Electricity auction market simulation with multi-agent model

In electricity market, the generations are in possession of market power, so it is important to estimate the impact of market rule. But it is difficult to analysis this kind of problem by traditional game theory, because of the complex market rules. In this paper an agent-based simulation model is put forward. The model can describe the complicated electricity market rules, and it is more important that suppliers' bidding strategies may be naturally depicted as the behavior seeking for their maximization profits. When the market equilibrium states are obtained, the impacts of the market rules can be evaluated. A modified Roth-Erev reinforcement learning algorithm is developed as the generations agent bidding strategy learning algorithm. The key difference from Roth-Erev reinforcement learning algorithm is the propensity is standardized so as to converge at the Nash equilibria more easily than Roth-Erev algorithm. The impacts of nodal price are discussed. It shows the agent-based simulation call estimate the impact of the complicated electricity market rules.