Oil-Filled Submarine Cable Leak Localization Method Based on Deep Neural Networks

This study focuses on the problem of leak localization in high-pressure oil-filled submarine cables. These cables employ oil-immersed insulation technology, which not only provides excellent insulating performance and high current-carrying capacity but also adapts to complex and harsh seabed conditions. Despite this, oil-filled submarine cables still face significant challenges in leak detection. Traditional methods involving manual inspection and underwater robots are costly and thus not widely applicable. Additionally, advanced sensor technologies are limited due to the unique structure of the cables. Existing leak detection methods mainly rely on analyzing the flow state of insulating oil, but the accuracy of this approach needs improvement, and it cannot determine the leak damping at the leak point. To address this, the study develops an efficient leak localization and leak damping identification model that combines a reduced-order leak flow model with deep neural networks. This model aims to quickly locate the leak points and digitally twin the leak conditions. Research results show that this identification model has high accuracy and feasibility, effectively reducing the cost of leak localization in oil-filled submarine cables.