Collaborative Research: CISE MSI: RDP: CPS: Enhancing Efficiency and Sustainability in Electric Transportation and Power Systems through Interaction-Aware Management

The rapid evolution in the power grid is driven by sustainability concerns, with focus shifting significantly towards renewable energy sources and away from fossil fuels. Meanwhile, the rise of electric vehicles (EVs) has accelerated the electrification of transportation. Accordingly, this project proposes an interaction-aware management framework to improve the efficiency and sustainability of these two independent and self-interested systems. Since both the EV transportation system and the power-grid system are managed by distinct stakeholders and operate within different domains, they function independently and without coordination, impacting their overall efficiency and causing issues such as voltage instability, frequency fluctuations, high financial costs, and long charging durations. Past work has accumulated abundant knowledge on how to design each system independently; however, strategies to achieve synergistic outcomes beneficial to both parties remain under-explored. It is therefore crucial to develop a collaborative framework that considers how each system responds to the other's actions, such as how power grids adjust electricity prices based on EV charging demands and how EVs choose charging stations based on price and availability. This collaboration is expected to benefit both EV drivers and power-grid operators, reducing costs and improving sustainability. In terms of broader impact, the project also includes capacity-building, education, and outreach initiatives to promote the participation of underrepresented minorities in the modern EV-related workforce. Technically, the project will focus on the following three key components: (1) a robust multi-agent reinforcement-learning control model for power systems to dynamically adjust electricity prices and charging-power rates, which integrates a human charging-behavior model for enhanced accuracy and efficiency; (2) a mean-field game-based control method for large-scale EVs to autonomously select charging stations in a decentralized fashion with awareness of potential charging rates; and (3) an incentive-driven collaboration mechanism to facilitate socially optimal actions between the power grid and EV operators using graph-based multi-agent reinforcement learning and Shapley value. The project will use real-world data to validate its approach and ultimately contribute to the development of a more sustainable transportation and power infrastructure. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.