QUANTITATIVE METHODS FOR RESERVOIR CHARACTERIZATION AND IMPROVED RECOVERY: APPLICATION TO HEAVY OIL SANDS

Improved prediction of interwell reservoir heterogeneity has the potential to increase productivity and to reduce recovery cost for California's heavy oil sands, which contain approximately 2.3 billion barrels of remaining reserves in the Temblor Formation and in other formations of the San Joaquin Valley. This investigation involves application of advanced analytical property-distribution methods conditioned to continuous outcrop control for improved reservoir characterization and simulation, particularly in heavy oil sands. The investigation was performed in collaboration with Chevron Production Company U.S.A. as an industrial partner, and incorporates data from the Temblor Formation in Chevron's West Coalinga Field. Observations of lateral variability and vertical sequences observed in Temblor Formation outcrops has led to a better understanding of reservoir geology in West Coalinga Field. Based on the characteristics of stratigraphic bounding surfaces in the outcrops, these surfaces were identified in the subsurface using cores and logs. The bounding surfaces were mapped and then used as reference horizons in the reservoir modeling. Facies groups and facies tracts were recognized from outcrops and cores of the Temblor Formation and were applied to defining the stratigraphic framework and facies architecture for building 3D geological models. The following facies tracts were recognized: incised valley, estuarine, tide- to wave-dominated shoreline, diatomite, and subtidal. A new minipermeameter probe, which has important advantages over previous methods of measuring outcrop permeability, was developed during this project. The device, which measures permeability at the distal end of a small drillhole, avoids surface weathering effects and provides a superior seal compared with previous methods for measuring outcrop permeability. The new probe was used successfully for obtaining a high-quality permeability data set from an outcrop in southern Utah. Results obtained from analyzing the fractal structure of permeability data collected from the southern Utah outcrop and from core permeability data provided by Chevron from West Coalinga Field were used in distributing permeability values in 3D reservoir models. Spectral analyses and the Double Trace Moment method (Lavallee et al., 1991) were used to analyze the scaling and multifractality of permeability data from cores from West Coalinga Field. T2VOC, which is a numerical flow simulator capable of modeling multiphase, multi-component, nonisothermal flow, was used to model steam injection and oil production for a portion of section 36D in West Coalinga Field. The layer structure and permeability distributions of different models, including facies group, facies tract, and fractal permeability models, were incorporated into the numerical flow simulator. The injection and production histories of wells in the study area were modeled, including shutdowns and the occasional conversion of production wells to steam injection wells. The framework provided by facies groups provides a more realistic representation of the reservoir conditions than facies tracts, which is revealed by a comparison of the history-matching for the oil production. Permeability distributions obtained using the fractal results predict the high degree of heterogeneity within the reservoir sands of West Coalinga Field. The modeling results indicate that predictions of oil production are strongly influenced by the geologic framework and by the boundary conditions. The permeability data collected from the southern Utah outcrop, support a new concept for representing natural heterogeneity, which is called the fractal/facies concept. This hypothesis is one of the few potentially simplifying concepts to emerge from recent studies of geological heterogeneity. Further investigation of this concept should be done to more fully apply fractal analysis to reservoir modeling and simulation. Additional outcrop permeability data sets and further analysis of the data from distinct facies will be needed in order to fully develop this new concept