Scholarlink

<p>As the advance of technology and recent decades’ power distribution networks become more sophisticated, resource management has to be flexible and cost effective. Existing strategy is based on the concepts such as static optimization and centralized framework that is not very efficient in responding to dynamic changes concerning demand, supply and the topology of the communication network. To meet these challenges, this paper develops a new Elastic Planning Algorithm (EPA) by using Multi-Agent Deep Reinforcement Learning (MADRL), which can achieve decentralized and dynamic resource allocation. The proposed approach focus on using cooperative learning among the agents to achieve the desirable performance criteria including energy loss, resource use and fault repair time in various networks. It is shown in the MATLAB/Simulink and OpenDSS based simulations that the proposed algorithm yields a 28% improvement in energy losses, 32% augmentation in the resource usage, and 20% better fault recovery time than the conventional techniques. Based on these findings, the EPA framework demonstrates the ability for reorganizing of the resource management in the power distribution networks for future energy systems’ robustness.</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

Elastic Planning Algorithms for Flexible Resource Management in Power Distribution Networks

  • 期刊论文

  • 发表日期：

    2025年9月9日

查看全文
  • Guowei Liu,
  • Hao Deng,
  • Hao Dai,
  • Lisheng Xin,
  • Nan Ma,
  • Liming Wang

摘要

As the advance of technology and recent decades’ power distribution networks become more sophisticated, resource management has to be flexible and cost effective. Existing strategy is based on the concepts such as static optimization and centralized framework that is not very efficient in responding to dynamic changes concerning demand, supply and the topology of the communication network. To meet these challenges, this paper develops a new Elastic Planning Algorithm (EPA) by using Multi-Agent Deep Reinforcement Learning (MADRL), which can achieve decentralized and dynamic resource allocation. The proposed approach focus on using cooperative learning among the agents to achieve the desirable performance criteria including energy loss, resource use and fault repair time in various networks. It is shown in the MATLAB/Simulink and OpenDSS based simulations that the proposed algorithm yields a 28% improvement in energy losses, 32% augmentation in the resource usage, and 20% better fault recovery time than the conventional techniques. Based on these findings, the EPA framework demonstrates the ability for reorganizing of the resource management in the power distribution networks for future energy systems’ robustness.