半月刊

ISSN 1000-1026

CN 32-1180/TP

+高级检索 English
基于深度强化学习的激励型需求响应决策优化模型
作者:
作者单位:

1.南瑞集团有限公司(国网电力科学研究院有限公司),江苏省南京市 211106;2.智能电网保护与运行控制国家重点实验室,江苏省南京市 211106

摘要:

随着中国电力市场化改革的推进,售电侧市场逐步开放,售电商可以聚合大量的分散负荷参与电力市场环境下的需求响应。文中提出以售电商和用户综合收益最大化为目标的基于深度强化学习的激励型需求响应建模和求解方法。首先,建立售电商和用户的需求响应模型,通过引入时间-价格弹性,改进现有的用户响应模型,考虑用户对相邻时段补贴价格差的反应。然后,基于马尔可夫决策过程框架构建补贴价格决策优化模型,并设计基于深度Q学习网络的求解算法。最后,以1个售电商和3个不同类型的用户为例进行仿真计算,通过分析算法收敛性和对比不同模型及参数下的优化结果,验证了改进模型的合理性和生成策略的有效性,并分析了激励型需求响应对售电商以及用户的影响。

关键词:

基金项目:

国家重点研发计划资助项目(2018AAA0101504);国家电网公司科技项目(5700-202019364A-0-0-00)。

通信作者:

作者简介:

徐弘升(1987—),男,通信作者,博士,高级工程师,主要研究方向:强化学习、深度学习与电力大数据分析。E-mail:xuhongsheng@sgepri.sgcc.com.cn
陆继翔(1973—),男,正高级工程师,主要研究方向:人工智能与电网调度。E-mail:lujixiang@sgepri.sgcc.com.cn
杨志宏(1968—),男,教授级高级工程师,主要研究方向:电网调控自动化和变电站自动化。E-mail:yangzhihong@sgepri.sgcc.com.cn


Decision Optimization Model of Incentive Demand Response Based on Deep Reinforcement Learning
Author:
Affiliation:

1.NARI Group Corporation (State Grid Electric Power Research institute), Nanjing 211106, China;2.State Key Laboratory of Smart Grid Protection and Control, Nanjing 211106, China

Abstract:

With the advancement of the electricity market reform in China, electricity side market is gradually opening up. The electricity retailers can aggregate a large number of distributed loads to participate in the demand response in the electricity market environment. In this paper, the modeling and solving methods of incentive demand response based on deep reinforcement learning are proposed to maximize the comprehensive profits of both retailers and customers. Firstly, the demand response models for retailers and customers are established. By introducing the time-price elasticity, the recent customer response model is improved, which takes the customer response to the subsidy price difference between adjacent periods into account. Then, based on the Markov decision process framework, an optimization model of subsidy price decision is constructed, and a solution algorithm based on deep Q-learning network is designed. Finally, simulation calculation is performed using one retailer and three different types of customers as examples. By analyzing the convergence of the algorithm and comparing the optimization results of different models and parameters, the rationality of the improved model and the effectiveness of the generated strategy are verified, and the impact of incentive demand response on the retailer and customers is analyzed.

Keywords:

Foundation:
This work is supported by National Key R&D Program of China (No. 2018AAA0101504) and State Grid Corporation of China (No. 5700-202019364A-0-0-00).
引用本文
[1]徐弘升,陆继翔,杨志宏,等.基于深度强化学习的激励型需求响应决策优化模型[J].电力系统自动化,2021,45(14):97-103. DOI:10.7500/AEPS20200208001.
XU Hongsheng, LU Jixiang, YANG Zhihong, et al. Decision Optimization Model of Incentive Demand Response Based on Deep Reinforcement Learning[J]. Automation of Electric Power Systems, 2021, 45(14):97-103. DOI:10.7500/AEPS20200208001.
复制
分享
历史
  • 收稿日期:2020-02-08
  • 最后修改日期:2021-01-19
  • 录用日期:
  • 在线发布日期: 2021-07-21
  • 出版日期:
相关附件