Uncertainty Quantification of Forecasted Oil Recovery using Dynamic Model Ranking with Application to a ME Carbonate Reservoir

Abstract

Abstract History matching, being an ill-posed optimization problem, attempts to render multiple realizations of reservoir models that satisfy a given objective function with applicable constraints. A variety of assisted history-matching (AHM) techniques is currently being developed and used with the main objective to generate statistically diverse ensembles of history-matched models to capture the uncertainty in the distribution of reservoir parameters. This paper targets the outstanding questions of how to a) rigorously quantify the uncertainty in the distribution of the most prominent reservoir parameters that govern the reservoir connectivity and b) rank the history-matched models and identify the model candidates for production forecasting without compromising the notion of uncertainty. A workflow has been developed that integrates the modules for AHM and dynamic model ranking (DMR) based on forecasted oil recovery factors (ORFs). A pattern recognition methodology based on a kernel -means clustering algorithm is used to identify key reservoir models. The reduced set of models is used to minimize the computational load for forecast-based analysis, while retaining the knowledge of the uncertainty in the recovery factor. The comprehensive probabilistic AHM and DMR workflow was implemented at the operator's North Kuwait Integrated Digital Oilfield (KwIDF) collaboration center. It delivers an optimized reservoir model for waterflood management and automatically updates the model quarterly with geological, production, and completion information. Introduction Reservoir characterization is one of the most important and comprehensive tasks in the process of field development planning (FDP). The key component in FDP is developing a history-matched reservoir model that correctly represents the physics of the actual reservoir, reconciles with logging and historic production data, and accounts for reservoir uncertainties. Because of the nature of the inverse problem, history matching delivers no unique solution, and multiple history-matched models can satisfy the objective function within given constraints. Recently, a variety of algorithms (Schulze-Riegert and Ghedan 2007; Oliver and Chen 2011; Rwenchungura et al. 2011; Maučec et al. 2013a and references therein) have been considered that enable AHM and uncertainty quantification; however, there are still unanswered questions:How to efficiently utilize and extract the knowledge from the many history-matched models for FDP.Which model should be chosen for the production forecasting?If it is possible to identify and optimize, on a few representative models for which FDP or any forecast-based analysis should be performed, while still comprehensively capturing the uncertainty. The present work focuses particularly on addressing these questions, with a field case study applied to the Sabriyah-Mauddud (SaMa) reservoir operated by the Kuwait Oil Company (KOC). As part of a comprehensive strategy to enhance the overall productivity of reservoirs through the application of intelligent digital oilfield (iDOF) concepts, KOC has already initiated an assessment of their major SaMa reservoir for conversion to an integrated iDOF master platform (Al-Abbasi et al. 2013; Al-Jasmi et al. 2013 and references therein). KOC is also aiming to increase efficiency by automating work processes and shortening observation-to-action cycle time. With these goals in mind, several next generation smart workflows have been delivered to KOC. These workflows combine subsurface waterflooding optimization (SWFO) (Khan et al. 2013), integrated production optimization (IPO), and simulation model update and ranking (SMUR) (Maučec et al. 2013a).