Page 98 - 《中国电力》2026年第3期
P. 98
第 59 卷 第 3 期 Vol. 59, No. 3
2026 年 3 月 ELECTRIC POWER Mar. 2026
引用格式:郑峰, 孙电, 黄丽丽, 等. 基于混合博弈强化学习的虚拟电厂市场交易策略[J]. 中国电力, 2026, 59(3): 94−102.
Citation: ZHENG Feng, SUN Dian, HUANG Lili, et al. Virtual power plant market trading strategy based on hybrid game reinforcement learning[J].
Electric Power, 2026, 59(3): 94−102.
基于混合博弈强化学习的虚拟电厂市场交易策略
郑峰 ,孙电 ,黄丽丽 ,杨峰 ,倪芸 2
2
1
2
2
(1. 国家能源集团长源电力股份有限公司,湖北 武汉 430000;
2. 国能长源能源销售有限公司,湖北 武汉 430000)
Virtual power plant market trading strategy based on hybrid game
reinforcement learning
2
2
2
1
ZHENG Feng , SUN Dian , HUANG Lili , YANG Feng , NI Yun 2
(1. Chn Energy Changyuan Electric Power Co., Ltd., Wuhan 430000, China;
2. Guoneng Changyuan Energy Sales Co., Ltd., Wuhan 430000, China)
Abstract: With the rapid development of regional distributed higher returns.
energy, the issues of small installed capacity and strong output This work is supported by Science and Technology Project of
variability have become increasingly prominent, resulting in Chn Energy Changyuan Electric Power Co., Ltd. (No.CYDL-
insufficient competitiveness when distributed energy 2024-14).
participates in market transactions independently. To enhance Keywords: virtual power plant operators; market trading;
its market participation capabilities, integrating distributed hybrid game
energy resources into virtual power plant has emerged as an
摘 要:随着地区分布式能源快速发展,其单机装机容
effective approach. Therefore, this study investigates market
量小和出力随机性强的问题愈发凸显,导致分布式能源
trading strategies for virtual power plant incorporating
在单独参与市场交易时竞争力不足。为提升其市场参与
distributed energy resources and proposes a trading strategy
能力,整合分布式能源形成虚拟电厂(virtual power plant,
based on hybrid game-based reinforcement learning. First,
VPP)已成为一种有效途径。因此,针对含分布式能源的
establish revenue models for energy suppliers and load
VPP 市场交易策略进行研究,提出一种基于混合博弈强
aggregators based on the operational characteristics of internal
化学习的交易策略。首先,根据虚拟电厂内部单元的运
units within the virtual power plant. Then, to ensure the overall
行 特 性 构 建 能 源 供 应 商 和 负 荷 聚 合 商 的 收 益 模 型 ; 然
profitability of operators within the virtual power plant, a social
后,为了保证虚拟电厂内部运营商的整体收益建立社会
welfare maximization model is established. Finally, the
福利最大化模型;最后,基于 Stackelberg 博弈和演化博弈
transaction model is solved using a hybrid game-based 的混合博弈强化学习算法求解该交易模型。算例分析表
reinforcement learning algorithm combining Stackelberg and 明,基于混合博弈强化学习算法的双层模型求解效果优
evolutionary game theory. Case studies demonstrate that the 于其他传统智能算法,求解时间减小近 50%;此外,VPP
two-layer model based on hybrid game-theoretic reinforcement 同 时 参 与 能 量 市 场 和 辅 助 服 务 市 场 时 , 可 获 得 更 高 的
learning algorithms outperforms traditional intelligent 收益。
algorithms, reducing computation time by nearly 50%. 关键词:虚拟电厂运营商;市场交易;混合博弈
Furthermore, when virtual power plants participate in both DOI:10.11930/j.issn.1004-9649.202506047
energy markets and ancillary service markets, they can achieve
0 引言
收稿日期:2025−06−17; 修回日期:2026−01−05。
基金项目:国家能源集团长源电力股份有限公司科技项目 “双碳”目标背景下,以风电和光伏为代表
(CYDL-2024-14)。 的分布式能源快速发展 [1-5] 。然而,分布式能源的
94

