Data-Driven Optimization of Integrated Energy-Transport Networks in Smart Cities: A Multi-Agent Reinforcement Learning Framework

Abstract:

This study proposes a novel multi-agent reinforcement learning (MARL) framework to optimize the integration of electric vehicle (EV) charging infrastructure with renewable energy grids in urban environments. Addressing the critical challenge of imbalanced spatiotemporal demand in smart cities, our approach leverages real-time data from 15,000 IoT sensors across transportation networks, energy grids, and weather systems in Zurich, Singapore, and Tokyo. We develop a decentralized MARL system where agents representing EV charging stations and renewable energy sources learn optimal scheduling and pricing strategies through interactions with their local environments and each other. Integration with blockchain technology facilitates transparent and efficient peer-to-peer energy trading among agents, while spatial equity analytics ensure equitable distribution of charging infrastructure benefits. Comprehensive evaluations through 18-month simulations demonstrates 15% reduction in grid stress during extreme weather events and 23% lower carbon emissions compared to conventional systems. Our findings establish a replicable model for resilient, human-centric urban infrastructure that aligns with SDGs 7 (Affordable Energy), 11 (Sustainable Cities), and 13 (Climate Action).