Scholarlink

This paper proposes an efficient and resilient decision support system that uses unsupervised learning with explainable clustering to justify the selection of representative feeders in medium voltage distribution networks for estimating energy losses. The process of classifying feeders considers various technical factors affecting the operation of electrical distribution networks, such as the number of medium/low voltage distribution substations, the rated power of transformers, the average cross-section, and the length of the feeders. The most important stage involves grouping feeders through a clustering process, applying the K-means algorithm to determine the optimal number of categories (clusters) with similar technical characteristics. Explainable clustering offers more profound insights into feeder categories and the selection of representative feeders linked to the identified categories. The decision support system was tested using a database comprising 25 medium voltage feeders from a rural area belonging to a Romanian Distribution Network Operator. The results demonstrated that the explainable clustering integrated into the decision support system can be more effective at identifying representative feeders than traditional methods used by the Distribution Network Operator, providing advantages in calculating energy losses.

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 Decision Support System Incorporating Explainable Clustering for Selecting Medium Voltage Representative Feeders to Improve the Energy Loss Estimation Process

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mihai-Alexandru Baciu,
  • Gheorghe Grigoras,
  • Bogdan-Constantin Neagu,
  • Elena-Crenguta Bobric

摘要

This paper proposes an efficient and resilient decision support system that uses unsupervised learning with explainable clustering to justify the selection of representative feeders in medium voltage distribution networks for estimating energy losses. The process of classifying feeders considers various technical factors affecting the operation of electrical distribution networks, such as the number of medium/low voltage distribution substations, the rated power of transformers, the average cross-section, and the length of the feeders. The most important stage involves grouping feeders through a clustering process, applying the K-means algorithm to determine the optimal number of categories (clusters) with similar technical characteristics. Explainable clustering offers more profound insights into feeder categories and the selection of representative feeders linked to the identified categories. The decision support system was tested using a database comprising 25 medium voltage feeders from a rural area belonging to a Romanian Distribution Network Operator. The results demonstrated that the explainable clustering integrated into the decision support system can be more effective at identifying representative feeders than traditional methods used by the Distribution Network Operator, providing advantages in calculating energy losses.