Dynamic Incentive Strategies for Smart EV Charging Stations: A LLM-Driven User Digital Twin Approach

Electric vehicles and vehicle-to-grid technology are pivotal to modern demand response systems, yet their effectiveness is hindered by uncertainties in user behavior and low participation. To address these challenges, this paper presents a collaborative multi-agent demand response framework enhanced by large language models. By constructing user digital twins that integrate multidimensional user profile features, user decision-making patterns can be accurately predicted. Furthermore, a data and knowledge dual-driven dynamic incentive mechanism is introduced, combined with a network-constrained distributed optimization model, to optimize grid-user interactions while ensuring economic efficiency and security. Simulation results indicate significant improvements in peak load mitigation and charge-discharge strategies. Experimental validation highlights the system’s advantages in load balancing, user satisfaction, and grid stability, providing policymakers with a scalable V2G management tool that fosters sustainable vehicle-grid synergy.