Development of Agent-Based Global Energy Model for Considering International Competition of Energy Strategy

摘要

We introduce our new global energy model formulated as multi-agent simulation with reinforcement learning. Typical global energy models are optimization model and they are suitable to deal with worldwide and long-term energy problem. However, it is very difficult for conventional optimization model to deal with the local problem of conflicting interests with other nations. Therefore, we developed a new model for those problems utilizing the advantages of the conventional one. In new multi-agent based model, the purpose of an agent is to minimize its own energy system cost. It learns the optimal strategy by reinforcement learning to achieve its objective. In this paper, an agent represents a group of nations cooperating to decide energy policies. They can charge energy premium prices on their fuel exports or impose tariff on their fuel import based on their strategy to minimize their own energy cost. We tested this new model with simplified case and confirmed the new multi-agent model works as designed.