基于深度强化学习的配电网多主体协同电压控制方法
A Multi-Agent Cooperative Voltage Control Method of Distribution System Based on Deep Reinforcement Learning
耿天旭 1梁俊宇 2龚新勇 3张贵鹏 4王笑雪5
作者信息
- 1. 云南电网有限责任公司昆明禄劝供电局,云南昆明 651500
- 2. 云南电网有限责任公司电力科学研究院,云南昆明 650217
- 3. 云南电网有限责任公司昆明供电局,云南昆明 650106
- 4. 云南电网有限责任公司生产技术部,云南昆明 650032
- 5. 河北工业大学电气工程学院,河北天津 300000
- 折叠
摘要
随着配电市场化的推进和普及,配电网中分布式电源、分布式储能、微电网等众多可控资源以多利益主体形态呈现,这使得配电网无法强制对其进行调度与控制.为了充分利用多主体可控资源,该文提出了一种基于深度强化学习的、考虑多主体参与的配电网电压优化控制方法.首先,建立配电网运营商和多利益主体的主从博弈电压控制模型.随后,为了保障多主体间的公平性和隐私性,提出多主体基于不完全信息和深度强化学习Actor-Critic算法的动态决策方法,使其在自身利益最大化的同时为配电网提供电压控制辅助服务.最后,在改进的IEEE33 和IEEE 123 算例中对所提电压控制方法的可行性和有效性进行了验证.
Abstract
With the promotion of the market-orientated reform of power distribution,numerous controllable resources(such as distributed generators,distributed energy storage systems,etc.)in the distribution network are presented in the form of multi-stakeholders,which makes it it impossible for the distribution network to forcibly dispatch and control them.To make full use of controllable resources,this paper proposes a voltage optimization control method for distribution networks based on deep reinforced learning and considering multi-agent participation.Firstly,a Stackelberg Game voltage control model of the distribution network operator and the multi-stakeholder is established.Second,to ensure the fairness and privacy among multi-agents,a multi-agent dynamic decision-making method based on incomplete information and deep reinforcement learning Actor-Critic algorithm is proposed to maximize its own interests and provide auxiliary voltage control service for the distribution network.Finally,the feasibility and effectiveness of the proposed method are verified in the improved IEEE33 and IEEE 123 examples.
关键词
配电网电压控制/电力市场/多利益主体/深度强化学习/主从博弈Key words
voltage control of distribution network/electricity market/multi-stakeholder/deep reinforcement learning/Stackelberg Game引用本文复制引用
基金项目
中国博士后科学基金资助项目(2021T140174)
中国南方电网有限责任公司创新项目(YNKJXM20222334)
出版年
2024
NETL
NSTL
万方数据