Toward Urban Electric Taxi Systems in Smart Cities: the Battery Swapping Challenge
Yang Wang,Wenjian Ding,Liusheng Huang,Zheng Wei,Hengchang Liu,John A. Stankovic
DOI: https://doi.org/10.1109/tvt.2017.2774447
IF: 6.8
2017-01-01
IEEE Transactions on Vehicular Technology
Abstract:Despite the clear benefits of electric vehicles (EVs) in terms of reducing greenhouse gas emissions and traditional energy consumptions, the popularization of EVs remains a challenge in the short run. There are currently two major ways of refueling electric taxicabs (ETs): recharging and battery swapping. ETs usually choose battery swapping so as not to waste precious time during their daily work. While previous studies focused on planning battery swapping stations for private EVs, we investigate ways of supporting the upgrade of an entire urban taxi system, with demands differing both in scale and nature. Furthermore, most ETs are fully charged in the early morning and need to swap battery in approximately the same time period of the day, which results in a bottleneck for battery swapping and hurts the quality of service (QoS) of the taxicab fleet. With this insight, we model the ET fleet as a mobile sensor network, analyze the historical sensing data of taxi routes, and evaluate the battery swapping demand profile and the power consumption of individual taxis, as well as the driving time between positions in the road network. Based on these inputs, we propose a method to calculate an optimized battery swapping station scheme, then describe a real-time algorithm to schedule a subset of the unoccupied taxicabs to swap batteries early by giving them allowances to avoid congestion. Our strategies are then evaluated via a real-world 366-day, 3 976-taxi dataset. The results demonstrate that compared to uniform deployment, our planning scheme reduces the average time cost by 67.2%; and based on our deployment, our scheduling strategy decreases the in-station queuing time by 45.63%, 52.26%, and 29.57%, and the average driving time cost to the stations by 42.3%, 43.69%, and 28.77% compared to actual taxicab refueling, random scheduling, and a static heuristic strategy, respectively. Furthermore, our approach reduces the number of battery swapping that last more than 30 min by 40.8%, which is a significant improvement on the QoS of the whole system.