Pre-spending asset-team models point to correct development decisions.

In today's global business environment, exploration and production (E&P) asset teams are making increasingly challenging and costly decisions earlier in the life cycle
of oil and gas reservoirs. Accurate modeling of field development scenarios before making significant capital investments can optimize hydrocarbon extraction and yield tremendous savings of both time and money. Advancements in software technology and the availability of high-performance computing at increasingly lower prices are revolutionizing traditional reservoir management workflows.
Traditional serial workflows
Historically, reservoir simulation and associated workflows have been time- and compute-intensive, forcing asset teams to choose between more thorough analyses and acceptable turnaround times. Most complex reservoir problems have been solved by approximation, sacrificing some degree of fidelity in one or more modeled domains.
Traditional reservoir management workflows have consisted of serial processes in which each technical domain hands off its version of the earth model to the next specialist in a highly siloed manner. This familiar process has been especially cumbersome due to the inherent difficulty of transferring digital data and models between disparate software applications. Tools and technologies available to date have focused primarily on technical rigor within each discipline rather than support for integration among multiple domains.
As a result, each discipline tends to build the best model possible within its particular silo instead of optimizing the entire asset to meet strategic business objectives. Disjointed, multistep linear processes prevent asset teams from evaluating a sufficient number of development scenarios or fully capturing all the inter-dependencies in the value chain, creating greater risks.
Today, however, a next-generation reservoir modeling and decision management framework (Figure 1) is enabling teams to solve more complex problems in shorter timeframes without sacrificing accuracy. How? Through more consistent, highly iterative, multidiscipline workflows built on advanced technologies that preserve high resolution where necessary and integrate subsurface with surface domains under conditions of uncertainty. This new framework can be used for optimization as well. It enables the use of grid computing, cluster technology and workflow automation, and it provides open interfaces to a variety of engines at levels of rigor appropriate for the problem at hand.
Next-generation workflows
Let's briefly consider key parts of next-generation reservoir management workflows.
Earth modeling. Historically, asset teams may have included one or two static geological models in their reservoir simulation efforts. With current technology they can carry numerous equally probable earth models - from tens to hundreds - all the way from seismic to simulation. Today's models often have sufficient detail to accurately match the geoscientist's concept of the reservoir. Static models with 100 million cells are not unheard of. However, even with the newest and most sophisticated simulators, it can be difficult and time-consuming to apply uncertainty and/or optimization techniques to models of this size.
For this reason, next-generation workflows must be flexible enough to accommodate different earth modeling approaches, including:
• creating "scalable" models, in which size and resolution depend on specific problems to be solved;
• sizing models based on finishing complex calculations in set timeframes (often overnight), based on available compute power available, without upscaling;
• building the largest possible earth model and upscaling to the appropriate size; and
• building large models without upscaling.
Upscaling. An averaging process designed to reduce model resolution while maintaining critical geologic features, upscaling often loses extreme data values and combines cells that impact fluid flow. Reservoir simulation professionals, therefore, would prefer simulating on the scale of the full earth model, either without upscaling or through more "intelligent" upscaling.
Next-generation gridding technology is capable of preserving original geologic detail where needed (such as around channels, faults and well locations) while coarsening areas where there are no hydrocarbons (such as shales), maintaining the continuity of high- and low-permeability features (Figure 2).
Another innovative way of reducing simulation model size is automated lumping of layers based on petrophysical attributes in the vertical direction, which eliminates the laborious and often trial-and-error task of determining which layers to lump.
Flow modeling. Reservoir simulation models traditionally stop at the sandface or the tubing head, and results are handed off for surface network and facilities calculations. This method is no longer sufficient for today's complex situations where, for example, a common surface infrastructure will be used to develop multiple fields. It is critical to avoid over-designing or under-sizing production facilities while determining the optimal number and timing of wells per field. To properly estimate facilities performance requires a simulation methodology that incorporates tightly coupled, simultaneous modeling of both the surface network and subsurface reservoirs within a single calculation. With next-generation simulation technology, the entire surface/ subsurface system (Figure 3) can be seamlessly simulated to perform these types of calculations. Reservoirs can easily be added to the system as new fields are brought on production.
