Enhancing Arc Fault Detection in Industrial Applications: A Sample-Rate Super-Resolution Approach

This paper addresses the challenge of detecting series arc faults in industrial power systems, where detection devices typically operate with low-sample-rate open-loop current transformers that struggle to capture high-frequency arc signals. While high-sample-rate equipment in laboratory settings achieves satisfactory detection performance, real-world industrial applications often face limitations due to the inability to collect high-frequency signals. To overcome this, we propose a novel arc fault detection method that leverages sample-rate super-resolution via TimeGAN to upscale low-sample-rate data. This technique restores critical arc signal details, significantly improving fault detection accuracy even with limited hardware. Our experimental results demonstrate that using super-resolved data substantially enhances detection performance, offering a robust solution for real-world arc fault detection in industrial environments. This work contributes to advancing fault detection methodologies by introducing a cost-effective approach that can be applied to existing low-sample-rate detection systems, with broader implications for improving electrical safety in industrial power networks.