電力自由化市場エージェントモデルにおけるエージェントレベルでの学習能力評価

概要

論文の詳細を見る
ABM (Agent Based Modeling/Model) becomes a new approach in simulation technology. ABM is composed by a set of autonomous agents and the validity of ABM can be estimated by evaluating the characteristics not only at system level but at agent level. Since any agent of ABM behaves adaptively, the validity must be proved by showing a fact that the agent's learning proceeds correctly in the simulation. In this paper we propose a metric of quantifying learning process and a method which evaluates agent's learning by the metric. We built an ABM simulator of a competitive market in electric power and the method was applied to the simulation results. Then we derived the conclusion that the ambiguity in rule decision decreases, agents learn rules correctly and the metric works effectively.
日本シミュレーション学会の論文
2001-03-15