Twenty years ago, most wells could be modeled as simple vertical tubes. Today, complex directional wells with potentially large extended reach, multilaterals, and downhole measurement and control devices require much more sophisticated models than first-generation simulators can handle. Next-generation simulation technology treats any wellbore as an extension of the surface network, with fully implicit formulation in most cases. As such, network components such as downhole valves and chokes can be modeled, accommodating "smart wells" and multilaterals.
First-generation simulators also tend to lose efficiency as earth models grow in size, incorporating numerous faults, pinch-outs, unconformities, inactive cells and other features known as "non-neighbor" connections. They also have difficulty with intelligently upscaled models, which include connections between grid blocks of different resolution.
New simulation technology, however, which uses an unstructured formulation, thrives on such complex models while delivering exceptionally reliable results in excess of five times faster than current commercial simulators.
Integrated optimization
Prior to the development of the next-generation reservoir and decision management framework shown in Figure 1, there was no way to fully integrate uncertainty and optimization techniques with rigorous reservoir modeling, flow simulation and economics to explicitly manage risk and achieve business objectives.
As part of the decision-making process, uncertainties are parameters that cannot be fully controlled - for example, geological realization, property distributions and fluid contacts within the earth model; or costs, future operating and capital expenses, and oil and gas prices within the fiscal model. Uncertainties, therefore, are often modeled by continuous or discrete probability density functions. Sound decision-making also requires the ability to comprehend multiple, distinct development scenarios and the range of uncertainties present within each one.
There are, however, certain decision variables that can be controlled - such as facilities size, number and location of wells, and time and manner in which fields are produced. These can be varied during the optimization process until key business objectives are achieved. This does not mean optimization always converges to a single mathematical or "provable" optimal solution. The goal is to progressively improve the objective by finding "better" solutions. It may be necessary to identify and carry forward multiple solutions to minimize the impact of uncertainty.
An integrated reservoir and decision management system might also have to honor certain constraints, for example, maximum rates or process capacity, which would be honored either in the flow simulator or other engines that make up the workflow. Other constraints and requirements, like risk tolerances, would need to be calculated and specified during formulation of the optimization problem.
The framework described in this article has been used successfully with spreadsheets and various commercial applications for flow simulation, integrated performance modeling, etc. For uncertainty analysis, it enables multiple efficient ways to sample and identify key uncertainties that drive the overall project risk and economics.
With decision trees or representative case studies, solutions are not driven by the objective or potential risk. E&P problems typically have such a large number of alternatives that one cannot simply search exhaustively for solutions, particularly when uncertainties exist. What is needed is a flexible approach that accounts explicitly for uncertainty by requiring that the objective meet a requirement on its statistical risk using a global optimizer. A simulation-optimization approach that uses a global stochastic search algorithm is available in this next-generation framework. The optimization solver used here employs standard and heuristic global search methods ("metaheuristics"). Other optimization engines can also be incorporated into the same framework.
With these capabilities, the framework can be used effectively for optimizing development planning or matching historically observed production data.
Solutions to changing needs
The oil and gas industry today continues to change in order to meet the growing complexity of technical and business challenges. Asset teams need greater speed to reduce cycle times, while E&P organizations are looking for streamlined processes and cost reduction. Because reservoirs are more complex and more difficult to find, decision-makers require better understanding of the uncertainties that are most likely to impact project economics.
All of these changes require greater cross-discipline integration and accuracy. To fully manage the risk from uncertainties and to make optimal decisions, reservoir management workflows must be iterated multiple times in shorter timeframes. Next-generation technologies and methods are emerging to provide robust solutions to today's challenges.