First EAGE Integrated Reservoir Modelling Conference - Are we doing it right?

This presentation focuses on various aspects of how the results from stochastic, petro-elastic seismic inversions can be used in 3D static models. Scale issues can be handled in the inversion, in a downscaling step, or by working on a single scale. For an update to be consistent in both time and depth, the inversion should also do geologically meaningful structural updates. The uncertainty estimates provided by the stochastic inversion should drive the amount of variability in the static model(s).

We showcase a method to provide 4D seismic information in the reservoir engineering domain, making integration into AHM loops more straightforward whilst ensuring that the 4D signal, production data and geology are coherent at the well through calibration of the poro-elastic model. An example is shown on a field undergoing production and water injection.

Modeling of reservoirs has become more mathematically sophisticated over the last twenty years. In particular stochastic models have become the norm in very many, if not most reservoir studies. The reasons for this are threefold 1) The advance of stochastic reservoir modeling as a science exposed the shortcomings implicit in the older deterministic models, namely underestimation of the role of heterogeneity. 2) Stochastic methods provide the best mechanism to incorporate various data types into a model. 3) They allow for some understanding and modeling of the uncertainty present in the reservoir. We have arrived at a point in time where many of the standard techniques for modelling are available in all of the principal commercial software packages (which is not to say that the methods cannot and should not be improved further). It may therefore be a good time to ask the wider question about how they are being used in practice particularly with regard to the major issue of understanding risk. In this presentation we look at how closed mathematical models, while being a fundamental building block, do not lead, on their own, to a good methodology for understanding risk. An analogy with the mathematisation of risk in banking and the credit markets is noted. The conclusion, which is very obvious, is to increase the amount of speculative geological model building and peer to peer discussion about scenarios. However, the current practice of reservoir modeling provides stumbling blocks making users reluctant, or even unable to easily explore differing scenarios. Indeed this difficulty extends beyond heterogeneity modeling into geophysics, petrophysics etc. To change this is a big task, which we don’t pretend to have resolved here, but by shining light on the problem, and by making a small suggestion to improve matters for heterogeneity modeling, we hope to open discussion about the subject.

A new approach for making identical 3D Training Images (TI) for multipoint statistics (MPS) and object based reservoir models of complex geological analogues is presented. It bridges the current gap in bringing complex facies assemblages quickly into a reservoir model without losing the essential geometric shapes and geological associations. It also provides an opportunity to validate the analogue selection. Ultimately this method allows geologist to transfer outcrop data and conceptual models directly and effectively into reservoir grids, and to efficiently form geologically relevant complex reservoir facies models.

See Abstract

Proper sampling of plugs and upscaling are key steps in generating reservoir models. At core scale many carbonate reservoir rocks exhibit extreme variations in permeability which can frequently not be captured by conventional core plugging. Whole core studies, which investigate the dynamic behaviour at core scale are costly and time extensive. We present a set of 3D microblock models at decimeter scale, which aim to represent the full heterogeneity of carbonate reservoir rocks. The textures range from highly irregular algal mudstones through clast supported wacke- packstones to bioturbated mudstones. The structure of the depositional sedimentary elements at this resolution were mapped in cores and translated into the micromodels using various geostatistic techniques. In example multipoint geostatistics was used to model complex shapes like floating coral debris. Permeability was measured every centimeter using a minipermeameter. The core material was then plugged at centimeter scale and porosity was measured at the generated mini-plugs. The properties were mapped, geostatistically evaluated and simulated using a 3D modeling package. Applying conventional flow simulators in lab mode we study the two-phase flow behaviour of the modeled reservoir rocks and compare to analytical upscaling predictions. We can show how flow is dominated by high permeability streaks at core scale and can simulate under which conditions such behaviour is preserved at reservoir cell scale.

Light-UP (for Light Uncertainty Process) is the reference TOTAL in-house tool for uncertainties management. Light-UP provides (1) the assessment of the probabilized HIIP distribution thanks to the quantification of uncertainties around the Base Case model and (2) the methodology and tools to build “tailor-made” models that drive fluid flow simulation for reserves evaluation with or without dynamic uncertainties.

Reservoir forecasts tend to be optimistic. Forecasts for IOR/EOR projects tend to be particularly optimistic. Sources of the optimism can be divided into several broad categories including: Data – quantity, quality, sampling bias; Static Modeling – model complexity, particularly for permeability contrasts; Model parameter/algorithms choice; and, Dynamic Modeling – model detail/complexity, up-scaling, well location optimization. In addition, human factors also tend to drive projects towards optimistic forecasts. Based studies of a number of reservoirs representing a variety of lithology types and depositional environments with data densities ranging from low (greenfield) to extremely high (multi-pattern pilots) observations on modeling and forecast accuracy can be made relative to IOR/EOR forecast results, in particular. Among the most critical modeling parameters are the areal grid size and the semivariogram range parameter. Optimistic estimates of the in place hydrocarbon volume is also one of the most significant sources of optimistic forecasts. Some of this latter bias is due to sampling, particularly for green-field developments, and some due to inappropriate use of analogs. This bias can be reduced with uncertainty-based analyses and workflows and an appropriate suite of analogs. Well location optimization based on stochastic models is an under-appreciated source of forecast optimism.

In recent years, the petroleum industry achieved marked advances in the development of full 3D geomechanical models (both static as well as dynamic). If robust, these 3D models provide the benefit for more accurate well and field development planning in structurally complex settings such as areas with significant topography or/and faulting, or/and when substantial pore pressure changes in response to injection or/and depletion are likely to cause changes in the in situ stress field. These stress changes may then be accompanied by phenomena such as fault and fracture reactivation, reservoir compaction and surface subsidence etc. Furthermore, proper 3D geomechanical modeling is often required when wells are planned to intersect salt diapirs or other frictionless bodies/layers with different material characteristics. In many cases, these 3D geomechanical models also require advanced finite element modeling coupled with reservoir simulation. To maximize the value from such modeling, in particular understanding the in situ stresses, pore pressure, and material properties and their variability (lateral, in depth as well as with time) are of critical importance to better evaluate and mitigate risk during the drilling, completion, and production phase of a field development. In other words, a sound and accurate 3D structural model is critically important, since it not only covers the structural elements and layering (stratigraphy) of the reservoir but also the overburden all the way to the surface (or seafloor in offshore environments); it thereby provides the basis for a 3D geomechanical model. When combined with a well calibrated fracture network model (e.g., DFN), 3D geomechanical models can be utilized to investigate the possibility for natural fractures to be or become critically-stressed in response to production or injection and evaluate the impact on the reservoir-wide permeability tensor as a function of time. This presentation will provide an overview of the basics of 3D static as well as dynamic geomechanics and how it is linked and integrated with structural modeling, discrete fracture network models as well reservoir simulation. Current model limitations and a future outlook as to where technology develop is and should head is also included in the discussion.