This study aims to apply numerical rate transient analysis (RTA) in a semi-automated fashion using public data. The main objective is to have an initial calibration of reservoir properties, such as relative permeability, and an enhanced, physics-based method for production forecasting. Numerical RTA is applied to approximately 100 wells to match actual production data. The primary objective is to minimize errors in predicted rates of oil, gas and water vs. actual production by adjusting reservoir parameters. The ultimate goal is an optimized, semi-automated RTA workflow requiring minimal human intervention, providing a robust starting point for detailed numerical RTA.
Public production data from 300 Permian wells are leveraged in the study. 100 of the wells are used for RTA calibration, while the other 200 are used as blind tests in a hindcasting exercise. A Permian field-wide equation of state model is used to estimate relevant PVT properties. A common PVT (Rsi, Boi, μoi) and relative permeability set are assumed for the entire well population, iteratively adjusting relative permeability, and other relevant reservoir parameters, to minimize overall errors. A relevant error minimization function (root mean square) is defined to capture key study objectives across the combined 100 well set.
Two significant benefits emerge from this novel approach. First, the relative permeability curves, informed by the entire well population, provide an excellent starting point for detailed numerical RTA, surpassing the current methodology's time-consuming iterations. Second, the approach enhances aspects of EUR forecasting with public data. By linking oil rates and GOR through physics-based models calibrated to the entire well population, the workflow offers improvements to EUR forecasting. The methodology's accuracy is validated through hindcasting on 300 nearby wells, comparing against traditional decline curve forecasts with one and then two years of production history.
The success of this automated, numerical RTA workflow carries significant implications. It not only benefits projects using public data, such as regional neural networks, acquisitions, and trades but also reduces the time required (currently 80-160 hours per field) to establish baseline reservoir properties in new areas. Successful implementation and further work could streamline workflows across diverse projects, ushering in a more efficient era in RTA calibration and production forecasting with broader applicability and accelerated project timelines.
