Online Dynamic Pricing for Electric Vehicle Charging Stations with Reservations

Jan Mrkos, Antonín Komenda, David Fiedler, Jiří Vokřínek

Abstract

This paper introduces a novel model for online dynamic pricing of electric vehicle (EV) charging services that integrates reservation, parking, and charging into a comprehensive bundle priced as a whole. Our approach focuses on the individual high-demand, fast-charging location, employing a Poisson process as a model of charging reservation arrivals, and develops an online dynamic pricing strategy optimized through a Markov Decision Process (MDP). A key contribution is the novel analysis of discretization error introduced when incorporating the continuous-time Poisson process into the discrete MDP framework. The MDP model's feasibility is demonstrated with a heuristic dynamic pricing method based on Monte-Carlo tree search (MCTS), offering a viable path for real-world applications.

Introduction

As electric vehicles (EVs) become increasingly prevalent, highway charging hubs are facing significant congestion. Current pricing models often present a confusing array of connection fees, per-minute idle charges, and per-kWh energy costs. Furthermore, the standard first-come, first-served approach creates travel uncertainty for long-distance drivers. This research, conducted at the Czech Technical University in Prague, addresses these challenges by applying Revenue Management principles to EV charging, treating station capacity as a finite resource similar to airline seats or hotel rooms.

Positioning the Research

To understand where this work fits within the broader landscape of EV charging research, it is helpful to categorize pricing strategies. While many surveys look at centralized or global-view systems, this paper focuses on a decentralized, local-view approach. By optimizing for the individual charging station (CS), the model provides a practical tool for operators to manage local demand and maximize revenue through sophisticated pricing mechanisms.

Taxonomy of EV charging pricing research, highlighting this work's focus on a decentralized, local view.

Methodology

The proposed solution rests on three primary pillars:

  1. The Bundle Model: Instead of fragmented fees, the service is sold as a single "product" encompassing the reservation, the parking spot, and the electricity. This allows the operator to internalize the opportunity cost of a slot; as capacity decreases, the price for long sessions rises to reflect the potential loss of other customers.
  2. Markov Decision Process (MDP): The problem is framed as a sequence of decisions where the "state" includes available capacity, time, and the specific request, while the "action" is the price offered.
  3. Monte Carlo Tree Search (MCTS): To overcome the scalability issues of traditional methods like Value Iteration (VI), the authors employ MCTS (specifically the UCT variant). This allows the system to "look ahead" into future scenarios and scale to real-world station sizes efficiently.

Technical Contributions: Discretization Error Analysis

A standout technical contribution of this work is the analysis of discretization error. Charging requests occur in continuous time (modeled as a Poisson process), but the MDP operates in discrete time intervals. The authors developed a novel metric to quantify the error introduced by this approximation, demonstrating how to select a time-step length that maintains high accuracy (e.g., a relative error of 0.06) without requiring excessive computational power.

Datasets and Evaluation

The study utilized a real-world dataset from a German charging station operator, containing 1,513 charging sessions over two years. This data informed the distributions for charging duration and start times. To stress-test the model, the researchers generated high-demand synthetic scenarios to evaluate the dynamic pricing and scalability of the MCTS algorithm.

Key Results

  • Revenue Optimization: The MCTS approach significantly outperformed flat-rate pricing baselines in high-demand scenarios.
  • Near-Optimality: In small-scale tests, MCTS achieved revenue results nearly identical to the exact mathematical solutions provided by Value Iteration.
  • Real-Time Viability: The MCTS algorithm generated price offers in seconds, making it suitable for integration into smartphone apps or vehicle head units.
  • Demand Shifting: The model successfully shifted demand, ensuring more efficient utilization of charging slots during peak hours compared to static pricing.

Limitations and Future Work

While the model provides a robust framework, the authors acknowledge several areas for future refinement. Currently, the model does not account for "no-shows" or strategic driver behavior (such as booking multiple slots). Additionally, it assumes a constant power flow rather than the variable "charging curves" typical of real EVs. Future research will involve testing the model in multi-agent simulations to observe the interactions of a city-wide network of smart drivers reacting to dynamic prices.

Summary updated

Abstract

This paper introduces a novel model for online dynamic pricing of electric vehicle charging services that integrates reservation, parking, and charging into a comprehensive bundle priced as a whole. Our approach focuses on the individual high-demand, fast-charging location, employing a Poisson process as a model of charging reservation arrivals, and develops an online dynamic pricing strategy optimized through a Markov Decision Process (MDP). A key contribution is the novel analysis of discretization error introduced when incorporating the continuous-time Poisson process into the discrete MDP framework. The MDP model's feasibility is demonstrated with a heuristic dynamic pricing method based on Monte-Carlo tree search, offering a viable path for real-world applications.
BibTeX
@article{mrkos2025online,
  author = {Jan Mrkos and Antonín Komenda and David Fiedler and Jiří Vokřínek},
  title = {{Online Dynamic Pricing for Electric Vehicle Charging Stations with Reservations}},
  journal = {IEEE Trans. ITS},
  year = {2025},
  doi = {10.1109/tits.2025.3565906},
}
Figure