| 吴凡,杨永标,徐青山,赵祎静.基于DDPG和电价感知的中长期双边竞价策略优化模型[J].电力需求侧管理,2026,28(1):113-119 |
| 基于DDPG和电价感知的中长期双边竞价策略优化模型 |
| Mid-to long-term bilateral bidding strategy optimization using DDPG with price perception |
| 投稿时间：2025-09-22 修订日期：2025-10-25 |
| DOI：10.3969/j.issn.1009-1831.2026.01.016 |
| 中文关键词: 中长期双边交易 多周期电价感知 深度确定性策略梯度算法 强化学习 策略优化 |
| 英文关键词: mid-to long-term bilateral transactions multi-period price perception DDPG reinforcement learning strategy optimization |
| 基金项目:国家重点研发项目"低碳高可靠城市配电系统示范工程"(2024ZD0800800) |
|
| 摘要点击次数: 0 |
| 全文下载次数: 0 |
| 中文摘要: |
| 随着电力市场改革的不断深化,用户侧参与中长期电力市场交易的活跃度持续提升,其对电价变化的感知与响应行为已成为影响市场资源配置效率的重要因素。针对传统中长期竞价模型在刻画用户动态行为和处理连续策略优化方面的不足,引入多周期电价感知机制与深度确定性策略梯度(deep deterministic policy gradient,DDPG)算法,构建发电商中长期双边竞价策略优化方法。首先,建立基于Sigmoid结构的多周期非线性电价感知模型,引入周期调节与不确定扰动因素,刻画用户在不同电价周期下的响应阈值与行为差异;其次,基于DDPG算法设计发电商策略学习框架,并设置差异化奖励函数,兼顾利润、社会福利与市场公平性,实现连续动作空间下的策略自适应优化;最后,通过仿真实验验证所提模型在提升市场收益、促进负荷调节和实现多目标均衡方面的有效性。研究结果表明,该方法能够更精准地模拟电力用户行为,优化发电商竞价策略,增强电力市场的运行稳定性与社会福利水平。 |
| 英文摘要: |
| As electricity market reform deepens, the participation of end users in long term electricity transactions is intensifying, and their price perception and demand response behavior have become critical determinants of resource allocation efficiency. To address the limited capability of conventional bidding models to represent dynamic user behavior and to handle continuous decision spaces, a bilateral bidding strategy optimization framework is developed that integrates a multi period price perception mechanism with the deep deterministic policy gradient (DDPG) algorithm. A nonlinear price perception model with a Sigmoid structure is first established, in which cycle adjustment factors and stochastic perturbations are introduced to characterize response thresholds and behavioral heterogeneity under different price regimes. On this basis, a DDPG based policy learning architecture is constructed, and a differentiated reward function is designed to jointly account for profit maximization, social welfare improvement, and market fairness enhancement, thereby enabling adaptive optimization in a continuous action space. The proposed framework is evaluated through numerical simulation studies on representative market scenarios. The results indicate that market revenues can be increased, load modulation and temporal shifting can be promoted, and a favorable trade off among multiple operational objectives can be achieved. Moreover, user behavior is captured with higher fidelity, generator bidding strategies are more efficiently optimized, and the stability and social welfare of the electricity market are enhanced. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|