In this dissertation, I present a body of work that advances our understanding of the technical and economic potential for vehicle grid integration based on a variety of methodological approaches that quantify the opportunity at multiple scales, across multiple geographies, and that cover scenarios with both personally owned plug-in electric vehicles (PEVs) and shared autonomous electric vehicles (SAEVs). The key research questions addressed in this dissertation include:

* How can charging infrastructure be cost-effectively deployed to maximize utilization and value to PEV drivers?

* How much flexibility exists in the charging demand from PEVs?

* What is the economic opportunity to manage the charging of PEVs to occur at lower cost time periods?

* How will fleets of electrified autonomous vehicles serving mobility on-demand differ in how they are managed to minimize the cost of charging or to serve as a source of electricity for buildings?

These questions are motivated by the fact that transportation electrification and emerging forms of mobility are dramatically changing how the transportation system is planned, operated, and analyzed. PEVs present new challenges and constraints around the siting and operation of refueling infrastructure. Electric load from PEVs can exacerbate grid congestion at either transmission or distribution scales if left unmanaged. Sharing and autonomy are changing mobility which will have unique implications for the grid integration of PEVs.

Meanwhile, there are strong social and environmental forces compelling planners, regulators, and private industry to electrify transportation as soon as possible. The transportation sector is the largest emitter of greenhouse gases in the United States. With the exception of the great recession, emissions in the transportation sector have been growing for the last three decades, in contrast to the electric power and industrial sectors which have been on a downward trend in emissions. Transportation, therefore, represents one of the primary challenges to achieving deep decarbonization of the U.S. economy.

In the electric power sector, policy and economic forces are upending incumbent generation technologies (coal and natural gas) in favor of lower cost and lower carbon alternatives, particularly wind and solar power. As these intermittent renewable resources increase in capacity, the incidence of renewable energy (RE) curtailment increases due to time periods when supply is greater than demand and generators are turned down or shut off from the level that they would otherwise be producing. Curtailment raises the overall system cost of supplying electricity. In addition, some utilities must meet an energy production standard to satisfy state mandates for renewable production. Renewable curtailment forces utilities to either acquire more RE or introduce sources of grid flexibility to relieve the curtailment. One low cost strategy to mitigate these challenges is to manage the temporal profile of electricity demand to make use of the renewable resources when they are available.

PEVs are generally analyzed through modeling using one of two approaches, statistical modeling and activity-based modeling. Statistical models typically summarize or infer travel patterns from travel survey data and use them to characterize the need for PEV charging and the temporal opportunities to charge. The key disadvantage of such approaches is that they cannot account for the individual mobility constraints of travelers and they typically require an assumption that charging infrastructure is unlimited. Another common approach is to develop Markov Chain models of mobility and PEV charging. In these models, transitions between states are treated as random events. Because they lack a representation of the causal mechanism for these transitions, these models are difficult to generalize and their utility is degraded if applied in prospective contexts assuming a transportation system with dramatically different characteristics than present.

Activity-based models make use of travel diaries from surveys or GPS data logging which are then provided as input to mobility and charging simulations. Agent-based models are a subset of activity-based models, in so far as they treat travelers individually and require a representation of each individual's activity schedule in order to model the travel necessary to engage in those activities. What distinguishes agent-based models are two key features: 1) wrapping the individuals in a virtual environment (e.g. the transportation system) with detailed representation of transportation supply and 2) dynamically simulating the agents' interactions with the virtual environment and with each other. These interactions open the opportunity to model the choices of the agents based on empirical studies of human behavior as well as to make agent behavior contingent on the time-evolving state of their environment and other agents.

In the electric power and grid modeling domain, load from PEVs are typically represented as static or derived from very rudimentary estimation techniques. Studies either ignore flexibility entirely or they make simplistic assumptions about the timing and degree to which PEV load can be shaped. The inaccuracy in these modeling choices have had a relatively low impact in the recent past due to the still relatively low penetration of PEVs in the national vehicle fleet. But within a decade it will no longer suffice to ignore or simplify PEV load, which could eventually make up more than 20% of U.S. electricity demand.

