DR-Gym: Open-Source Environment for Utility Demand-Response Training

Researchers have launched DR-Gym, an open-source environment compatible with Gymnasium, aimed at training and assessing demand-response initiatives from the viewpoint of electric utilities. Extreme weather conditions and fluctuating wholesale electricity markets pose significant financial risks to residential consumers, yet demand response at the distribution level is still largely untapped. Although these programs can provide financial credits during peak pricing periods to protect consumers, optimizing the sequential decision-making process is a complex task for reinforcement learning. While there is an abundance of historical smart meter and wholesale pricing data available, it does not adequately reflect the dynamic interaction between utility pricing signals and customer responses. DR-Gym tackles this issue by creating interactive scenarios for training reinforcement learning agents. This research is documented in arXiv:2605.12462v1.