Scholarlink

<p>This paper introduces GRATE-DRL-AI (Graph-Embedded Transfer and Deep Reinforcement Learning Artificial Intelligence), an advanced AI-driven framework that integrates deep learning and optimization techniques to enhance next-generation distribution system planning (DSP). The proposed approach combines graph learning, transfer learning, deep reinforcement learning (DRL), and physics-guided neural networks to address the increasing complexity and uncertainty in modern power distribution networks with high penetration of distributed energy resources (DERs). By leveraging deep learning for spatial feature extraction and optimization techniques for decision-making, GRATE-DRL-AI ensures improved computational efficiency, cost-effectiveness, and resilience. Comparative evaluations on IEEE 33-bus and 123-bus systems demonstrate that the proposed framework outperforms traditional methods such as mixed-integer conic programming (MICP), genetic algorithms (GA), and conventional reinforcement learning (RL). The results indicate significant reductions in total costs, faster computation times, and enhanced robustness against uncertainties. These findings highlight the transformative potential of deep learning-driven optimization for smart and sustainable power distribution planning.</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

Integrating Deep Learning and Optimization for Next-Generation Distribution System Planning

  • 期刊论文

  • 发表日期：

    2025年8月5日

查看全文
  • Harshit Singh,
  • Sachin Singh,
  • Rajiv Kumar Singh,
  • Fidele Maniraguha

摘要

This paper introduces GRATE-DRL-AI (Graph-Embedded Transfer and Deep Reinforcement Learning Artificial Intelligence), an advanced AI-driven framework that integrates deep learning and optimization techniques to enhance next-generation distribution system planning (DSP). The proposed approach combines graph learning, transfer learning, deep reinforcement learning (DRL), and physics-guided neural networks to address the increasing complexity and uncertainty in modern power distribution networks with high penetration of distributed energy resources (DERs). By leveraging deep learning for spatial feature extraction and optimization techniques for decision-making, GRATE-DRL-AI ensures improved computational efficiency, cost-effectiveness, and resilience. Comparative evaluations on IEEE 33-bus and 123-bus systems demonstrate that the proposed framework outperforms traditional methods such as mixed-integer conic programming (MICP), genetic algorithms (GA), and conventional reinforcement learning (RL). The results indicate significant reductions in total costs, faster computation times, and enhanced robustness against uncertainties. These findings highlight the transformative potential of deep learning-driven optimization for smart and sustainable power distribution planning.