| 引用本文: | 武晓冬,赵宇旸,孟祥齐,等.基于 CVaR 约束的双层主从博弈优化模型及其在多政策场景下的应用研究[J].电力系统保护与控制,2026,54(09):14-25. |
| WU Xiaodong,ZHAO Yuyang,MENG Xiangqi,et al.A bi-level leader-follower game optimization model based on CVaR constraints and its application in multi-policy scenarios[J].Power System Protection and Control,2026,54(09):14-25 |
|
| 摘要: |
| 在政策不断更新与现货电价剧烈波动的背景下，售电公司面临收益不确定与风险控制的双重挑战，亟需构建兼顾收益与风险的购售电决策模型。提出一种基于条件风险价值 (conditional value-at-risk, CVaR) 约束的双层主从博弈优化方法：上层售电公司制定长协购电方案与分段售电策略，下层用户以成本最小化原则响应，实现购售电收益 - 风险的协同优化。模型基于历史电价分布生成多场景数据，在对 CVaR 约束进行线性化处理后，采用交替迭代的最优响应方法求解主从博弈均衡，并与传统粒子群算法及基于 CVaR 约束的粒子群算法进行对比验证。结果表明，该方法在满足用户侧 80% 绿电比例约束的前提下，平均收益提升约 7%~9%，在置信水平 α=0.95 下的 CVaR 指标最高下降 96%，显著增强了收益稳健性与风险适应性，为售电公司在高波动电力市场中的策略优化提供了有效参考。 |
| 关键词: 售电公司 双层博弈 蒙特卡洛模拟 CVaR 约束 电力现货市场 |
| DOI：10.19783/j.cnki.pspc.251245 |
| 分类号: |
| 基金项目:国家自然科学基金项目资助(52507121)；山西省基础研究计划项目资助(202503021212088)；山西省高等学校科技创新项目资助(2025L006) |
|
| A bi-level leader-follower game optimization model based on CVaR constraints and its application in multi-policy scenarios |
|
WU Xiaodong1, ZHAO Yuyang1, MENG Xiangqi1, WEN Xueru2
|
|
1. School of Electric Power and Architecture, Shanxi University, Taiyuan 030013, China; 2. School of Electrical and Electronic Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
|
| Abstract: |
| Against the backdrop of continuously evolving policies and highly volatile spot electricity prices, electricity retailers face dual challenges of revenue uncertainty and risk control, necessitating decision-making models that balance profit and risk. This paper develops a bi-level leader-follower game optimization framework based on conditional value-at-risk (CVaR) constraints. In the upper level, the retailer formulates long-term contract procurement plans and tiered retail pricing strategies, while in the lower level, users respond based on a cost-minimization principle, achieving coordinated optimization of profit and risk in electricity trading. The model generates multi-scenario data based on historical electricity price distributions. After linearizing the CVaR constraints, an alternating iterative best-response approach is adopted to solve for the Stackelberg game equilibrium. The proposed method is further validated through comparison with traditional particle swarm optimization and CVaR-constrained PSO methods. Results show that, under the 80% green electricity consumption constraint on the user side, the proposed method improves average profit by approximately 7%~9%, while reducing the CVaR metric by up to 96% at a confidence level of α=0.95. This significantly enhances revenue robustness and risk adaptability, providing an effective reference for electricity retailers in optimizing strategies under highly volatile electricity markets. |
| Key words: electricity retailer bi-level game Monte Carlo simulation CVaR constraint electricity spot market |
