Scott Kelley
Michael Kuby
 School of Geographical Sciences and Urban Planning, Arizona State University, Tempe, AZ, United States
 
 
 Highlights

•We surveyed drivers of compressed natural gas vehicles while refueling.
•We mapped their stops pre-/post-refueling in GIS and generated shortest paths.
•We isolated drivers who refueled either closest to home or most on their way.
•CNG drivers refuel at stations on the way instead of close to home by a 10:1 margin.
•This suggests using flow-based models over point-based for optimal station location.


Keywords
Station; Infrastructure; Convenience; Deviation; Location; Alternative fuel vehicle


Abstract
Limited refueling infrastructure is an oft-cited barrier to alternative fuel vehicle (AFV) adoption, but empirical data on AFV driver refueling behavior are rare. To address this need, we surveyed 259 drivers of compressed natural gas (CNG) vehicles in Southern California at five stations across the metropolitan area. The key survey questions concerned the stops immediately before and after refueling and the driver’s home location. Using GIS, we analyze the least travel-time routes and the station chosen to provide insight into what drivers consider to be their most convenient refueling location. Specifically, we focus on whether they select stations nearest to home or on routes that require the least deviation. When faced with a choice between the two—that is, when no station satisfies both criteria—we found that ten times as many CNG drivers selected the station most on their way between their origin and destination than chose the station closest to their home. This finding supports the notion that optimal location models for planning early AFV refueling infrastructures should maximize convenience by serving the routes that drivers use frequently rather than their home locations, and that locations near high-volume roads may be ideal candidates for early station sites.



Article Outline
1. Introduction
2. Data and methodology
2.1. Survey
2.2. Deviations
2.3. Closest facility vs. least deviation analysis

3. Results
3.1. Analysis of station chosen
3.2. Comparison of the four groups
3.3. Comparison of the two groups faced with a choice
3.4. Subjective vs. objective detours

4. Discussion
5. Conclusion
Acknowledgements
References


Figures

Fig. 1.

Locations of stations where CNG surveys were collected, as well as other CNG refueling stations with available public refueling. Note the competing stations surrounding the SoCal Gas Trillium station in Anaheim.

Fig. 2.

Comparison of least travel-time direct path and refueling path, which forms the basis for deviation calculations.


Fig. 3.

Example of a refueling route from an origin at work (1) to a destination at home (2) where drivers are faced with a choice between a station that requires the least deviation (Burbank) or is closest to home (Glendale).


Fig. 4.

Scatter plot of difference (in minutes) between least deviation route and route traveled vs. difference (in minutes) of travel time from station to home and closest station to home. Note that there are 102 drivers at y = 0 (least deviation station), and 2 drivers at y < 6 s longer than their least deviation. There are 10 drivers at x = 0 (station closest to home), and one driver at x < 2 s farther than the closest station.

Fig. 5.

Desire lines graphic of CNG home locations and station at which driver refueled. Note the other publicly available CNG stations on the map, which represent other candidate sites for CNG refueling. Fraction in legend shows the number of drivers who refueled at the station closest to home out of the total number of drivers surveyed at that station.



Tables

Table 1. Deviation, closest facility, and least deviation analysis results for each station.

Table 2. Categorization of refueling station selection of all CNG drivers surveyed. The two cells of particular interest, in which drivers were faced with a choice of either closest to home or least deviation and chose one or the other, are in white, while the other cases are shaded.

Table 3. Incorporation of marginal cases into the absolute 2×2 classification, by rank of stations.

Table 4. Summary statistics for independent groups selecting a refueling station.

Table 5. Primary reason for choosing refueling station.

Table 6. Difference of means (independent samples t-test) results for choice groups.