不同定价方式下电力拍卖市场运行特征的比较研究(一)模型与算法

COMPARISONS AMONG PRICING METHODS IN POOL-BASED ELECTRICITY MARKET BY AGENT-BASED SIMULATION Part One Model and Algorithm

由于电力拍卖市场具有明显的寡头市场特征,且由于发电领域依然存在规模经济性,电力拍卖市场的定价规则出现了多种选择,如边际成本定价、按报价支付定价和当量电价定价法。文中构造了一个基于智能代理的仿真模型讨论3种定价方法的市场运行特征。该模型用具有分散自主决策和智能学习的代理表示发电厂商,在重复进行的报价博弈中,发电厂商通过利润高低按照生物的条件反射原理学习并改进报价策略,追求其利润最大化,随着迭代次数的增加,市场将逐步收敛于均衡位置。然后,通过不同定价方式的拍卖市场均衡状态的比较给出了3种定价方式的市场运行特点。给出了智能代理仿真的模型及其发电厂商的学习算法。最后用一个算例阐述了学习算法的特点。

Because of the existences of market power and economies of scale, various pricing methods are proposed for pool based electricity market, such as uniform clearing pricing method (UCP), pay as bid pricing method (PAB) and the electricity value equivalent pricing method (EVE). In this paper, an agent-based simulation model was developed to compare the market characteristic of the different pricing methods. In the proposed model, the generators improve the bidding strategies by using reinforcement learning algorithm in repeating bidding game to seek for their maximum profit. Then the correlativity about the three pricing methods are discussed, and the generators learning algorithm is given. At last, a simple example is employed to explain the characters of learning in agent-based simulation.