Modeling Charging Choices of Small-Battery Plug-In Hybrid Electric Vehicle Drivers by Using Instrumented Vehicle Data

Abstract

Methods for modeling plug-in hybrid electric vehicle drivers’ choices of whether to charge at the end of a trip are compared. Instrumented vehicle data from 125 preproduction Toyota Prius plug-in hybrids were combined with electric vehicle supply equipment (i.e., charging station) location data from multiple sources, including PlugShare and the U.S. Department of Energy. The effects of factors including battery state of charge (SOC), dwell time, and location on the probability that a driver will charge during a stop were then modeled for locations where charging was possible. The amount of energy that can be transferred during a charging session was a better predictor of charging behavior than SOC and dwell time considered independently. Results were sensitive to data and methods used to identify charging locations, which indicated that caution is needed in this area. Finally, a latent class logit model was investigated and found to provide a better fit to the observed data than did the mixed logit model. The results of the latent class model indicated that all users were more likely to charge after 8:00 p.m. and at times when doing so provided a larger gain in SOC. For some classes of users, charging was more likely at home or after the last trip of the day.