Partial cooperation models are studied for many years to solve the bilevel programming problems where the follower’s optimal reaction is not unique. However, in these existed models, the follower’s cooperation level...Partial cooperation models are studied for many years to solve the bilevel programming problems where the follower’s optimal reaction is not unique. However, in these existed models, the follower’s cooperation level does not depend on the leader’s decision. A new model is proposed to solve this deficiency. It is proved the feasibility of the new model when the reaction set of the lower level is lower semicontinuous. And the numerical results show that the new model has optimal solutions when the reaction set of the lower level is discrete, lower semi-continuous and non-lower semi-continuous.展开更多
为研究中长期市场交易对现货市场运行的影响,分析发电商在现货市场中的报价策略,提出一种模拟电力现货市场发电商竞价均衡的双层优化模型和多智能体深度强化学习(multi-agent deep reinforcement learning,MADRL)求解算法。引入供需比...为研究中长期市场交易对现货市场运行的影响,分析发电商在现货市场中的报价策略,提出一种模拟电力现货市场发电商竞价均衡的双层优化模型和多智能体深度强化学习(multi-agent deep reinforcement learning,MADRL)求解算法。引入供需比表征现货市场供需关系,并使用前景理论刻画发电商的有限理性行为特征,以此分析中长期市场交易对现货市场中发电商报价策略的影响。在MADRL求解过程中,将发电商建模为智能体,现货市场出清建模为环境,通过迭代求解得到均衡状态下各发电商的报价策略和现货市场出清价格。以中国东部区域包含8家发电商的实际电力系统为例开展仿真,结果表明,该MADRL算法可以有效求解各发电商的报价策略,准确模拟不同中长期市场设置对现货市场运行的影响。研究结论可为电力交易机构评估发电商竞价行为和制定市场规则提供参考依据。展开更多
基金supported by the National Natural Science Foundationof China (70771080)the National Science Foundation of Hubei Province(20091107)Hubei Province Key Laboratory of Systems Science in Metallurgical Process (B201003)
文摘Partial cooperation models are studied for many years to solve the bilevel programming problems where the follower’s optimal reaction is not unique. However, in these existed models, the follower’s cooperation level does not depend on the leader’s decision. A new model is proposed to solve this deficiency. It is proved the feasibility of the new model when the reaction set of the lower level is lower semicontinuous. And the numerical results show that the new model has optimal solutions when the reaction set of the lower level is discrete, lower semi-continuous and non-lower semi-continuous.
文摘为研究中长期市场交易对现货市场运行的影响,分析发电商在现货市场中的报价策略,提出一种模拟电力现货市场发电商竞价均衡的双层优化模型和多智能体深度强化学习(multi-agent deep reinforcement learning,MADRL)求解算法。引入供需比表征现货市场供需关系,并使用前景理论刻画发电商的有限理性行为特征,以此分析中长期市场交易对现货市场中发电商报价策略的影响。在MADRL求解过程中,将发电商建模为智能体,现货市场出清建模为环境,通过迭代求解得到均衡状态下各发电商的报价策略和现货市场出清价格。以中国东部区域包含8家发电商的实际电力系统为例开展仿真,结果表明,该MADRL算法可以有效求解各发电商的报价策略,准确模拟不同中长期市场设置对现货市场运行的影响。研究结论可为电力交易机构评估发电商竞价行为和制定市场规则提供参考依据。
