Multi-Charging Station Pricing Strategy for Real-Time Congestion Management in Coupled Traffic-Power Networks

In high-density cities, coordinating charging station (CS) real-time pricing to mitigate traffic congestion is considered a promising approach. However, the increasing complexity of transportation systems and multi-player dynamic interactions limit the effectiveness of existing pricing strategies in real-time congestion management. To address these, a real-time coordinated pricing framework for multiple CSs in coupled traffic-power networks based on Simulation of Urban MObility (SUMO) simulation is proposed in this paper. First, based on the established CS selection model for EVs, SUMO is utilized to simulate the real-time vehicle and road conditions of the transportation system under given charging prices, providing feedback on charging demand as well as individual and global traffic congestion costs. Next, an Adversarial Diverse Ensemble (ADE) surrogate-based traffic state prediction model is developed to overcome the time-intensive nature of SUMO and integrate it into the multi-agent reinforcement learning (MARL) framework. Then, a Graph Attention Network-based bi-level multi-agent Soft Actor-Critic (GAT-BLMASAC) algorithm is proposed to solve the coordinated pricing problem in a bi-level architecture, where the upper level optimizes bidding strategies of generation companies (GenCos) as price-makers and the lower level coordinates pricing among CSs as price-takers. Finally, simulations on a real-world Hong Kong transport network demonstrate that the pricing strategies derived from the proposed framework significantly mitigate traffic congestion while enhancing the total profit of CSs. The proposed algorithm exhibits robust convergence in the multi-leader-multi-follower game, offering a scalable solution for coordinated CS pricing in high-density urban settings.