Study of power market simulation based on distributed multi-agent system

Commonly used power market simulation systems can only simulate the small-scale market with a simple result and do not consider the market participators as those who have the adaptive learning mechanism. By modeling the market participators as Agents, we approached a MAS-based distributed simulation system on JADE platform, which has the ability to simulate complex, large-scale markets with different strategies on different Agents. The Q-Learning and some other methods are introduced into the bidding strategy of the system. The system performance and convergence of Q-Learning strategy are evaluated by designed instances. As a result, the distributed system performance is better than the centralized system in many respects and Q-Learning strategy always leading to a high profit.