Abstract: The presence of multi-source disturbances such as the connection of power electronic devices, load changes, switch operations, and environmental noise can overwhelm the signals indicating potential hazards in new distribution network lines, making it difficult to accurately assess the status and development trends of these hazards. Starting from the real-time data collected by the on-site traveling wave monitoring terminal, this paper proposes a defect signal extraction technology for new distribution network based on eaglefish optimization algorithm and variable scale mode filtering. The aim is to achieve precise extraction of defect emission signals under complex new distribution network architectures. Firstly, the signals collected from the line are decomposed using the eaglefish optimization algorithm and VMD algorithm to obtain appropriate modal functions. Then, a shift-invariant sparse learning function is introduced to construct a variable-scale modal filtering algorithm, which reconstructs the modal functions through filtering. Defect information is obtained via the envelope frequency. Finally, through case studies, the accuracy and practicality of this algorithm are verified.