An Intelligent Localization Method for Defects in High Voltage Cable Buffer Layers Based on Multi-Scale Spatial Feature Fusion

This paper presents an innovative approach for localizing defects in high voltage cable buffer layers through a method based on multi-scale spatial feature fusion. Addressing the crucial need for accurate identification of erosion defects, this research leverages Frequency Domain Reflectometry (FDR) enhanced with multi-scale Convolutional Neural Networks (CNNs) to improve localization precision. The methodology employs a fusion of features extracted at multiple scales, enabling the system to effectively identify defects across various temporal resolutions. The proposed framework is thoroughly evaluated using a com- prehensive dataset simulating different defect scenarios in cable buffer layers. The results demonstrate significant improvements in localization accuracy and reliability, providing a promising solution for enhancing the safety and operational efficiency of high-voltage transmission networks.