Automated Anomaly Detection in GIS SF6 Leakage Using Distributed Acoustic Sensor Networks and Deep Learning

**Abstract:** Gas-Insulated Switchgear (GIS) systems rely heavily on Sulfur Hexafluoride (SF6) as an insulating and arc-quenching medium. Accurate and rapid detection of SF6 leakage is critical for maintaining system reliability, worker safety, and environmental sustainability. Current leak detection methods are often intrusive, periodic, and lack real-time sensitivity. This paper introduces a novel, non-intrusive anomaly detection system leveraging a geographically distributed network of acoustic sensors and Deep Learning (DL) models to identify minute SF6 leakages in GIS enclosures. The system analyzes subtle acoustic signatures indicative of leak dynamics, achieving a consistent false alarm rate of < 2% while detecting deteriorations less than 10 grams of SF6 per hour – significantly exceeding the sensitivity of conventional methods. We present the system architecture, data acquisition methodology, DL model training and validation, and estimated commercialization roadmap. --- *This document is a part of the Freederia Research Archive. Explore our complete collection of advanced research at [en.freederia.com](https://en.freederia.com), or visit our main portal at [freederia.com](https://freederia.com) to learn more about our mission and other initiatives.*