Scholarlink

<p>This paper focuses on the real-time pricing problem for the welfare equilibrium of multitype electricity users in smart grids and proposes a real-time pricing model based on a bi-level structure incorporating Double Q-learning and prioritized experience replay (DQL-PER). Firstly, welfare models for power suppliers and different types of users are constructed, taking into account their diverse electricity consumption behaviors and demands to achieve welfare equilibrium. Then, the models of each stakeholder are respectively transformed based on the DQL-PER algorithm. This study verifies the convergence of a model comprising 200 residential, 100 commercial, and 50 industrial users with 1-hour time granularity within 5,000 iterations. Simulation experiments demonstrate the model’s effectiveness across different pricing schemes and demand volatility scenarios. The comparison of user welfare using different algorithms shows that the DQL-PER algorithm achieves significant improvements compared with Q-learning (QL) and Double Q-learning (DQL) algorithms. This research provides a novel and practical solution for real-time electricity pricing in smart grids.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Real - Time Pricing Model for Welfare Equilibrium of Multitype Electricity Users in Smart Grid: A bi-level Structure with Double Q-learning and Prioritized Experience Replay Algorithm

  • 期刊论文

  • 发表日期：

    2025年10月6日

查看全文
  • Haixiao Song,
  • Yan Gao,
  • Zhongqing Wang

摘要

This paper focuses on the real-time pricing problem for the welfare equilibrium of multitype electricity users in smart grids and proposes a real-time pricing model based on a bi-level structure incorporating Double Q-learning and prioritized experience replay (DQL-PER). Firstly, welfare models for power suppliers and different types of users are constructed, taking into account their diverse electricity consumption behaviors and demands to achieve welfare equilibrium. Then, the models of each stakeholder are respectively transformed based on the DQL-PER algorithm. This study verifies the convergence of a model comprising 200 residential, 100 commercial, and 50 industrial users with 1-hour time granularity within 5,000 iterations. Simulation experiments demonstrate the model’s effectiveness across different pricing schemes and demand volatility scenarios. The comparison of user welfare using different algorithms shows that the DQL-PER algorithm achieves significant improvements compared with Q-learning (QL) and Double Q-learning (DQL) algorithms. This research provides a novel and practical solution for real-time electricity pricing in smart grids.