We introduce a novel Machine Learning (ML) approach for processing distributed fiber-optic sensing (DFOS) data that enables dynamic flow profile monitoring using a fiber-optic e-line cable deployed in a gas condensate well and compare it to a conventional approach. DFOS technology has the potential to provide more efficient and dynamic flow profiles compared to traditional methods, particularly in high rate gas wells where production logs (PL) are recorded at reduced rates to avoid tool lifting.
Distributed acoustic and temperature sensing (DAS & DTS) data were acquired simultaneously while the well was producing ~70 MMSCF/D gas. Conventional PL data was also acquired under the same condition to validate the flow profiling results obtained from DFOS measurements. This paper describes a novel data processing approach where ML based models for pattern recognition were applied to obtain the signatures of different fluid types. Flow profiling is achieved by applying multiple data models to address three key questions for inflow profiling: (1) which zones are producing? (2) what is the phase? and (3) what is the flow rate?
A blind test was set up to avoid results contamination. The processing and interpretation of DFOS data and PL data were carried out independently and results were compared only when the work on both datasets was completed. The comparison demonstrates a good match between two measurements for gas inflow profile with an average error of about 1% in relative gas rate allocation along the four producing perforated intervals. Flow profile in a single-phase gas producing well was accurately determined by DFOS data analysis and the liquid production rate was then re-calculated using condensate-gas ratio (CGR) to obtain liquid and gas production rates at standard surface condition. The well was connected to a test separator during the entire acquisition period, and accurate gas, condensate and water production rates were obtained in real-time at surface condition.
The hybrid processing technique was applied for the first time among our well stock and resulted in accurate gas inflow profiling. To further validate the performance of the presented approach, the authors intend to repeat the test in other high rate gas producing wells, including wells with permanently installed fiber. Multi-disciplinary teamwork involved collaboration between operator and vendor and allowed for efficient operational execution. The result of the risk assessment ensured the selection of the best candidate well ensuring minimum sand production at the optimum production rate, optimization of stationary time for DFOS data acquisition and cable armor erosion model.