Scholarlink

<p>Recent disasters demonstrate how modern power distribution systems are susceptible to such extremes; therefore, intelligent adaptive resilience is required. The conventional concepts can basically embrace the application of standard and rule-of-thumb decision-making techniques which are incompetent to address the multifaceted characteristic of a power grid system prolonging the period of power outage and raising the energy loss rates. In response to these challenges, this research proposes the development of a new graph neural network-reinforcement learning approach that can improve the reliability of distribution networks in addition to optimizing their performance. Contrary to existing methods, the proposed method uses GNNs that capture the precise characteristic of power systems in terms of topology and time and RL that supports the optimal decision in network reconfiguration and load prioritization. The simulation results show enhancements of various aspects like outage duration reduced by 50%, ENS lowered to 0.65&#xa0;MWh from 1.25&#xa0;MWh, and SAIDI improved to 1.6&#xa0;h from 3.2&#xa0;h. Employing the MATLAB/Simulink platform and the application of power system stability, this framework is validated through comprehensive simulations on IEEE benchmark systems for a future-ready scalable power systems solution.</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

AI-driven resilience analysis of distribution networks under extreme events

  • 期刊论文

  • 发表日期：

    2025年6月4日

查看全文
  • Hao Dai,
  • Guowei Liu,
  • Lisheng Xin,
  • Longlong Shang,
  • Hao Deng,
  • Nan Ma

摘要

Recent disasters demonstrate how modern power distribution systems are susceptible to such extremes; therefore, intelligent adaptive resilience is required. The conventional concepts can basically embrace the application of standard and rule-of-thumb decision-making techniques which are incompetent to address the multifaceted characteristic of a power grid system prolonging the period of power outage and raising the energy loss rates. In response to these challenges, this research proposes the development of a new graph neural network-reinforcement learning approach that can improve the reliability of distribution networks in addition to optimizing their performance. Contrary to existing methods, the proposed method uses GNNs that capture the precise characteristic of power systems in terms of topology and time and RL that supports the optimal decision in network reconfiguration and load prioritization. The simulation results show enhancements of various aspects like outage duration reduced by 50%, ENS lowered to 0.65 MWh from 1.25 MWh, and SAIDI improved to 1.6 h from 3.2 h. Employing the MATLAB/Simulink platform and the application of power system stability, this framework is validated through comprehensive simulations on IEEE benchmark systems for a future-ready scalable power systems solution.