Proxy Models for Fast Transfer of Static Uncertainty to Reservoir Performance Uncertainty

Abstract Petroleum reservoirs are heterogeneous and, therefore, uncertain. Heterogeneity and uncertainty are important for reservoir management. The transfer of geologic uncertainty through process performance is commonly achieved with flow simulation; however, full physics flow simulation requires significant professional and computer time. A proxy model tuned to the specific recovery process provides a means to quickly predict performance uncertainty caused by multiple geologic models and uncertain operating conditions. A limited number of flow simulation runs are used to calibrate the proxy, then calculations proceed quickly. The classical response surface approach is also illustrated. Details are given to the application of proxy models to the steam-assisted gravity drainage process. A proxy model based on the Butler steam-assisted gravity drainage theory is developed to predict the oil flow rate, cumulative oil production, and cumulative steam injection profiles during the rising and spreading steam chamber periods of a steam-assisted gravity drainage well pair. A synthetic example shows the efficiency of the methodology in terms of computation time and predictability. Investment decisions follow an information cycle that starts with acquisition, processing, and interpretation of subsurface data, creating the information needed to construct geostatistical models used to assess reservoir performance using criteria such as the number and type of wells. Different field development scenarios are considered, and decision criteria are applied to select and implement the optimum production scheme. Once wells are drilled and production starts, more information is available for optimizing decisions. The decisions made early in a reservoir life cycle are perhaps the most important, and they have to be made with limited information. Generating multiple geostatistical realizations to represent the state of uncertainty related to the actual reservoir properties is becoming common ( Deutsch and Journel, 1998). The available well data early in reservoir development sample less than one-trillionth of the reservoir. Seismic data provide more extensive data coverage but at a larger scale and with less precision related to reservoir facies, porosity, and permeability. The uncertainty is large; however, 100 or fewer realizations are commonly considered adequate to permit assessment of resources at 10, 50, and 90% confidence levels.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.