Scholarlink

As an important energy conversion equipment in the power grid, power transformers have a complex temperature field, which is related to their own operation life and the stability of the power system. Therefore, accurately obtaining the characteristics of the temperature field and hot spot temperature of transformers has important practical value. At present, there are two methods for constructing the temperature field of transformers, including model driven and data driven. However, existing data-driven methods couple temporal and spatial information when constructing networks, resulting in imbalanced spatiotemporal modeling and making the network training process difficult. Therefore, we propose a transformer temperature field prediction method based on spatiotemporal decoupling training, including spatial feature vector quantification and temporal feature modeling. It can balance the spatiotemporal modeling process, reduce the difficulty of model training, and effectively improve the accuracy of transformer temperature field prediction.

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

Decoupled Spatial–Temporal Model for Temperature Field Prediction of Transformer

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hao Liu,
  • Jinrui Gan,
  • Qiang Zhang,
  • Jie Tong,
  • Zhonghao Zhang,
  • Pengfei Tang,
  • Qiong Fang,
  • Songyuan Li,
  • Chi Zhang

摘要

As an important energy conversion equipment in the power grid, power transformers have a complex temperature field, which is related to their own operation life and the stability of the power system. Therefore, accurately obtaining the characteristics of the temperature field and hot spot temperature of transformers has important practical value. At present, there are two methods for constructing the temperature field of transformers, including model driven and data driven. However, existing data-driven methods couple temporal and spatial information when constructing networks, resulting in imbalanced spatiotemporal modeling and making the network training process difficult. Therefore, we propose a transformer temperature field prediction method based on spatiotemporal decoupling training, including spatial feature vector quantification and temporal feature modeling. It can balance the spatiotemporal modeling process, reduce the difficulty of model training, and effectively improve the accuracy of transformer temperature field prediction.