基于Q学习算法的发电公司决策新模型

提出了基于Q学习的发电公司决策新模型,应用于求解发电公司在具有不完全信息的电力市场环境下为获取最优长期利润而制定的决策问题.以电力市场重复运行具有的典型Markov过程特性,应用Q学习算法构建以长期利润最优为目标的发电公司决策模型,并通过算例仿真验证了该模型的有效性.所提出的决策新模型可以根据发电公司对市场状态变化的不同预估状态转移概率模拟该公司在市场环境下的不同竞价策略,并给出不确定市场环境下的最优决策.

A Q-Learning Algorithm Based Decision Model for Generation Company

A novel decision model based on Q-learning algorithm for power suppliers who aim at long term profit maximization was proposed.Since a repeatedly operated electricity market can be modeled as a classic Markov decision process problem, Q-learning algorithm can be utilized to build a long-term decision model for power suppliers.The effectiveness of the proposed model was shown by the numerical tests.This model can also be used to simulate different strategies of power suppliers via different types of estimated state transition probability matrix,and to give the optimal decisions finally.