Operators are looking for tools and workflows that can deliver that crucial competitive edge and improve the accuracy of predictions of reservoir behavior in newer oil and gas fields. When it comes to subsurface geology, these trends converge at the reservoir model (Figure 1). Reservoir modeling is consistently shown to be the best means of integrating and understanding subsurface data and generating decision-critical information for the commercial production of hydrocarbons.

While reservoir modeling has improved hugely over the last few years, limitations do still exist – limitations that are becoming increasingly apparent as the oil and gas industry moves into more complex geological settings and more economically marginal prospects. Emerson is developing a new, flexible, and responsive workflow to meet the current and future challenges of reservoir modeling called model-driven interpretation. This workflow lets users capture and quantify geologic risk while building a reservoir model directly from the data.

What’s in a reservoir model?

Geo Improving Figure 1

FIGURE 1. A successful reservoir model represents the data and the structural framework, supports decision-making, and describes key risks and uncertainties that affect reservoir management decisions. (Images courtesy of Emerson Process Management)

There are four key elements reservoir models need to be able to deliver value for the operator. Reservoir models need to represent the data – whether they are derived from seismic images, property logs, regional geology, or even production data – providing a realistic depiction of the geometry and properties that impact fluid flow and volumes.

Reservoir models also need to be able to provide an accurate representation of the structural framework – the faults and geologic horizons in the reservoir – and an accurate tracking of heterogeneities and lithologies and how the different flow properties are distributed in the reservoir.

Reservoir models also must answer the right questions, providing an important tool for decision-making support that can make predictions and test hypotheses via a drilling campaign. This requires the tight integration of data and the ability for interpreters, modelers, reservoir engineers, and drillers to work together toward common objectives.

Finally, reservoir models need to describe key risks where uncertainties that affect reservoir management decisions are captured. It is an increased awareness of these risks – particularly early in the interpretation and model-building phases – that improves results throughout the reservoir life cycle.

What challenges do we face today?

Interpretation and reservoir modeling workflows are dated. They are conceptually identical to the pencil-to-paper methods used 30 years ago. Modern software has certainly improved and streamlined the workflow, but several philosophical limitations remain from the paper days. These include an increased reliance on a single model; the inability to account for ambiguity within the data; difficulties in quantifying uncertainties in static reservoir properties; and disjointed, “siloed,” and time-consuming workflows.

Conventional geophysical interpretation today remains geared toward producing a single model or scenario for the configuration of subsurface features – despite the data supporting many different interpretations. This can be a real lost opportunity for reservoir modeling, with many alternative hypotheses simply discarded.

Geo Improving Figure 2

FIGURE 2. Fault and horizon uncertainty envelopes allow the interpreter to accurately represent the limitations of the data and simulate an ensemble of plausible subsurface models.

There also is an inherent ambiguity in the data evaluated in many reservoir models today. The physical limitations in seismic acquisition technology result in only a portion of the earth response being captured in a seismic image, leading to uncertain estimates of horizon or fault locations. This ambiguity also increases rapidly as the interpreter moves away from control points such as well logs. This can result in many configurations or scenarios (fault configurations, for example) supported by the data but unable to be distinguished based on the data alone.

Another limitation is that uncertainties in static reservoir properties (for example, the structure or interpretation, depth conversion, fault model, or facies distributions) often are difficult to quantify in reservoir models, particularly in frontier areas where there is little well control. It is these properties that are the largest contributors to the commerciality of a prospect today, and yet it is currently difficult to quantify the uncertainties within them.

Finally, there are the disjointed and time-consuming workflows that characterize much reservoir modeling today. Too often workflows are segmented and siloed with time-consuming and resource-intensive iterations and quality control. With geophysicists interpreting thousands of points at seismic scale and geomodelers doing the best they can to fit the model to the interpretation, data and crucial decision-making information are often overlooked.

Solving the challenges

This requires a new, flexible, and responsive workflow that can handle multiple models, capture the limitations of the data, and quantify geologic risk as early in the reservoir modeling process as possible.

Model-driven interpretation accomplishes this by uniting geophysicists and geologists on a common platform where interpretation and uncertainty quantification take place in real time as the model is being built (Figure 2).

Rather than creating one model with thousands of individual measurements, the new model-driven interpretation workflow creates thousands of models by estimating uncertainty in the interpreter’s measurements. The software can then generate statistically significant ensembles of models based on these probability distributions and provide immediate value to geoscientists.

The interpretation is based on uncertainty information being collected and paired with an interpreted geologic feature (horizon, fault, contact, etc.), thereby more accurately representing the limitations of the data and the interpreter’s vision for the geologic structure. In this way the new method can show what parts of the model are most uncertain and can quickly indicate where more detailed investigation is needed and where new data need to be acquired.

Users also will receive instant feedback on the consequences of a measurement, with fewer points and clicks required to map a geologic feature. Through this, interpreters will be able to rapidly map the key features of the reservoir using a sparse representation with no additional quality-control phases.

Model-driven interpretation allows geoscientists to guide and update a 3-D geologically consistent structural model directly from the data.

Structural modeling algorithms also can be applied to interactively construct a geologically consistent model of the static reservoir with fault uncertainty added to simulate the position and geometry of faults in the model.

With horizon uncertainty workflows, users can generate structural model realizations that account for a wide range of uncertainty parameters, including horizon positions, zone logs, or even uncertainty in the velocity model. It is this ability to analyze the full range of structural uncertainties that enables geoscientists to create suites of model realizations that satisfy many external constraints from well picks to velocity uncertainty and horizon or fault positional uncertainty.

Finally, static bulk volumes of the reservoir can be computed. Using these volumes, interpreters can compute a posterior probability distribution to derisk the prospect and ensure valuable input to future field development planning.

Let the model drive

So what impact can this new workflow have in supporting commercial decision-making? First, the new workflow generates a more complete representation of the data irrespective of the quality of the data, with the uncertainty staying with the interpretation throughout the modeling workflow.

The workflow also can generate early estimates of reservoir volumes, enabling geoscientists to quickly build risked models of static reservoir volumes and generate the best possible estimates to support commercial decisions. Histograms and distributions of static reservoir volumes, for example, can be quickly constructed and analyzed, resulting in probability estimates directly used in financial modeling. Risk estimates for drilling decisions also can be generated through combining interpretation and structural uncertainty modules. Operators can thereby make risked predictions of horizon or fault positions and integrate this with LWD data and precision steering to reduce risk in drilling.

Emerson has recently published a white paper titled, “Five Benefits Operators Expect from Their Reservoir Models and How These Can Be Achieved.” To download a copy, visit marketing.roxar.com/LP=135.