Building Reservoir Models Based on 4D Seismic & Well Data in Gulf of Mexico Oil Fields

Abstract

Abstract A general methodology has been developed for building and history matching reservoir models based on 4D seismic, petrophysical, and production data. The methodology consists of: construction of structural maps of different reservoir zones; building 3D facies models; geostatistical construction of 3D volumes of reservoir properties (porosity, net-to-gross ratio, permeability) constrained by seismic and well log data; construction of fluid flow barriers; and matching of seismic attribute changes over time. A procedure has been developed for building confining surfaces of the reservoir zones matching well markers and velocity maps. Object based modeling tools have been applied for the construction of the 3D facies models constrained by seismic and well data, and consistent with depositional environment. The construction of the 3-D volumes of the reservoir properties is based on latest achievements in rock mechanics - sim2seis correlations, which correlate seismic attributes (for example, acoustic impedance AI) to porosity, net-to-gross ratio, pressure, water & gas saturation and rock properties. The procedure has been developed to calibrate sim2seis correlations based on well log and pressure data. The procedure is applied for effective porosity definition, in two parts. First, correlating 3D seismic AI with well logs averaged over the corresponding model cells derives an effective porosity trend. Second, geostatistical modeling tools are applied to define the 3D effective porosity volume constrained by well log data and the effective porosity trend. A new procedure has been developed for the construction of the fluid flow barriers from 4D seismic. The developed methodology has been applied in several Gulf of Mexico oil fields. It has been demonstrated that: a reasonable history match can be achieved without any adjustments of the reservoir models, and that fluid flow barriers derived from 4D seismic significantly improve history matching results. A good match of predicted results and field measurements from recently drilled wells (suggested by reservoir modeling) has demonstrated the prediction capabilities of the constructed reservoir models. Introduction In this paper we describe methodology that has been applied in several deepwater GOM (Gulf of Mexico) turbidite channel oil fields, focussing on one particular field example. In most of these cases, distances between wells are significantly larger than lengths of geological features. For this reason, it was very important to use seismic data to predict changes of rock properties (net-to-gross ratio, porosity, and permeability) between wells. Well log and core data are considered as hard data that should be matched exactly, and seismic data are applied as soft data (as a guidance) for the rock property interpolation between wells. Confining surfaces of the reservoir zones and seismic attributes are usually interpreted in two-way reflection time space (vertical units of seismic time). Intersections of well trajectories with the confining surfaces (well markers) can be derived from well logs. We describe a procedure for building the confining surfaces in depth space matching well markers and constrained by corresponding surfaces in the time space as well as velocity maps. It is important to model shale barriers that cannot be resolved with seismic data but can be observed in well logs. We apply geostatistical facies modeling tools to represent the shale barriers in the reservoir models. Effective porosity and/or net-to-gross ratio can be derived from seismic attributes (for example, acoustic impedance) using geophysical rock property modeling techniques, but these cannot be used directly in reservoir simulation models for two reasons. First, they do not exactly match values of the effective porosity and net-to-gross ratio derived from well log data. Second, there are large uncertainties in seismic attributes and parameters of correlations used to derive rock properties from seismic attributes. For these reasons, we use the rock properties derived from seismic as trends. Then, we use geostatistical modeling to match well data constrained by these trends and take into account uncertainties in seismic attributes.