Generation Expansion Planning Considering Investment Dynamic of Market Participants Using Multi-agent System

Abstract

This paper presents a model for generation expansion planning (GEP) using multi-agent system. A robust framework for modeling communications between market participants is proposed. Multi-agent simulation provides a capability for simulating interaction and communication among the market participants. The developed approach considers counteraction and influences of decisions made by generation companies (GenCos) in the level of investments, so GenCos investments in dynamic environment will be modeled. Reinforcement learning has been applied in order to smart out the agents and modeling the intellectual decision makers. In addition, fuel cost and load growth uncertainties have been considered. In this model, GenCos decisions are converge to the Nash equilibrium point. The environment of investment for each GenCo is defined in a way that it observes Markov property. So, the problem can be considered as an episodic decision making over years of planning. Using this approach, in a case study three incentive capacity mechanisms have been appraised.