Scholarlink

<p>Security-Constraint Unit Commitment (SCUC) is a complicated optimization problem in power system operation. Conventional model-driven methods have a significant computational burden, and traditional data-driven methods without enough historical data to consider multiple topological scenarios are challenging. Encountering the challenges and reducing the model complexity, this paper proposes RMGN-SCUC, an embedded mode hybrid data-model driven method that considers power grid topology information, and RMGN-SCUC is composed of two parts: the data-driven part ResMGCN for mapping decision-makings and the model-driven part mathematical models are used to obtain other desired results. First, ResMGCN learns from Case30 and Case118 samples modified by load disturbing and the synthetic network method, then outputs predicted unit states. Following, the predicted unit states input a simplified or original model to obtain the required results by solver-solving. The simulation results showed that the RMGN-SCUC effectively combines the advantages of both driven methods, reducing computation burden while maintaining results quality, achieving speedups of at least 50 × on different testing examples with high accuracy, and adaptability to multiple topologies.</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

RMGN-SCUC: A Hybrid Data-Model Driven Method for SCUC Considering Multiple Topologies

  • 期刊论文

  • 发表日期：

    2025年3月27日

查看全文
  • Lingfeng Kuang,
  • Xiaoqing Bai,
  • Peijie Li,
  • Zonglong Weng,
  • Jiale Zhang

摘要

Security-Constraint Unit Commitment (SCUC) is a complicated optimization problem in power system operation. Conventional model-driven methods have a significant computational burden, and traditional data-driven methods without enough historical data to consider multiple topological scenarios are challenging. Encountering the challenges and reducing the model complexity, this paper proposes RMGN-SCUC, an embedded mode hybrid data-model driven method that considers power grid topology information, and RMGN-SCUC is composed of two parts: the data-driven part ResMGCN for mapping decision-makings and the model-driven part mathematical models are used to obtain other desired results. First, ResMGCN learns from Case30 and Case118 samples modified by load disturbing and the synthetic network method, then outputs predicted unit states. Following, the predicted unit states input a simplified or original model to obtain the required results by solver-solving. The simulation results showed that the RMGN-SCUC effectively combines the advantages of both driven methods, reducing computation burden while maintaining results quality, achieving speedups of at least 50 × on different testing examples with high accuracy, and adaptability to multiple topologies.