Abstract
Battery Swapping and Charging Stations (BSCS) offer a solution to electric vehicle (EV) range anxiety by providing rapid battery replacement. However, optimizing station capacity—balancing parking, swapping robots, chargers, and batteries—is a complex challenge due to the coupled nature of EV arrivals and battery circulation. This paper introduces a Mixed Queueing Network (MQN) model and uses asymptotic analysis to evaluate performance as the number of batteries grows. The study identifies two distinct operating modes—charging-limiting and swapping-limiting—and provides analytical bounds to guide infrastructure planning.
Introduction
The adoption of Electric Vehicles (EVs) is often hindered by "range anxiety" and long charging times. Battery Swapping and Charging Stations (BSCS) address this by allowing drivers to swap depleted batteries for fully charged ones in minutes, similar to refueling a traditional internal combustion engine vehicle. Despite their benefits, building these stations requires significant capital investment.
To ensure a high Quality of Service (QoS), operators must minimize the "blocking probability"—the likelihood that an arriving EV is turned away because the station's buffer is full. This requires precise planning of four interdependent variables: the number of parking spaces, swapping islands, chargers, and spare batteries. Because the flow of arriving EVs and the internal circulation of charging batteries are tightly coupled, standard queueing models are insufficient to predict performance.
System Modeling and Methods
The authors model the BSCS as a Mixed Queueing Network (MQN). This model is "mixed" because it combines an open queue (EVs arriving from the outside world, receiving service, and departing) with a closed queue (a fixed number of batteries circulating between depleted and fully charged states).
As illustrated in Figure 1, the system architecture consists of several stages:
- Open EV-Queue: Arriving EVs wait in a buffer of size $V$.
- Swapping Service: $S$ swapping islands ($SS_1 \dots SS_S$) replace depleted batteries with fully charged ones.
- DB-Queue & Charging Service: Depleted batteries (DB) enter a queue to be recharged by $C$ chargers ($CS_1 \dots CS_C$).
- FB-Queue: Fully charged batteries (FB) are stored until needed for the next swap.
To analyze this system, the authors employ a two-dimensional Continuous-Time Markov Chain (CTMC) to derive balance equations and find the steady-state distribution.
A key methodological innovation is the asymptotic analysis performed as the number of batteries $B \to \infty$. By analyzing the system's ergodicity in this limit, the authors decouple the complex MQN into two simpler sub-networks. This allows for the derivation of exact analytical lower bounds for the blocking probability, providing a theoretical floor for station performance.
Key Insights and Results
The research identifies two primary operating modes that dictate station performance based on which component acts as the bottleneck:
- Charging-Limiting Mode: This occurs when the chargers are too slow or too few to keep up with demand. In this state, adding more batteries does not improve the blocking probability beyond a specific floor ($1 - C\mu/\lambda$). The authors refer to this as the "Infinite Battery Illusion"—extra batteries simply pile up in the depleted queue waiting for a charger.
- Swapping-Limiting Mode: This occurs when there are enough chargers to keep up, but the swapping robots or islands become the bottleneck. This mode represents the maximum possible performance for a given number of parking spaces and swapping islands.
The authors successfully derived exact mathematical thresholds to determine which mode a station is in. Furthermore, Monte Carlo simulations validated that these theoretical bounds remain highly accurate (typically within 2% error) even when real-world variances, such as non-exponential service times, are introduced.
Conclusion
This work provides a robust framework for multi-stage capacity planning. By understanding the transition between charging-limiting and swapping-limiting modes, operators can avoid over-investing in batteries when chargers are the true bottleneck. While the model assumes infinite physical space for battery storage, it remains a practical tool for station design. Future research may extend this model to include cost-optimization strategies, balancing infrastructure expenses against specific QoS requirements.
Summary updated