AI for Demand Flexibility in Electricity Markets

Cong Chen – Dartmouth College

The rapid growth of distributed energy resources—such as home batteries, rooftop solar panels, and electric vehicles—has transformed how households participate in electricity markets. As demand-side technologies become more widespread, understanding how customers make decisions under uncertainty is increasingly important for improving system efficiency and resilience. However, modeling real-world customer behavior remains challenging. This talk presents two AI-based approaches to better capture demand flexibility. First, we develop a scalable mean-field learning framework to analyze population-level decisions of large numbers of heterogeneous households. This method enables efficient modeling of aggregate demand-side behavior while accounting for individual uncertainty. Second, we introduce large language model-powered AI agents that act as digital proxies for energy customers. These agents simulate credible and diverse behavioral responses to dynamic electricity prices, policy incentives, and rare but high-impact events such as blackouts. Together, these approaches provide practical tools for generating credible behavioral insights, supporting improved forecasting, market design, and resilience planning in modern electricity markets.