有限信息环境下基于学习自动机的发电商竞价策略

电力市场是典型的不完全竞争市场,发电商可以通过策略性报价以提高自身收益。现有的发电商报价策略研究通常假设发电商能利用充分的市场信息,但这种假设在市场启动初期往往不成立。为解决发电商在有限信息环境下的报价策略问题,文中提出了一种改进的强化学习自动机算法,该方法对外部信息量要求较低,且计算复杂度小,易于实现。此外,将发电商报价和市场出清的过程建模为重复博弈而非广泛使用的马尔可夫博弈,避免了马尔可夫博弈要求系统状态具有时间相关性这一强假设。最后,算例验证了该算法的有效性。

Learning Automata Based Bidding Strategy for Power Suppliers in Incomplete Information Environment

The electricity market is a typical imperfectly competitive market where power suppliers can increase their profits through strategic bidding. Existing research on bidding strategies for power suppliers usually assumes that the power suppliers can use sufficient market information, which is often not true if the market is just launched. To solve the bidding problem of power suppliers in the incomplete information environment, this paper proposes an improved reinforcement learning automata algorithm. This method requires little external information, and is easy to implement. In addition, this paper models the process of power supplier bidding and market clearing as a repeated game rather than the widely used Markov game, which avoids the strong assumption that Markov game requires time correlation between system states. Finally, the example result verifies the effectiveness of the proposed algorithm.