Search citation statements
Paper Sections
Citation Types
Year Published
Publication Types
Relationship
Authors
Journals
The Shushufindi field (Fig. 1) was discovered in 1972 and is sparsely developed, with only 181 wells covering an area of approximately 400 km2. Recovered whole core from the well SSF-151D presents very thin streaks of quartz material, along with carbon and amber, which form vertical permeability barriers in the cored section, approximately one inch thick. Production profiles for many wells in the field present a very gradual increase in the water cut from an active edgewater drive aquifer and demonstrate a layered system. This implies that the vertical permeability barriers are areally extensive, and that they have a substantial impact on production predictions and development scenarios. The challenge for the subsurface team is to represent these thin barriers in a predictive simulation model. A standard Kv:Kh ratio is insufficient to mimic the observed production response. A very fine-scale geomodel could be considered but is computationally expensive. Vertical transmissibility multipliers are useful in a simulation model, but experiments demonstrate that the lateral extent of the barriers is discontinuous. Diagnostic plots of the well water/oil ratio (WWOR) versus the cumulative liquid production (WLPT) can identify groups of wells with layered production profiles. This information provides a basis to infer the potential for a layered production response from existing well logs. To introduce such discontinuities, a series of randomly generated vertical baffles have been created which provide an improved simulation response in a relatively coarse simulation model. Early water breakthrough, and excessive water production in some wells caused premature re-completions into other zones. This type of sub-surface modeling provides support for the introduction of intelligent completions to isolate specific layers as part of the development plan. This paper presents the techniques implemented by the team, the results to date, and a prognosis for the future of the field.
The Shushufindi field (Fig. 1) was discovered in 1972 and is sparsely developed, with only 181 wells covering an area of approximately 400 km2. Recovered whole core from the well SSF-151D presents very thin streaks of quartz material, along with carbon and amber, which form vertical permeability barriers in the cored section, approximately one inch thick. Production profiles for many wells in the field present a very gradual increase in the water cut from an active edgewater drive aquifer and demonstrate a layered system. This implies that the vertical permeability barriers are areally extensive, and that they have a substantial impact on production predictions and development scenarios. The challenge for the subsurface team is to represent these thin barriers in a predictive simulation model. A standard Kv:Kh ratio is insufficient to mimic the observed production response. A very fine-scale geomodel could be considered but is computationally expensive. Vertical transmissibility multipliers are useful in a simulation model, but experiments demonstrate that the lateral extent of the barriers is discontinuous. Diagnostic plots of the well water/oil ratio (WWOR) versus the cumulative liquid production (WLPT) can identify groups of wells with layered production profiles. This information provides a basis to infer the potential for a layered production response from existing well logs. To introduce such discontinuities, a series of randomly generated vertical baffles have been created which provide an improved simulation response in a relatively coarse simulation model. Early water breakthrough, and excessive water production in some wells caused premature re-completions into other zones. This type of sub-surface modeling provides support for the introduction of intelligent completions to isolate specific layers as part of the development plan. This paper presents the techniques implemented by the team, the results to date, and a prognosis for the future of the field.
The use of proxy models for optimisation of expensive functions has demonstrated its value since the 1990's in many industries. Within reservoir engineering, similar techniques have been used for over a decade for history matching, both in commercial tools and in-house software.In addition to efficient history matching, proxy models have a distinct advantage when performing uncertainty quantification of probabilistic forecasts. Markov Chain Monte Carlo (MCMC) methods cannot realistically be applied directly with reservoir simulations, and even fast proxy models can fail dramatically to adequately represent the range of uncertainty if implemented without due care.A pitfall of the use of proxy models is that they are considered 'black box' and their quality is difficult to measure. Engineers prefer to deal with deterministic simulation models which they can evaluate and understand.The main pitfall of simple random walk MCMC techniques, which have begun to appear within reservoir engineering workflows, is a focus on theoretical properties which are not observed in practical implementations. This gives rise to potential gross errors, which are not generally appreciated by practitioners. Advances in recent years within the field of Bayesian statistics have significantly improved this situation, but have not yet been disseminated within the oil and gas industry. This paper describes the limitations of random walk MCMC techniques which are currently used for reservoir prediction studies, and shows how Hamiltonian MCMC techniques, together with an efficient implementation of proxy models, can lead to a more reliable and validated probabilistic uncertainty quantification, whilst also generating a suitable ensemble of deterministic reservoir models. Scientific comparison studies are performed for both an analytical case and a realistic reservoir simulation case to demonstrate the validity of the approach.The benefit of this methodology is to allow asset teams to effectively manage reservoir decisions using a robust and validated understanding of uncertainty. It lays the scientific foundations for the next generation of uncertainty tools and workflows.
A dry gas reservoir offshore North Trinidad has been on production since 2002 with a high recovery to date. Notwithstanding this maturity, significant volumetric uncertainty was suspected and this had the potential to alter the field development plan. Hence, a robust quantification of hydrocarbon volumes was deemed necessary through a study that rigorously captured risks and uncertainties and provided higher confidence in the predicted remaining reserves. Experimental design based uncertainty analysis, Markov Chain Monte Carlo for Bayesian optimization and proxy modelling assisted history matching methodologies were used. Scoping Latin Hypercube (LHC) simulations were run to test a wide range of volumetric scenarios. MCMC allowed Bayesian optimization to be done in an unbiased manner but being computationally expensive, a proxy modelling workflow was utilized. Top LHC cases were selected to train a proxy. The efficiency of the proxy workflow enabled hundreds of full field proxy simulations to be run in a short time. All experiments were combined and an optimum filtering criterion established to short list potentially matched cases. Finally, stochastic volumes were selected from this filtered pool. Due to the use of Bayesian algorithm, posterior uncertainty distributions were often different, and thus quite insightful, when compared to the prior distributions. The shape of prior distributions had limited impact on the final result. A limited number of LHC runs scoping a wide sample space were sufficient to train a robust proxy model. Acceptable history matches for all wells could be achieved just after the second MCMC cycle. LHC workflow allowed wider uncertainty sampling but typically poor matches, whereas MCMC proved to be a rapid & highly effective history matching technique but it tended to smooth out extremes volumetric results and converged towards a high confidence median. A key result was the observation of a very strong correlation between the "global" history matching error and field gas-in-place or recoverable volumes. Use of this Proxy based workflow allowed significant reduction in computational costs. A powerful cross-plot analysis methodology is presented that demonstrates convergence of hydrocarbon volumes to a narrow range. Results showed that irrespective of the deterministic starting inputs, reduction in history match error was only achieved for a narrow posterior range of uncertainties and volumes, even where a large initial uncertainty was believed to exist. These results were successful in achieving the objective by largely reducing the volumetric uncertainty and re-confirming the initial view.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.