Agent-based modeling of the demand-side system reserve provision

Market simulators based on agent-based modeling techniques are frequently used for electricity market analyses. However, the majority of such analyses focus on the electricity markets bidding strategies on generation-side rather than on the demand-side. Meanwhile, the behavior of the demand-side in the system reserve provision has been less investigated. This paper presents a novel system reserve provision agent which is incorporated into a stochastic market optimization problem. The agent for the system reserve provision uses the SA-Q-learning algorithm to learn how much system reserve to offer at different times, while seeking to increase the ratio between their economic costs and benefits. The agent and its learning process are described in detail and are tested on the IEEE Reliability test system. It has been shown that incorporating the demand-side market strategies using the proposed agent improves the performance and the economic outcome for the consumers.