Analyzing Multiple Realizations
of Structural Complexity
Brian Lynch, Jane Wheelwright, Pascal Luxey and Graham Brew
Dynamic Graphics Ltd, Wokingham, UK.
Calculating Multiple Realizations
As exploration and production targets get ever more complex, so does the potential uncertainty of our structural models. Hence a realistic reservoir characterization might involve a suite of potential structural realizations, rather than a single deterministic model. Variations between the realizations could include iterations of fault and horizon positions, fluid contacts, reservoir properties, and velocities (Figure 2).
Figure 1
Figure 2
Algorithms under development allow for the parameterization and calculation of large suites of nominally equi-probable realizations as variations on a single base case. Increasing desktop processing power, and the proliferation of Linux clusters, allow for the generation of hundreds or thousands of these structural realizations within reasonable timeframes. Cellular output from each realization can be run through reservoir simulation for history matching and validation, and new visualization software permits these realizations to be analyses and interrogated on the desktop. Furthermore, volumetric calculations and sensitivity analyses of these complex models can be performed on each realization—thus provided comprehensive risk assessments.
This article describes the uses of a small set of examples of these multi-realized models.
Calculating Volumetric Distributions
One of the most powerful applications of the multiple realizations is their use in calculating the range of potential volumetric distributions. Algorithms under development allow the calculation of bulk rock volume, or property volumetrics, from each of the multiple realizations. These results can then be analysed to see the effect of the structural uncertainty, and the sensitivity of the volumetrics to the different uncertainty parameters.
In Figure 3, the bulk rock volume of a fault block is investigated in the framework of uncertainty. Parameterized freedom is given to the positions of the fault, the oil-water contact, and the reservoir stratigraphy. The results clearly demonstrate that the resulting volumes are most sensitive to the position of the bounding fault.
Figure 3
Capturing Multiple Realizations
The calculation of multiple realizations is only the beginning of the analysis. Depending on the total number of realizations, different approaches are taken for the investigation of the effects of uncertainty. One obvious avenue of investigation is volumetrics (discussed below). But the validity of the structural realizations is best investigated through visualization of the multiple realizations.
These visualization are best appreciated on a live video monitor. However, Figure 5 gives a flavor of viewing a small number of realizations. In this example each of the picks in the well data has a degree of positional uncertainty. This uncertainty is parameterized and modeled, resulting in 125 structural realizations. The ″envelope″ of the movement of each of the structural horizons is captured by the ″trailing dots″ that are output from each realization as they animate across the screen.
Figure 4
Relatively small numbers of realization can be investigated through a sequence of animations, such as in Figure 5. However, for larger numbers of realizations it helps to capture the effect of the uncertainties into a single display. Figure 4 shows the summary of several hundred realizations of a model in which the position of the primary horizon, the position of the fault and the rotation angle of the channel are given parameterized uncertainty. The cellular grids are colored according to the probability of one of the structural elements being present in that locality (assuming that each realization is equi-probable). The vast majority of cells have a zero probability of having these elements present, as such these cells have been tuned off in this display. The image gives a fast summary of the effects of the uncertainty, and could be invaluable in well planning and risk assessment.
Figure 5
Visualizing Cross-Disciplinary Data
An important component when investigating uncertainty is access to all available and relevant information. Drilling and development decisions that involve the entire multidisciplinary asset team should be made with access to all the available multidisciplinary data. A visualization software environment allows the rapid and easy visualization of many different subsurface datasets including seismic, reservoir simulations, well data and logs, production data and structural interpretations. Temporal data can also be visualized and animated through time. These data can be displayed together with multiple structural realizations to fully investigate the subsurface uncertainty.
Figure 6
Although best seen using live computer software, Figures 6 and 7 offer a glimpse of the types of data juxtaposition that are available using the visualization environment on the desktop. The software allows the multidisciplinary team to make better, faster decisions.
Figure 7
Multiple Property Models
The multiple structural realizations calculated by the methodology described herein can be extended through property modeling. Uncertainty is inherent in any property modeling owing to the sparsely distributed ground truth (usually well data). Combined with the uncertainty in structural frameworks, this results in numerous degrees of freedom in the property model.
Below are a very small number of property realizations from a much larger set of models that investigated the effects of these uncertainties. The influence of even minor variations would have a dramatic affect to the outcome of the planned wellbore.
Multiple Property Model 1
Multiple Property Model 2
Multiple Property Model 3
Multiple Property Model 4