Power trip : smart management of shared autonomous electric vehicle fleets




Dean, Matthew David

Journal Title

Journal ISSN

Volume Title



The climate crisis requires substantial shifts in the transportation and energy sectors. Greater use of intermittent renewable energy sources requires demand- and supply-side flexibility in electricity markets. Deployment of on-demand, shared, fully automated, and electric vehicle (SAEV) fleets offers natural synergies in meeting such challenges. Smart charging (and discharging) of electric vehicles (EVs) can shift loads away from peak demand to reduce, or at least delay, expensive infrastructure upgrades, while fleet managers lower emissions and power costs in real time. This dissertation explores (1) optimization-based idle-vehicle dispatch strategies to improve SAEV fleet operations in the Austin metro, (2) integration of power and transportation system (EV-use) modeling across the Chicago metro area, and (3) a case study of demand response participation and charging station siting in a region with multiple energy suppliers. Optimizing SAEV repositioning and charging dispatch strategies jointly lowered rider wait times by 39%, on average, and increased daily trips served per SAEV by 28% (up to 6.4 additional riders), compared to separate range-agnostic repositioning and heuristic charging strategies. Joint strategies may also decrease the SAEV fleet’s empty travel by 5.7 to 12.8 percentage points (depending on geofencing and charging station density). If fleets pay dynamic electricity prices and wish to internalize their upstream charging emissions damages, a new multi-stage charging problem is required. A day-ahead energy transaction problem provides targets for a within-day idle-vehicle dispatch strategy that balances charging, discharging, repositioning, and maintenance decisions. This strategy allowed the Austin SAEV fleet to lower daily power costs (by 15.5% or $0.79/day/SAEV, on average) while reducing health damages from generation-related pollution (2.8% or $0.43/day/SAEV, on average). Fleet managers obtained higher profits ($8 per SAEV per day) by serving more passengers per day than with simpler (price-agnostic) dispatch strategies. This dissertation also coupled an agent-based travel demand simulator (POLARIS) with an electricity grid model (A-LEAF) to evaluate charging impacts on the power grid across seasons, household-EV adoption levels, SAEV mode shares, and dynamic ride-sharing assumptions in 2035 for the Chicago, Illinois metro. At relatively low EV penetration levels (8% to 17%), an increase in electricity demand will require at most 1 GW of additional generation capacity. Illinois’ transition to intermittent variable renewable energy (VRE) and phase-out of coal-fired power plants will likely not noticeably increase wholesale power prices, even with unmanaged personal EV charging at peak hours. However, wholesale power prices will increase during peak winter hours (by +$100/MWh, or $0.10/kWh) and peak summer hours (+$300/MWh) due to higher energy fees and steep congestion fees on Illinois’ 2015-era transmission system. Although a smart-charging SAEV fleet uses wholesale prices to reduce electricity demand during peak hours, spreading charging demand in hours before and after the baseline peak creates new "ridges" in energy demand, which raise prices for all. These simulation results underscore the importance of investing in transmission system expansion and reducing barriers to upgrading or building new transmission infrastructure. If vehicles and chargers support bidirectional charging, SAEVs can improve grid reliability and resilience at critical times through demand response (DR) programs that allow load curtailment and vehicle-to-grid (V2G) power. Scenario testing of DR requests in Austin ranging from 1 MW to 12 MW between 4 and 5 PM reveals break-even compensation costs (to SAEV owners) that range from $86/kW to $4,160/kW (if the city imposes unoccupied travel fees), depending on vehicle locations and battery levels at the time of the DR request. Smaller requests can be met without V2G by reducing charging speeds, usually from 120 kW speed to Level 2 charging. Finally, an incremental charging station heuristic was designed to capture differences in land costs and electricity rate structures from different energy suppliers in the same region. The daily amortized costs over 10 years of hardware, installation, and land costs were estimated to be nearly $0.30/SAEV/day, compared to $0.38/SAEV/day with a baseline heuristic strategy ignoring land costs and marginal costs of expanding existing sites. SAEV charging costs showed no substantial difference between heuristic strategies, although combined daily energy fees were more expensive at $0.43/SAEV/day. Including land costs in charging station investment heuristics is necessary, and modelers should include spatially varying energy prices since the average daily per-vehicle energy costs are higher than the physical station costs. Taken together, this dissertation’s contributions offer hope for a decarbonizing world that provides affordable, clean, and convenient on-demand mobility.


LCSH Subject Headings