Agent-based modeling of the demand-side flexibility

The paper presents an agent-based approach to model the flexibility of the demand-side. It uses Q-learning algorithm to model a behavior of a demand-side agent, so to investigate the elasticity of the demand to the change in price. Often, market simulation models assume that the demand elasticity is known, however due the lack of data this elasticity is not easy to determine. The objective of this paper is to evaluate the flexibility of the total system demand, and the shift in the consumption with the price, i.e. increase in the demand when the price is low, and a decrease in the demand when the price is high. The here presented model of a demand-side agent is incorporated into the market simulator with double-sided auctions, and is tested on the Slovenian market. However, this approach can be used to estimate flexibility in any system for which the forecasted demand data and generation offers are know.