Interpretation of Stratified Multiphase Flow via Distributed Acoustic Sensing: Insights from Radial-Axial Coupling Experiments

Summary

Accurate production profile characterization in multilayer and multiphase reservoirs and wellbores is essential for evaluating reservoir performance and optimizing production. Although distributed acoustic sensing (DAS) offers continuous monitoring capabilities, deciphering flow profiles remains challenging due to complex acoustic interactions between localized radial core inflow and turbulent axial pipe flow. For this study, we constructed a full-scale physical facility coupling radial core inflow (utilizing core samples with 20-md, 200-md, and 2,000-md permeabilities) with axial flow along the wellbore to systematically elucidate DAS frequency-response mechanisms across diverse optical fiber deployment positions, core permeabilities, flow rates, and phase holdups. The results reveal that reservoir permeability and flow rate primarily control DAS spectral energy distribution and characteristic frequency evolution, whereas phase holdup mainly affects response amplitude through changes in fluid/structure interaction. Furthermore, a novel frequency-band intelligent optimization approach of binary search combined with a prefix-sum algorithm was established to enhance signal extraction efficiency. Based on these findings, a dual-regime flow interpretation framework was developed for distinct production regimes: For high-rate flows (>100 m³/d), an automated image processing workflow integrated with Doppler-based flow velocity estimation achieved an inversion accuracy exceeding 90%; for low-rate flows (<100 m³/d), frequency-band energy (FBE) response charts were formulated, yielding robust flow rate correlations (R2 of 0.85). The proposed framework was validated by field DAS data sets from two oil wells. The layer-specific liquid-production error in a vertical well (Well A) was 10.43%, matching the reservoir permeability profile. In a horizontal well (Well B), the sectional liquid-production error was 15.22%, successfully identifying the main water-producing zone and guiding subsequent chemical water-plugging interventions. By combining full-scale physical modeling, efficient algorithm design, and field-closure validation, we provide a robust and practical workflow for interpreting production profiles in vertical, deviated, and horizontal wells.