Toward sustainable urban mobility: Advanced power-sharing strategies for electric vehicle charging

The rapid growth of Electric Vehicle (EV) adoption poses significant challenges to electricity demand management, with risks of transformer overloads and demand spikes in urban power grids. To address these challenges and enhance grid sustainability, this study introduces a power-sharing strategy for public fast-charging stations. This strategy enables public charging stations to dynamically distribute available electrical power among multiple EVs that are plugged in simultaneously. Effective coordination of charging can minimize the charging costs and prevent grid overloads, addressing the price and time sensitivity of EV drivers. To achieve this coordination, we formulate the EV charging problem as a multi-agent Markov decision process and develop a unified multi-objective control framework based on Multi-agent reinforcement Learning. Specifically, we extend the Multi-Agent Cooperative Soft Actor-Critic architecture by embedding a power-sharing mechanism and a set of jointly optimized objectives, including charging waiting time, charging cost, and power-grid load balancing. This integrated algorithmic design provides the foundation that links the problem formulation to system-level performance improvements. Using real-world data from Shenzhen, China, the proposed framework outperforms the baseline shortest distance (SD) algorithm. Results demonstrate that grid-aware algorithms achieve a more balanced distribution of visits, reducing power pressure at high-demand stations and improving utilization at less frequented ones. Notably, the proposed MARL algorithm with power-sharing achieves the lowest average station load, representing a 7.9% reduction compared to the SD algorithm. This work contributes to the design of smart, clean, and efficient urban transportation systems, aligning with the vision of sustainable cities by promoting intelligent infrastructure for EV charging.