TL;DR
This paper models electric vehicle (EV) charging as a queueing system to understand congestion in parking lots and the power grid. They develop analytical bounds and approximations for the probability of an EV getting fully charged, offering insights for infrastructure planning and resource allocation.
What Problem Does This Paper Address?
The increasing adoption of electric vehicles (EVs) presents a dual challenge: potential congestion at charging stations and strain on the electricity distribution grid. As more EVs need charging, limited parking spaces and available power can lead to inefficiencies and performance issues. This paper aims to provide a framework for analyzing these problems, particularly from an infrastructure investment and policy perspective, by treating EV charging as a queueing system. They highlight the need for models that operate on longer, "investment" time-scales, beyond the typical operational analyses.
Methods
The authors employ queueing theory to model the EV charging process. Their approach involves:
- Queueing Model: They define a system with a finite number of parking spaces ($K$) and charging stations, where EVs arrive randomly, have random parking durations, and random energy demands. Crucially, EVs can leave without being fully charged, and arriving EVs are blocked if all spaces are occupied.
- Bounds: They derive upper and lower bounds for the fraction of EVs that get fully charged in stationarity. These bounds are based on analyzing two special cases: when the number of charging stations equals the number of parking spaces ($K=M$) and when there are infinitely many spaces ($K=\infty$).
- Fluid Approximation: They develop a fluid approximation for the system, which simplifies the complex queueing dynamics into a continuous model. This approximation helps in understanding the system's behavior under heavy traffic conditions and provides a closed-form expression for the success probability (fraction of fully charged EVs).
- Diffusion Approximation: For a more refined analysis, they introduce a diffusion approximation. This technique models the system's fluctuations around its fluid limit, providing a stochastic differential equation that describes the system's evolution. This allows for a deeper understanding of the system's dynamics and potential for transient behavior.
Datasets
Not discussed in the paper. The paper focuses on theoretical modeling and numerical evaluation of the proposed models.
Key Insights and Ideas
- Dual Congestion: The paper clearly articulates that EV charging faces two primary sources of congestion: the limited number of parking spaces with chargers and the finite available power.
- Investment-Scale Modeling: The emphasis on developing models for longer, investment-oriented time-scales is crucial for strategic planning of charging infrastructure.
- Queueing as a Natural Fit: Treating EV charging as a queueing system, with arrivals, service (parking and charging), and potential blocking, provides a robust framework for performance analysis.
- Success Probability as a Key Metric: The probability of an EV leaving with a fully charged battery is identified as a critical performance indicator, directly impacting user satisfaction and the overall utility of the charging infrastructure.
- Approximations Provide Tractability: While exact solutions can be complex, the fluid and diffusion approximations offer valuable insights into system behavior, especially under heavy load conditions.
Key Results
- Bounds on Success Probability: The paper provides analytical upper and lower bounds for the fraction of EVs that get fully charged. These bounds are shown to be tight for certain parameter ranges (e.g., $M > 0.7K$).
- Fluid Approximation Formula: A fixed-point equation is derived for the steady-state fluid limit of the uncharged EV population, and an explicit formula for the success probability is presented for different load conditions.
- Diffusion Limit Equation: A 2-dimensional stochastic differential equation is established for the diffusion limit, describing the system's fluctuations. This equation is related to known results for Erlang B systems.
- Numerical Validation: Figures 1-3 illustrate the tightness of the bounds and the accuracy of the fluid approximation for various scenarios (moderately, critically, and over-loaded systems). The fluid approximation is noted to work well for $K=50$.
Limitations and Future Work
- Model Assumptions: The authors acknowledge that their model relies on several simplifying assumptions, including exponential arrival, parking, and charging times, and mutually independent random variables.
- Markovian Assumptions: The analysis is primarily based on Markovian assumptions, which might not always hold in real-world scenarios.
- Extension to General Distributions: The paper mentions that extending the model to handle generally distributed random variables for parking and charging times is a direction for future work.
- Time-Varying Rates and Multiple Classes: The authors plan to extend their analysis to time-varying arrival rates, multiple customer classes, and multiple parking lots.
- Distribution Network Analysis: The paper notes that the distribution network for multiple parking lots can be modeled as a tree network, suggesting further research in this area.
- Invariant Distribution for Finite K: The authors suggest that deriving the invariant distribution for the diffusion process with finite $K$ could lead to further improvements, as the $K=\infty$ case did not yield better results than the Erlang A lower bound.
Summary updated