Balancing Electric Scooter Battery Swapping Network by Spatio-Temporal Recommendation

As a promising battery-sharing service, battery swapping has become one of the most important electric-scooter (called e-scooter) energy refueling mechanisms. However, by analyzing a battery swapping dataset of 108,574 e-scooter drivers and 41,358 batteries and surveying 224 of these drivers, we find that drivers are facing insufficient energy replenishment during the battery swapping process, i.e., swapping out an undercharged battery from the battery swapping station. To identify the root cause, we perform a data-driven analysis and reveal an imbalance in spatial and temporal swapping behavior patterns, which causes a significant decrease in battery charging time. Inspired by these findings, we design iSwap, a novel spatio-temporal battery swapping recommendation system, which improves drivers’ swapping experience by proactively coordinating spatio-temporal unbalanced behavior patterns. iSwap considers not only drivers’ individual temporal preference for swapping time recommendation (i.e., when to swap) but also the complex mutual influence among drivers for swapping station recommendation (i.e., where to swap). To perform the two tasks efficiently, we propose a swapping behavior-aware phase demand balancing schema and a submodularity-based near-optimal swapping station recommendation algorithm. Finally, through the evaluation, iSwap improves the probability of swapping out a fully charged battery by 42.1% and driver satisfaction by 16.8% compared to the ground truth.