Feature Extraction Techniques for Noisy Distributed Acoustic Sensor Data Acquired in a Wellbore

Distributed acoustic sensor (DAS) is a promising technology for real-time monitoring of wellbores and other infrastructures. However, the desired signals are often overwhelmed by background and environmental noise inherent in field applications. We present a suite of computationally inexpensive techniques for the real-time extraction of gas signatures from noisy DAS data acquired in a 5163-ft-deep wellbore. The techniques are implemented on three well-scale DAS datasets, each representing multiphase flow conditions with different gas injection volumes, fluid circulation rates, and injection methods. The proposed denoising techniques not only helped in optimizing the gas slug signature despite the high background noise but also reduced the DAS data size without compromising the signal quality.