Scholarlink

This study addresses the computational challenges in simulating the temperature field of GIS for digital twin applications. A non-intrusive ROM framework is proposed, integrating POD for data dimensionality reduction and RBFNN for surrogate modeling. Focusing on a 550 kV three-phase common-enclosure GIS, steady-state temperature fields under varying ambient temperatures and current loads are simulated via multi-physics coupling. The POD extracts dominant modes capturing over 99.99% system energy with only two modes, while the RBFNN maps operating parameters to temperature fields. Validated against 20 test cases, the ROM achieves a maximum relative error of 0.95% and a 150× speedup compared to full-order simulations. This approach enables real-time temperature prediction for GIS digital twins, significantly enhancing operational reliability and design optimization.

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

A Reduced-Order Model for GIS Equipment Temperature Field Based on POD-RBF Neural Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Wenxin Li,
  • Zhilin Chu,
  • Qingyu Wang,
  • Siyuan Liu,
  • Huidong Tian

摘要

This study addresses the computational challenges in simulating the temperature field of GIS for digital twin applications. A non-intrusive ROM framework is proposed, integrating POD for data dimensionality reduction and RBFNN for surrogate modeling. Focusing on a 550 kV three-phase common-enclosure GIS, steady-state temperature fields under varying ambient temperatures and current loads are simulated via multi-physics coupling. The POD extracts dominant modes capturing over 99.99% system energy with only two modes, while the RBFNN maps operating parameters to temperature fields. Validated against 20 test cases, the ROM achieves a maximum relative error of 0.95% and a 150× speedup compared to full-order simulations. This approach enables real-time temperature prediction for GIS digital twins, significantly enhancing operational reliability and design optimization.