preparing for mass-production. This vast increase of grid-connected vehicles offers opportunities to use electric vehicles as a large-scale distributed storage system. In a liberalized electricity market, aggregators are typically seen as the actors who will manage this storage system.

The central problem addressed in this dissertation is an aggregator’s large-scale control of the power transfer between electric vehicles and the grid. To control this power transfer, the aggregator requires a set of algorithms, models and interactions, called a “GIV (Grid-Integrated Vehicle) control approach”. In designing a GIV control approach for a large amount of electric vehicles, classic optimization techniques fall short due to their limited applicability in high-dimensional online control problems. Therefore, a scalable GIV control approach is required, which achieves consistent high quality decision making for an increasing amount of electric vehicles. A particular challenge in the design of such an approach is the need for frequent adjustment of an aggregator’s control decisions to a continuously changing environment, in which electric vehicles are being plugged in and unplugged by their owners.

This dissertation proposes three GIV control approaches, each designed to provide large-scale control of EVs in different business cases of an aggregator. The first GIV control approach is a three-step market-based approach to charge electric vehicles in response to a dynamic electricity pricing scheme. The second GIV control approach is a reinforcement learning approach to learn a cost- effective day-ahead schedule. The third GIV control approach is a bin-based scheduling approach to provide regulation services with electric vehicles. While each GIV control approach is applicable in a different business case, their design is based on a common blueprint for large-scale control, called “aggregate and dispatch”. In this type of control, an aggregator calculates aggregated decisions for the EV fleet, which are translated to individual EV decisions by a dispatch mechanism.

The third approach has been validated and compared with other approaches in the EV fleet at the University of Delaware. In-field results show that this approach is applicable in a real-world scenario, while combining the advantages of both centralized and distributed GIV control approaches.