This dissertation addresses these gaps by coupling models of electric mobility and the grid at multiple scales. Each paper presented in this dissertation was produced in collaboration with co-authors across multiple projects and contexts. I employ reduced-form models in the context of optimization to solve the charger scheduling and vehicle mobility problems, as well as detailed agent-based models that simulate context-specific traveler behaviors and the dynamics of resource-constrained charging infrastructure.

To address the infrastructure siting problem, I develop a spatially explicit agent-based simulation model that represents charging infrastructure, charging behavior, competition for scarce chargers, and driver adaptation. A differential evolution and a heuristic optimization scheme are employed to find a cost-effective distribution of charging infrastructure.

I then address the question of flexibility in two ways. First I develop a scheme for optimizing the charging profiles of individual PEV drivers based an objective that simultaneously considers the costs of charging and the benefits associated with providing ancillary services to the grid. Then I employ a much higher fidelity approach to simulate both the electrified mobility system as well as the power sector. I develop the BEAM modeling framework (Behavior, Energy, Mobility, and Autonomy), which is an agent-based model of PEV mobility and charging behavior designed as an extension to MATSim (the Multi-Agent Transportation Simulation model). I apply BEAM to the San Francisco Bay Area and conduct a preliminary calibration and validation of its prediction of charging load based on observed charging infrastructure utilization for the region in 2016. I link the BEAM model with PLEXOS, an industry standard production cost model that accurately characterizes grid dispatch constraints.

Finally, I consider the impact of grid-integrated fleets of SAEVs providing mobility on-demand. In two separate studies I develop models to consider how such fleets could be used to serve building energy demand during power outages as well as a more general analysis of the battery and charging infrastructure requirements to serve nationwide mobility.

The key findings across all of this work are the following:

* In today's energy markets, PEV flexibility can reach values of $155/year/vehicle for NYISO and $98/year/vehicle for CAISO. The annual cost savings due to optimizing dispatch come more from savings on the price of energy (74% in CAISO and 61% in NYISO) but providing ancillary services (in the form of regulation) also contributes value to the solution (26% in CAISO and 39% in NYISO).

* When we project the energy market of California to a future year when renewables make up 50% of the annual energy produced, PEV flexibility is even more beneficial to the power sector, primarily in lowering grid operating cost and the amount of RE that must be curtailed to avoid over-generation when supply and demand are mismatched. For example, if treated as flexible loads, 2.5 million smart charging PEVs avoid 50\% of incremental system operating costs annually and reduce renewable energy curtailment by 27% annually relative to when the same number of unmanaged charging PEVs are added to the grid.

* When SAEVs serve power to buildings during an extreme outage, the fleet can generate 32%-40% more revenue than is earned serving mobility alone. While the overall value of providing on-demand power depends on the frequency and severity of power outages, the results show that serving power demand during large-scale outages can provide a substantial value stream, comparable to what can be earned providing grid services.

* All mobility in the United States currently served by 276 million personally owned vehicles could be served by 12.5 million SAEVs at a cost of $0.27/vehicle-mile or $0.18/passenger-mile. The energy requirements for this fleet would be 1142 GWh/day (8.5% of 2017 U.S. electricity demand) and the peak charging load 76.7 GW (11% of U.S. power peak).

In total, this body of work contributes new insights into the opportunity for electric mobility to reduce the cost of operating the electric grid, enabling deeper and faster adoption of renewable power in the electric sector, and providing reliable mobility to travelers in the transportation sector. The domain of vehicle grid integration is still relatively new, there are many areas of research that require additional attention, such as increased research on traveler preferences around PEV charging, the intersection between electric mobility and the distribution grid, electrification of medium and heavy duty vehicles, as well as properly incentivizing electric vehicles to ride hail drivers in the gig economy.