StatMod MC yields far greater vertical detail than available from standard seismic inversions. It integrates high resolution well data with low resolution 3D seismic, and provides high vertical detail near and away from well control. StatMod MC generates reservoir models with geologically plausible shapes and provides a clear quantification of uncertainty to assess risk. Highly-detailed petrophysical models are generated, ready for input to reservoir flow simulation.
Realistic, Highly Detailed 3D Numerical Models of Rock and Reservoir Properties
Fugro-Jason StatMod MCStatMod™ MC combines geostatistics and advanced statistical physics with innovative seismic inversion methods to integrate disparate data from multiple sources and generate reservoir models that can be used to reliably quantify
uncertainty for risk assessment and reduction. StatMod MC goes beyond traditional geostatistics and seismic inversion to:
* Integrate high resolution well data with low resolution 3D seismic
* Improve the vertical detail over deterministic seismic inversions
* Produce reservoir property models with geologically-plausible shapes
* Quantify model uncertainty for scenario analysis and risk assessment
* Generate highly detailed petrophysical models ready for input to reservoir flow simulation
With StatMod MC, geologists, geophysicists and other geoscientists can build highly detailed realistic 3D numerical reservoir models with more accurate estimates of uncertainty and less bias.
Joint inversion of impedance and lithofacies
StatMod MC simultaneously inverts for impedance and discrete property types, or lithofacies, instead of taking a two-step approach as is done by conventional inversion and geomodeling algorithms. Not only does simulating lithofacies and impedance realizations in one sweep save time as opposed to doing so sequentially, it also enhances the accuracy of the results as significant synergies between the two can be leveraged during the inversion. Once such highly detailed models of impedance and lithofacies have been generated, any number of additional petrophysical properties (e.g. porosity) can be jointly cosimulated from them.
The way StatMod MC conceptually works is deceptively simple and consists of the following steps:
1. Statistical Modeling. Each source of input information (e.g., wells, cores, seismic) is represented in the form of a probability distribution function (PDF) characterized in geostatistical terms (histograms and variograms). The histograms and variograms are obtained from log analysis, rock physics modeling and geological insight. The histograms define the likelihood of different values at any given point, while the variograms give essentially the ‘characteristic scale’ and texture of the geological features in lateral and vertical directions.
2. Bayesian Inference. Bayesian inference techniques are used to merge these individual PDFs together and obtain a posterior PDF conditioned on all known and assumed information. This posterior PDF represents the overlap between all of the input PDFs—think of it as some sort of ‘evidence fusion’. The advantage of this approach is that the weight assigned to each input data source is automatically determined by the algorithm, thus removing subjectivity.
3. Inversion and Cosimulation. A customized Markov Chain Monte Carlo algorithm is used to obtain statistically fair samples from the posterior PDF. A fair sample in this case means volumes of rock and reservoir properties of interest (e.g., P-impedance, lithotype, porosity, water saturation). Because all of the input data is effectively inverted simultaneously, significant synergies can be exploited, thus producing models that are of greater detail, accuracy and realism than otherwise possible. The Geostatistical Inversion and cosimulation procedures are iterated until a model is found that matches all information, from geological expectations to well logs, seismic, and production history.
4. Uncertainty Assessment. Estimates of uncertainty are made by producing a series of slightly different realizations and scenarios. Different realizations are produced by repeating the above steps with different random seeds. Different scenarios are produced by targeting the uncertainty in the more sensitive parameters. Together, such analyses give an intuitive and accurate handle on development risk and uncertainty, given what is known about the subsurface.
Uncertainty
Geostatistical Inversion makes it possible to generate multiple predictions, each of which honors the known input information and is a plausible model of what the reservoir might look like. This is a significant advantage. Multiple plausible predictions provide an intuitive understanding of uncertainty associated with any given model.
An accurate understanding of uncertainty is critical for risk assessment, scenario analyses and sensitivity analyses. These evaluations depend on the ability to identify which data sources are most likely to reduce the overall uncertainty prediction and to properly weight redundant data. For example, well data is almost always preferentially clustered in higher pay areas. The redundancy of the information coming from such wells needs to be properly accounted for in order to avoid biasing the predictions. Once uncertainty is adequately captured, different realizations and scenarios can be ranked--only then can risk be evaluated and informed decisions be made.
Reply With Quote
Bookmarks