Original scientific paper

https://doi.org/10.17559/TV-20260511003581

Battery Swapping Station Siting from Ride-Hailing Trajectories: A Robustness Study for Wuhan, China

Umut Ergunsü ; School of Business, Geely University of China, No. 123, Section 2, Chengjian Avenue, Chengdu Eastern New Area, Chengdu 641423, Sichuan Province, P. R. China *

* Corresponding author.

Abstract

Trajectory-based studies site battery swapping stations without asking how far the data determine the result. This paper asks which quantities the data identify, and its answer holds whatever the stops represent. Three days of ride-hailing trajectories from 29 143 vehicles in Wuhan are processed by cleaning, gap-aware stop detection, DBSCAN noise filtering and greedy p-median optimisation. Perturbing detection changes coverage at every scale by at most 4,0 points, so the demand surface is robust. The proven optimum differs from the greedy set in 13 of 40 sites for a 2,8 per cent gain, coverage-neutral perturbations relocate up to 25 stations, and only 6 of the 40 are selected by nearly all of 13 alternative solutions, so station identity is not stable. A knee criterion returns 19 to 23 stations and a proportional-decline rule 10 to 60. Forty stations cover 80,1 per cent of demand within three kilometres, 15,2, 14,2 and 6,0 points above random, district-proportional and k-medoids placement. Detection must split segments at sampling gaps, or mean duration inflates 29-fold. The result is a stable coverage curve from data cities already hold and a quantified boundary of what they leave open.

Keywords

battery swapping station; DBSCAN; p-median optimisation; ride-hailing trajectories; stop detection; Wuhan

Hrčak ID:

353079

URI

https://hrcak.srce.hr/353079

Publication date:

7.11.2026.

Visits: 0 *