Interpreting Downhole Pressure and Temperature Data from ESP Wells by Use of Inversion-Based Methods in Samabri Biseni Field

Increasing need for improved efficiency, service life and cost reduction using downhole real-time streaming sensor data is making electrical submersible pump (ESP) well operation management one of the most important issues in production optimization and improved oil recovery. Expanding the benefit of the downhole sensors is currently driving the need for embracing dynamic data-driven application systems (artificial intelligence, machine learning and deep learning) and big data tools by Oil and Gas industry to gain competitive advantage. One of the shortcomings for conventional data driven approach is that artificial intelligence (machine and/or deep learning) algorithms are totally decoupled from physics based modeling due to the lack of domain knowledge. As OEM for ESP, we have an industry proven ESP system simulator that can be used to generate training dataset for scalable data driven monitoring of ESP systems. Correct interpretation of temperature and pressure data can lead to improved accuracy of continuous downhole flow performance characteristics and reservoir properties such as static reservoir pressure and productivity index, which are key information to control and optimize ESP-based well production. In this work, a physics-based data driven model and inversion-based methods for model calibration and updating are developed for ESP well monitoring. The model is used as a forward engine and an inversion procedure is then added to interpret the measured data to estimate reservoir pressure, productivity index, downhole multiphase flow rates, and perform production allocation to improve hydrocarbon recovery and mitigate water/gas breakthrough risk. The new modeling framework introduces a fast and accurate forward model that incorporates specific measurements response functions for the physics-based data driven simulation model of permanent downhole gauge data in the ESP wells. Multiple inversion methods are used to interpret the downhole-measured data. Under the assumption of a subsurface multiphase flow model, the inversion approaches estimate well rates, back flow allocation, productivity index and reservoir pressure response specific to a given measurement domain by numerically reproducing the available measurements. The model and estimation techniques are evaluated with field data obtained from multiple wells located in a producing field. Many estimation simulations are performed using various sampling rates of the ESP AutographPC software. The satisfactory predictive accuracy of the physics-based data driven model makes the determination of multiphase flow and reservoir parameters computationally inexpensive, adaptive to operational changes, and suitable for online real-time system implementation.