Charging an Electric Vehicle-Sharing Fleet

Published: 16 June 2021

Electric vehicle (EV) sharing programs rely on publicly available charging infrastructure. Yet often there simply isn’t enough infrastructure to go around, creating a major barrier to success. In 2016, the vehicle sharing company Car2Go sold off its fleet of EVs in San Diego largely because the city’s charging infrastructure couldn’t keep pace with demand. Or so it seemed. With too many vehicles being dropped off at a few charging points in high-use areas, delays grew.

Meanwhile, Car2Go was only re-charging EVs once their energy levels dropped to 20%. Computational analysis by Desautels Professor Wei Qi has shown that Car2Go could have resolved the problem by optimizing their approach. By plugging in EVs once when they reach an energy level of 40%, the company could have powered up its vehicles more quickly and eliminated the delays. Charging cars sooner could help boost profits too. The research shows that customers choose shared EVs based in part on their charge, and the longer per-charge range could result in more overall use.

Authors: L. He, G. Ma, Wei Qi and X. Wang

Publication: Manufacturing and Service Operations Management, Volume 23, Issue 2, March 2021, Pages 471-487


Problem definition: Many cities worldwide are embracing electric vehicle (EV) sharing as a flexible and sustainable means of urban transit. However, it remains challenging for the operators to charge the fleet because of limited or costly access to charging facilities. In this paper, we focus on answering the core question how to charge the fleet to make EV sharing viable and profitable. 

Academic/practical relevance: Our work is motivated by the setback that struck San Diego, California, where car rental company car2go ceased its EV-sharing operations. We integrate charging infrastructure planning and vehicle repositioning operations that were often considered separately. More interestingly, our modeling emphasizes the operator-controlled charging operations and customers' EV-picking behavior, which are both central to EV sharing but were largely overlooked. 

Methodology: Supported by the real data of car2go, we develop a queuing network model that characterizes how customers endogenously pick EVs based on energy levels and how the operator implements a charging-up-To policy. The integrated queuing-location model leads to a nonlinear optimization program. We then propose both lower and upper bound formulations as mixed-integer second-order cone programs, which are computationally tractable and result in a small optimality gap when the fleet size is adequate. 

Results: We learn lessons from the setback of car2go in San Diego. We find that the viability of EV sharing can be enhanced by concentrating limited charger resources at selected locations. Charging EVs either in a proactive fashion or at the 40% recharge threshold (rather than car2go's policy of charging EVs only when their energy level drops below 20%) can boost the profit by more than 15%. Moreover, sufficient charger availability is crucial when collaborating with a public charger network. Increasing the charging power relieves the charger resource constraint, whereas extending per-charge range or adopting unmanned repositioning improves profitability. Finally, we discuss how EV sharing operations depend on the urban spatial structure, compared with conventional car sharing. 

Managerial implications: We demonstrate a data-verified and high-granularity modeling approach. Both the high-level planning guidelines and operational policies can be useful for practitioners. We also highlight the value of jointly managing demand fulfillment and EV charging.

Back to top