Introduction: Integrated reservoir characterization and modeling
Chapter DOI: 10.2516/ifpen/2014001.c000
The concept of reservoir model is not really new. Engineers have long used mathematical models when, for instance, calculating the oil in place. A reservoir model was then considered as a box with average values for porosity, saturation, thickness... The development of modern digital computers drastically changed this approach. Today, models are becoming increasingly complex with millions of grid blocks. Within this context, a key point is to integrate and reconcile all available data - geological, geophysical and production data - into reservoir models. Creating asset teams with geologists, geophysicists and reservoir engineers is a first step towards integration, but it is not enough. There is a clear need 1) for integrated modeling workflows focusing on a single objet, i.e., the reservoir model, and 2) for advanced techniques to adjust the reservoir model at any level of the workflows and strengthen its consistency with data.
Chapter 1: Uncertainty and reserve estimates
Chapter DOI: 10.2516/ifpen/2014001.c001
Characterizing underground geological formations is a tedious task for at least three reasons: geological heterogeneity, lack of data and measurement errors. Therefore, the description of such formations is uncertain. The identification of uncertainty sources and the understanding of the way they impact the volumetric estimates is essential to evaluate risk and make good decisions. This first chapter provides the basics to address this issue.
Chapter 2: Geological modeling - Geostatistics
Chapter DOI: 10.2516/ifpen/2014001.c002
Chapter 2 focuses on the modeling of the spatial distribution of petrophysical properties (facies, porosity, permeability, etc.) within reservoirs. This is performed within a stochastic framework since there is not enough data to map the exact distributions. We first recap the basics of geostatistics. Then, we show how to create possible images representative of the spatial distribution of petrophysical properties into the reservoir using either estimation or simulation. Estimation consists in interpolating the values of the property under consideration at unsampled locations from the observation of its values at nearby locations. It yields a single, statistically "best" estimate and a map of uncertainty. On the other hand, simulation entails the generation of outcomes or realizations of a random function. Each realization is considered as a possible representation of the spatial distribution of the petrophysical properties. The analysis of the differences between these realizations permits to quantify uncertainty.
Chapter 3: Static data integration - Examples
Chapter DOI: 10.2516/ifpen/2014001.c003
Chapter 3 deals with the integration of static data (measurements on cores, logs, seismic, etc.). The purpose of data integration is to compel the randomly drawn realizations to honor data at locations where there are data. This chapter introduces the types of data considered, the building of the grid supporting the geological model and various techniques developed to integrate data. A few case studies are also presented to illustrate how static data are incorporated into reservoir models. The first example describes an integrated workflow aiming to monitor the growth of a steam chamber in a field operated by steam-assisted gravity drainage from 4D-seismic. The second one emphasizes the ability and limits of seismic for discriminating geological objects. The last example shows a specific workflow developed to take the most from high-quality seismic data.
Chapter 4: Dynamic data integration – History-matching
Chapter DOI: 10.2516/ifpen/2014001.c004
Geological modeling comes up with a large number of geological models constrained to all available static data. Some of them are provided to a flow simulator to predict oil production. Then, the compilation of the computed production responses yields the range of uncertainty in the prediction and this range can be quite large. It is essential to narrow this range, hence to reduce uncertainties, by accounting also for dynamic data (production and 4D-seismic). This problem is well known as “history-matching”: the purpose is to determine reservoir models matching the production history. We emphasize why history-matching is a crucial step complementary to static data integration and provide the basics for modern history-matching.
Chapter 5: Consistency between geological and reservoir modeling
Chapter DOI: 10.2516/ifpen/2014001.c005
The petrophysical properties populating the geological model are very important parameters in terms of fluid displacement: they strongly influence the efficiency of sweeping or production. Unfortunately, there is only little information to describe how they spatially vary. These parameters are very specific for at least two reasons: they include a huge number of unknown values and they have a spatial structure. A key component to obtain a more accurate description of their spatial distributions is still history-matching. However, how can we adjust a permeability realization (or any realization of another petrophysical property) when there are so many unknowns and how can we make sure that the adjusted realization still respects the mean, variance and variogram? This is the issue addressed in this chapter, which presents distinct parameterization (or re-parameterization) techniques developed to make the history-matching problem manageable.