Mitigating Urban Infrastructure Scarcity: A Geo-Aware Closed-Loop Reinforcement Learning System for Dynamic and Socially Equitable Electric Vehicle Charging and Power Grid Balancing

• Abey city

  Lautech

  Author

The rapid proliferation of electric vehicles (EVs) in urban environments has created unprecedented strain on power distribution networks and exposed significant spatial inequities in charging infrastructure accessibility. While recent advances in reinforcement learning and multi-agent systems have demonstrated promise for EV scheduling and grid management, existing approaches treat infrastructure planning and real-time operational control as separate problems, resulting in suboptimal resource allocation and persistent social disparities. This study proposes a geo-aware closed-loop reinforcement learning framework that integrates federated deep reinforcement learning with geospatial analytics to jointly optimize EV charging station placement and real-time power grid balancing. The framework employs a hierarchical multi-agent architecture where local agents independently learn adaptive charging policies using Proximal Policy Optimization (PPO) while a central aggregator synchronizes global model parameters through federated averaging with fairness-aware weighting. Using real-world data from Chicago's urban landscape comprising 1,200 simulated EVs across 60 charging stations and a 33-bus feeder system, the proposed system achieves an 13.6% reduction in grid operating cost, a 21.4% increase in renewable energy absorption, and maintains Jain's fairness index consistently above 0.95 across income-diverse communities. The framework demonstrates robust adaptation to dynamic demand patterns and successfully allocates 520 charging stations that exceed a high reward threshold while improving population-weighted accessibility across all socioeconomic groups. This research contributes a replicable, privacy-preserving, and socially equitable approach to sustainable urban infrastructure planning.

• pdf

Copyright (c) 2026 Abey city (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.