Feasibility Analysis of Taxi Fleet Electrification Using 4.9 Million Miles of Real-World Driving Data 2019-01-0392
Ride hailing activity is rapidly increasing, largely due to the growth of transportation network companies (TNC’s) such as Uber and Lyft. However, traditional taxi companies continue to represent an important market share of hailed trips. Columbus Yellow Cab (CYC), a taxi company in Columbus, Ohio, allows traditional taxi hailing as well as on-request rides using an app. Data from CYC trips was provided to the National Renewable Energy Laboratory (NREL) to analyze the potential for electrification of their fleet, currently comprised entirely of gasoline and hybrid vehicles. CYC data contained information describing both GPS trajectories and taxi meter information. The data spanned a period of 13 months, containing approximately 70 million GPS points, 840 thousand trips, and 170 unique vehicles. A CYC taxi was found to have an average daily Vehicle Miles Traveled (VMT) of 154 miles with an annualized VMT of 29 thousand miles. Additionally, meter information facilitated an understanding of vehicle deadheading; 46% of miles traveled were driven without a passenger. A variety of scenarios were evaluated using CYC data and the Electric Vehicle Infrastructure Projection Tool (EVI-Pro) to understand challenges and opportunities associated with operating an electrified taxi fleet. Factors such as access to home charging, the availability of public charging, charging level, and the presence of a fleet depot are shown to be major variables affecting successful electric fleet operation. The analysis indicates that 93.7% of taxi travel days can be successfully completed by a 250 range EV and that 10.3% of fleet vehicles require Direct Current Fast Charging (DCFC). Charging event locations are aggregated geographically to inform future infrastructure considerations. Analysis of the comprehensive CYC data set extends the previous literature surrounding TNCs and fleet electrification; existing studies are typically performed with less resolved real-world information given the low availability of ride hailing data.
Matthew Moniot, Clement Rames, Erin Burrell