Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decentralized Collaborative Pricing and Shunting for Multiple EV Charging Stations Based on Multi-Agent Reinforcement Learning

Tianhao Bu, Hang Li, Guojie Li

TL;DR

This work extends decentralized EV charging optimization to a multi-charging-station setting where each charging pile is an autonomous agent. By formulating the problem as an MDP and applying centralized training with decentralized execution using MARL (VDN, with Q-Mix as an alternative), the authors develop a price-scheduling strategy to induce EV shunting across stations. A probabilistic EV user preference model, coupled with distance normalization, guides CS attractiveness and user choice, while a real-time pricing mechanism balances occupancy and profit. Simulations demonstrate improved convergence and higher accumulated rewards under pricing strategies, suggesting practical potential for reducing peak demand and improving station utilization in multi-station networks.

Abstract

The extraordinary electric vehicle (EV) popularization in the recent years has facilitated research studies in alleviating EV energy charging demand. Previous studies primarily focused on the optimizations over charging stations (CS) profit and EV users cost savings through charge/discharge scheduling events. In this work, the random behaviors of EVs are considered, with EV users preferences over multi-CS characteristics modelled to imitate the potential CS selection disequilibrium. A price scheduling strategy under decentralized collaborative framework is proposed to achieve EV shunting in a multi-CS environment, while minimizing the charging cost through multi agent reinforcement learning. The proposed problem is formulated as a Markov Decision Process (MDP) with uncertain transition probability.

Decentralized Collaborative Pricing and Shunting for Multiple EV Charging Stations Based on Multi-Agent Reinforcement Learning

TL;DR

This work extends decentralized EV charging optimization to a multi-charging-station setting where each charging pile is an autonomous agent. By formulating the problem as an MDP and applying centralized training with decentralized execution using MARL (VDN, with Q-Mix as an alternative), the authors develop a price-scheduling strategy to induce EV shunting across stations. A probabilistic EV user preference model, coupled with distance normalization, guides CS attractiveness and user choice, while a real-time pricing mechanism balances occupancy and profit. Simulations demonstrate improved convergence and higher accumulated rewards under pricing strategies, suggesting practical potential for reducing peak demand and improving station utilization in multi-station networks.

Abstract

The extraordinary electric vehicle (EV) popularization in the recent years has facilitated research studies in alleviating EV energy charging demand. Previous studies primarily focused on the optimizations over charging stations (CS) profit and EV users cost savings through charge/discharge scheduling events. In this work, the random behaviors of EVs are considered, with EV users preferences over multi-CS characteristics modelled to imitate the potential CS selection disequilibrium. A price scheduling strategy under decentralized collaborative framework is proposed to achieve EV shunting in a multi-CS environment, while minimizing the charging cost through multi agent reinforcement learning. The proposed problem is formulated as a Markov Decision Process (MDP) with uncertain transition probability.
Paper Structure (21 sections, 28 equations, 6 figures, 4 tables, 1 algorithm)

This paper contains 21 sections, 28 equations, 6 figures, 4 tables, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. BACKGROUND
  3. Decentralized collaborative framework (single CS)
  4. ENVIRONMENT FORMULATION
  5. Multi(dual)-CS based collaborative framework
  6. Probabilistic EV user preference model
  7. Distance normalization model
  8. Markov decision process
  9. State
  10. Action
  11. State transition
  12. Reward function
  13. PROPOSED PRICE SCHEDULING STRATEGY
  14. Price scheduling
  15. Training phase
  16. ...and 6 more sections

Figures (6)

  • Figure 1: Decentralized collaborative framework for a single CS.
  • Figure 2: Multi(dual)-CS based decentralized collaborative framework
  • Figure 3: Linear EV-CS distance model
  • Figure 4: EV distributions over price scheduling and baseline methods, under each scenario
  • Figure 5: VDN accumulation reward distribution
  • ...and 1 more figures