Scholarlink

This study proposes an innovative, intelligent system development for automatically localizing short-circuit faults. This system is based on the Linear Fusion of three powerful automatic regression models, aiming to maximize the accuracy and reliability of the proposed localization model by referring to information from different sources. The proposed approach is validated on a database of eleven short-circuit-type fault classes. This class examples are generated at various locations of three-phase High Voltage transmission lines of 30 kV, 50 Hz, and a length of 100 km [1]. The current and voltage signals are first segmented into 20ms windows and reduced to 60 descriptors using Discrete Wavelet Transform to facilitate regression. Three models, Support Vector Regression, Multilayer Perceptron, and Radial Basis Function, are then implemented to locate faults in high-voltage transmission lines. Finally, a fusion process is proposed to enhance the final decision reliability. The results are very satisfactory, with a prediction rate of 99.76% after fusion, a Mean Squared Error reduced to 0.03157, and a Root Mean Square Error reduced to 0.17767. These results demonstrate the proposed system’s effectiveness, thus offering a new contribution to the early localization of faults occurring in transmission lines.

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 Novel Short-Circuit Fault Localization Approach in Long High Voltage Lines Through DWT and Linear Fusion of Regression Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hendel Mounia,
  • Sakina Behilil,
  • Mostefa Brahami,
  • Imen Souhila Bousmaha

摘要

This study proposes an innovative, intelligent system development for automatically localizing short-circuit faults. This system is based on the Linear Fusion of three powerful automatic regression models, aiming to maximize the accuracy and reliability of the proposed localization model by referring to information from different sources. The proposed approach is validated on a database of eleven short-circuit-type fault classes. This class examples are generated at various locations of three-phase High Voltage transmission lines of 30 kV, 50 Hz, and a length of 100 km [1]. The current and voltage signals are first segmented into 20ms windows and reduced to 60 descriptors using Discrete Wavelet Transform to facilitate regression. Three models, Support Vector Regression, Multilayer Perceptron, and Radial Basis Function, are then implemented to locate faults in high-voltage transmission lines. Finally, a fusion process is proposed to enhance the final decision reliability. The results are very satisfactory, with a prediction rate of 99.76% after fusion, a Mean Squared Error reduced to 0.03157, and a Root Mean Square Error reduced to 0.17767. These results demonstrate the proposed system’s effectiveness, thus offering a new contribution to the early localization of faults occurring in transmission lines.