基于多主体的发电企业二阶段博弈模型仿真

Simulation for Two-phase Game Model of Power Plants Based on Multi-agent

不完全信息、垄断电力市场下,发电企业在博弈中需要准确预测对手信息,因此提出一种基于多主体平台(Swarm)的2阶段演化博弈模型。该模型依据获取信息量的多少把博弈过程划分为2阶段,刻画了发电企业充分利用在博弈过程中所获取的信息,不断修正自己行为规则的演化过程。为了研究个体决策对群体行为的影响,在模型中考虑了发电企业群体中成员的自学行为、效仿行为以及群体之间交互行为。模型假定发电市场是由采用不同模型的发电企业构成,在多主体平台Swarm上对该发电市场进行仿真,通过仿真论证了该模型能较其它模型更准确地预测对手的信息。在发电市场中引入退出机制,对由具有效仿行为的发电企业构成的发电市场进行仿真,发现留存在市场上的发电企业通常能较准确预测竞争对手的信息。

In incomplete information and monopolistic electricity market, power plant needs to predict competitors＇ information accurately. A two-phase game model based on multi-agent plat （Swarm） was proposed. The game process was divided into two phases according to the amount of information they received. The model depicted that power plant made full use of the obtained information and continually modified its behavior and rule in game. In order to study the effect of individual decision-making on population, both the self-educated behavior and imitative behavior of the member in power market, also alternating behavior between different populations were taken into account in the model. The power market, which consists of power plants adopting different models, has been simulated by Swarm. The result showed the two-phase game model can predict the information of their rivals more accurately than others. Power market with quit mechanism has been simulated. In power market, these power plants are able to imitate others＇ decision models. The simulation demonstrated that the survived power suppliers can accurately predict the information of their opponent.