中国电力 ›› 2026, Vol. 59 ›› Issue (1): 153-162.DOI: 10.11930/j.issn.1004-9649.202409028

• 新型电网 • 上一篇    

基于多主体博弈的电力市场不平衡资金分摊优化方法

王奖1(), 陈晓东1(), 许喆1(), 王景亮1, 王立鹏2   

  1. 1. 广州电力交易中心有限责任公司,广东 广州 510000
    2. 北京清大科越股份有限公司,北京 100102
  • 收稿日期:2025-09-12 修回日期:2025-12-19 发布日期:2026-01-13 出版日期:2026-01-28
  • 作者简介:
    王奖(1996),男,硕士,从事电力市场研究,E-mail:609057688@qq.com
    陈晓东(1975),男,博士,副教授,通信作者,从事电力市场研究,E-mail:chenxd2@csg.cn
    许喆(1989),女,硕士,副教授,从事电力市场研究,E-mail:xuzhe@csg.cn
  • 基金资助:
    广州电力交易中心科技项目(180000KC23080001)。

An optimization method for imbalance funds in a multi-agent game-theoretic electricity market

WANG Jiang1(), CHEN Xiaodong1(), XU Zhe1(), WANG Jingliang1, WANG Lipeng2   

  1. 1. Guangzhou Power Exchange Center Co., Ltd., Guangzhou 510000, China
    2. Tsinghua Keyue Co., Ltd., Beijing 100102, China
  • Received:2025-09-12 Revised:2025-12-19 Online:2026-01-13 Published:2026-01-28
  • Supported by:
    This work is supported by Science and Technology Project of Guangzhou Power Exchange Center Co., Ltd. (No.180000KC23080001).

摘要:

随着中国电力市场改革的推进,中国电力现货市场各个试点地区在实际结算过程中均出现了不同规模的不平衡资金,这极大地影响了每个市场主体的效益,降低了市场效率。因此,提出基于多主体博弈的电力市场不平衡资金分摊优化方法,考虑到各省在中长期交易规则、能源结构、现货市场建设等方面存在显著差异,使用基于主体(agent-based model,ABM)的建模方法,刻画系统运营商、发电商等各个市场主体,基于强化学习探索不同分摊方法下各个市场主体的行为决策。基于Stackel-berg博弈框架,挖掘各个市场主体在不同分摊方法下的内生演化机制,评估多种不平衡资金分摊方法对市场效率等方面的影响。基于ABM和强化学习刻画了多市场主体之间的互动特征,构建内生演化机制模拟市场主体的风险规避特征,有利于挖掘各省差异化的不平衡资金分摊优化方法,避免资源错配。仿真结果表明:ABM算法与内生演化机制充分反映了市场主体的行为策略。当不平衡资金分摊方法造成部分资金外流时,激励相容性指标达到0.22,各个市场主体呈现出强烈保守态势;当不平衡资金分摊方法通过价差惩罚引导市场主体时,各个市场主体均采取积极策略,有效提升了市场运行效率。

关键词: 电力市场, 不平衡资金, 现货市场

Abstract:

As China's power market reforms advance, pilot regions across the country have encountered imbalance funds of different scales during actual settlement processes in their spot power markets. This significantly impacts the profitability of each market participant and reduces market efficiency. Therefore, this study proposes an optimization method for allocating power market imbalance funds based on multi-agent games. Considering significant differences among provinces in medium-to-long-term trading rules, energy structures, and spot market development, an Agent-Based Model (ABM) is employed to characterize market participants such as system operators and power generators. Reinforcement learning is used to explore the behavioral decisions of these participants under different allocation methods. Second, within the Stackelberg game framework, we uncover the endogenous evolutionary mechanisms of market participants under different allocation methods, evaluating the impact of multiple imbalance fund allocation approaches on market efficiency and other metrics. By leveraging ABM and reinforcement learning to capture the interactive characteristics among multiple market participants and constructing endogenous evolutionary mechanisms to simulate risk-averse behaviors, this approach facilitates the identification of province-specific, optimized imbalance fund allocation proposed methods, thereby preventing resource misallocation. The results of the calculation examples show that the ABM algorithm and endogenous evolution mechanism fully reflect the behavioral strategies of market entities. When the imbalance fund allocation proposed method causes partial capital outflow, the incentive compatibility indicator reaches 0.22, and all market entities exhibit a strong conservative stance. When the imbalance fund allocation proposed method guides market entities through spread penalties, all market entities adopt proactive strategies, effectively improving market operational efficiency.

Key words: electricity market, imbalance funds, spot market


AI


AI小编
您好!我是《中国电力》AI小编,有什么可以帮您的吗?