Joint stochastic inversion of pre‐stack seismic data and well logs for high‐resolution reservoir delineation and improved production forecast

Abstract

Integrated reservoir characterization makes use of different varieties of data to construct detailed spatial distributions of petrophysical and fluid parameters. The benefit of data integration is the generation of consistent and accurate reservoir models that can be used for reservoir optimization, surveillance, and management. This paper describes a novel strategy for the static and dynamic characterization of hydrocarbon reservoirs based on the extensive use of 3D pre-stack seismic data, wireline logs, core data, geological information, and time records of fluid production measurements. A stochastic simulation procedure is used to extrapolate petrophysical variables laterally away from wells subject to honoring the existing 3D pre-stack seismic data. This procedure yields highresolution estimates of inter-well petrophysical parameters (and of bulk density, compressional-, and shear-wave velocities as a by-product) associated with pre-stack seismic data. A numerical study in two dimensions is performed to evaluate the estimation algorithm applied to pre-stack seismic data, as well as to assess different strategies for the direct estimation of petrophysical properties related to elastic parameters. The same numerical study is used to quantify consistency of the estimated reservoir parameters with the time record of fluid production measurements. The inversion algorithm is CPU intensive and is based on a global optimization technique. Examples of applications show that the inversion algorithm lends itself to accurate estimation of petrophysical properties, such as porosity, that honor both the pre-stack seismic data and the well logs. Depending on the number of wells and the distance between them, the inversion algorithm can produce estimates of inter-well petrophysical properties with a resolution midway between that of seismic data and well logs. Models generated with this inversion scheme yield highly accurate predictions of reservoir dynamic behavior when compared to predictions performed with standard geostatistical techniques.