Integrating Demand Response into agent-based models of electricity markets

This paper introduces a model for the decision-making of a Demand Response (DR) program operator as an adaptive agent participating in a competitive electricity market. The model focuses on an electricity retailer stimulating DR as a means of avoiding high balancing market prices. Nevertheless, we demonstrate the ability to extend the model to other actors who could capitalize on the value of demand flexibility - e.g. wind power producers looking to offset output variability. The model considers voluntary demand modifications whose materialization, subject to the uncertainty of consumer behavior, results in redistribution of consumption over a short time frame. As the retailer is modeled via an adaptive agent, it has the potential to learn from both consumer behavior and market outcomes. Here, we implement a reinforcement learning approach with the objective of allowing the agent to increase its profit by identifying the conditions under which DR should be stimulated. We validate the proposed agent-based model as a tool to quantify DR potential in a market setting.