Agent-based Simulation of Retailers under Competitive Environment using Classifier System

This paper presents a novel simulation system to analyze adaptive behaviors of agents in a deregulated electricity retail market. We develop a learning framework which enables the agents to autonomously acquire the action rules. In this paper, the XCS (extended classifier system) is employed as a learning algorithm of the agents. XCS can efficiently generate action rules in the dynamic environment such as the deregulated retail market affected by the interaction among many agents. The artificial retail market shows complicated behavior by the interaction among the agents, and therefore the agent-based simulation can provide some technical findings due to the interactions among autonomous agents which are not always rational in a sense of optimal behaviors. We provide new insights based on the behavior analysis of the agents in the artificial retail market simulation.