Efficient Big Data compression for distributed acoustic sensing via Gradient Quantization

Distributed Acoustic Sensing (DAS) data are vast, reaching terabytes, and can generate about 650 GB to 20 TB of data daily. To manage these large data volumes efficiently and ensure rapid network transfer and long-term storage, robust and reliable data processing techniques are essential. DAS enables real-time or near-real-time monitoring. Existing techniques are time-consuming, ineffective, and inefficient. In this study, we propose an enhanced version of the Gradient Quantization (GQ) model, focusing on improving the quantization steps during compression while preserving the real signals of DAS events. Additionally, we compared the results with six widely used data compression techniques. According to the results, our proposed method achieved higher compression and a higher space-saving ratio, outperforming other models while preserving the key features of DAS signals recorded by the Interrogator Unit (IU). In contrast, existing techniques often result in the loss of important event signals when higher compression is applied, thereby reducing data quality and usability, and limiting the precision of DAS in seismic data exploration. The proposed framework demonstrated superior performance, particularly in terms of data compression. Our approach reduced DAS data by a factor of 3.92 with a 74.46% space-saving ratio and achieved a Peak Signal-to-Noise Ratio (PSNR) of 65.81 dB, indicating high-quality signal preservation after compression.