Robust (in terms of describing “what is there”) and reliable (in terms of reservoir predictability) reservoir models are crucial elements of the reservoir characterization workflow. If these models fail, decisions based on them are likely to cause reduced field productivity and in the worst-case scenario, misplaced wells.

One of the key industry challenges currently facing reservoir engineers is updating an existing model to maintain model reliability whenever new data are acquired. Another challenge is to determine how to efficiently leverage all available data throughout the modeling process.

One key element of this is the integration of seismic data. Whether for exploration or production purposes, integration often is harder to achieve than it needs to be.

Thickness information is taken from wells or seismic attribute maps. (Images courtesy of Emerson Process Management)

As more and more seismic data become available, along with advanced techniques for extracting geological information from it, it is increasingly important for modeling software to be able to integrate these data quickly and easily.

The importance of property modeling
Property modeling is one area where seismic data can be combined with other data such as well data to generate accurate and well-constrained reservoir models.

Using seismic in property modeling, however, is not straightforward. Scale can be a challenge. The resolution of the seismic data normally is too coarse to resolve the true geometries of the heterogeneities, and uncertainty in depth can make it difficult to represent wells correctly. It is an uncertain and less well-understood area below the limits of seismic resolution but above the optimum scale for probabilistic modeling.

To meet this challenge, a new object-based facies modeling tool has been developed as part of Roxar RMS2010. The new tool – Sedseis – can incorporate information derived from seismic directly into the facies model, bridge the gap between the seismic scale and the scale relevant to stochastic techniques, and ensure seismic data are integrated quickly and easily into the reservoir model.

This is achieved by blending data extracted from seismic with geostatistical tools such as guide lines and trends to generate well-constrained sedimentary bodies. This ability to bridge the gap between both deterministic and statistical techniques gives the modeler access to the gray area between seismic resolution and data-constrained statistical modeling, resulting in a more realistic property model conditioned to well observations and with accurate volume calculations.

The nature of the model varies depending on the quality of the seismic data and the degree of the detectability of the sedimentary body. For example, with high levels of detectability, the whole seismic body can be extracted with upper and lower bounding surfaces and a complete definition of the body geometry. However, when this level of seismic data is not possible, the new tool also can incorporate thickness maps derived from seismic attributes or even from as little information as a polygon defining a lateral body position.

Since the uncertainty of the model increases as the seismic resolution decreases, the extraction of seismic on a sliding scale of detail allows the extension of more certain models into previously unconstrained scales of modeling.

This new technique is ideal when sand bodies can be identified from seismic attributes. The location of the sand objects can be defined simply by digitizing a polygon, using a set of points defining the “geobody,” or using a top and base map. The vertical shape of the objects can be described using the same tool used for object modeling with thickness information taken from wells or from seismic attribute maps.

Improved visualization capabilities aid the interpreter in making better decisions.

The extracted data from this new object-based facies modeling tool is used as input to stochastic object modeling techniques, supplying a deterministic constraint on the stochastic process. The result is higher resolution, a far more constrained and reliable model, and the efficient leveraging of seismic data.

Multipoint statistics
A second addition to the property modeling toolbox and another facies modeling tool for the reservoir modeler is a multipoint statistics (MPS) technique. Using a pixel-based (grid cell by grid cell) approach for building stochastic facies realizations, MPS allows the reservoir engineer to condition 3-D training images of the interpreted heterogeneities in the reservoir in addition to wells and seismic.

While other approaches might struggle as the number of wells increases and the quality of the seismic data is improved, MPS actually improves in performance.

A facies probability function has been generated based on a seismic attribute and well observations with the shape of the objects taken from a 3-D training image. The lower part shows the results as objects are located where the probability is high, the shapes of the sand bodies are realistic, and the wells are represented correctly.

New ways of storing, visualizing seismic data
New ways of storing and visualizing seismic data also need to be developed for high-quality seismic data to become an integral element of the reservoir modeling and characterization workflow, and for the existing model to be easily updated whenever new data are required.

Efficient handling of seismic data will not only provide new ways for performing quality control on the structural models but also will open up new functionalities and workflows that significantly speed up the modeling process and the quality of the model.

With this in mind, the latest module comes with improved functionality for importing, visualizing, and sampling seismic data to and within 3-D grids. Enhancements include the ability to handle large seismic cubes and the seamless importing of both 2-D and 3-D seismic data.

MPS is used on high-quality seismic data.

During the importing, data are transformed into a compact representation that is optimized for visualization and calculations while preserving the precision of the original data. Furthermore, the advanced compression when loading means there is no need to clip and scale the data. Once the data have been loaded and the project has been saved, the reservoir modeler does not require access to the original SEG-Y files.

There also are vastly improved visualization capabilities when comparing the two models.

In addition, the new seismic module provides fast and accurate visualization of seismic datasets. Options include the ability to generate probes of any shape – slices, boxes, or tubes – as well as probes from multiple sets of seismic data and other data types such as 3-D grid parameters and velocity models. Furthermore, fully interactive opacity control and color manipulation capabilities provide many possibilities for inspecting data.

Powerful seismic tools also are available for converting both grid parameters and velocity models to volume data, allowing the interpreter to visualize any grid parameters and velocity models for further use in property modeling. Data also can be converted from time to depth in one step using either a velocity model or a velocity cube as input.

Finally, throughout the reservoir modeling process, the interpreter can go back to seismic data at any time and verify that correct decisions were made.

It is this quick and easy integration and leveraging of seismic that ensures robust and reliable models, improved decision making, more justifiable investments, and improved field performance.

Comments