An Integrated Reservoir Characterization Study Matching Production Data and 4D Seismic
- Vincent Kretz (Institut Français du Pétrole) | Mickaële Le Ravalec-Dupin (Institut Français du Pétrole) | Frederic Roggero (Institut Français du Pétrole)
- Document ID
- Society of Petroleum Engineers
- SPE Reservoir Evaluation & Engineering
- Publication Date
- April 2004
- Document Type
- Journal Paper
- 116 - 122
- 2004. Society of Petroleum Engineers
- 5.6.4 Drillstem/Well Testing, 5.5.8 History Matching, 5.1.5 Geologic Modeling, 5.2.1 Phase Behavior and PVT Measurements, 5.1 Reservoir Characterisation, 4.1.2 Separation and Treating, 5.5.2 Construction of Static Models, 5.1.9 Four-Dimensional and Four-Component Seismic, 1.2.3 Rock properties, 5.6.9 Production Forecasting, 5.1.8 Seismic Modelling, 5.5 Reservoir Simulation, 3.3 Well & Reservoir Surveillance and Monitoring, 4.1.5 Processing Equipment
- 0 in the last 30 days
- 433 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 12.00|
|SPE Non-Member Price:||USD 35.00|
Our purpose is to evaluate the benefit of using time-lapse seismic in addition to production history for reservoir characterization. A new inversion methodology, based on the gradual deformation method (GDM), has been developed to simultaneously integrate different sources of information like 4D-seismic-related data and production data. This methodology was validated successfully on a geologically realistic 3D synthetic reservoir model.
An optimization procedure was developed to identify reservoir models consistent both with an initial 8-year period of production data and with 4D-seismic-related data. The procedure is based on two complementary numerical tools developed at the Inst. Français du Pétrole: the Fast Fourier Transform - Moving Average (FFT-MA) generator and the GDM. The first tool generates stochastic realizations for the reservoir model; the second tool allows us to perturb the realizations from a small number of parameters while preserving the spatial-variability model. The optimization process was repeated starting from different initial reservoir models. The predictive quality of the constrained reservoir models was estimated by comparing the simulated forecasts and the reference data for a second 8-year period of production data.
The proposed inversion method proved its efficiency for constraining geostatistical reservoir models to different sources of data. It is shown that conditioning reservoir models to saturation changes in addition to production data results in significant improvement in the reliability of production forecasts. The present study goes one step further than the study by Kretz et al.,1 which was dedicated to the integration of accurate gas-saturation data into the reservoir model. Because such accurate data are rarely available, we focus on a more realistic case: we estimate the usefulness of binary data, indicating the presence or absence of gas.
One of the most challenging fields in reservoir engineering is the integration of all available data for the characterization of reservoirs and the reduction of uncertainties in oil and gas production. In this paper, we focus on the identification of permeability and porosity distributions in heterogeneous and multiphase petroleum reservoirs by matching the observed dynamic behavior. During the field life, a large quantity of data, called dynamic data, is collected. Dynamic data traditionally involve production history, water cuts, and gas/oil ratios. Since the late 1990s, they also have consisted of 4D seismic data (i.e., repeated 3D seismic acquisitions). 4D seismic is a potentially powerful source of data for reservoir monitoring because it provides information over large areas. As long as data are collected and processed adequately, differences between the seismic data sets collected at successive times should provide information about the spatial distribution of saturation and pressure changes caused by fluid production or injection. Saturation values can be derived from the seismic data using Gassmann-based2 formulations or calibrated relationships.
A great deal of work has been dedicated to the conditioning of reservoir models to production data (for example, see Refs. 3 through 8). A few papers are also referenced in the literature for the use of 4D-seismic-related data in addition to traditional production data.
Basically, two kinds of approaches can be distinguished:
The first approach (Fig. 1) uses inverted 4D seismic data.9-13 The seismic data are inverted to directly point out changes in fluid saturation, impedance, or elastic properties between the surveys. Then, the inversion of fluid-displacement data needs only one forward simulator: a fluid-flow simulator.
The second approach (Fig. 2) does not use any kind of inverted seismic data14-16: the seismic data are just another set of matching data. Thus, the inversion process includes not only a fluid-flow simulator but also a forward seismic-wave-propagation simulator.
Our purpose is to evaluate the benefits of using inverted time-lapse seismic in addition to production history for reservoir characterization. For the sake of simplicity, we consider saturation data rather than the raw seismic cube. Saturation data are considered as the results of an inverted time-lapse seismic survey. We assume that seismic variations are caused by saturation changes only. The potential influence of pressure is beyond the scope of this paper, but it is the subject of ongoing research. In Ref. 1, we developed a matching methodology to build reservoir models constrained to production data as well as gas-saturation data. Because accurate gas-saturation data cannot be obtained easily, we focus now on low-level 4D seismic-related data. These data are indicators stating the presence or absence of gas. We assume that this information level can be reasonably provided by the present seismic techniques. Thus, we aim to build reservoir models constrained to fit production data and gas-presence indicators.
Dynamic data are used to condition reservoir models through a process known as history matching. The issue is posed as an inverse problem: the model parameters are adjusted iteratively so that the dynamic behavior simulated for the reservoir model fits the observed dynamic data. In the following sections, we introduce a new methodology and present an application case.
|File Size||1 MB||Number of Pages||7|