Seismic to Simulation Fracture Characterization of a Green Carbonate Reservoir in Presence of Large Uncertainties

Abstract

Abstract This Middle East case study addresses the modeling of a green carbonate reservoir which production is controlled by fractures. The workflow integrates 3D seismic, cores, logs and production data to assess and minimize the production forecast uncertainties. The 3D seismic was used both for the characterization of layer-bound fractures, by means of fracture-related attributes, and to locate sub-seismic faults and associated fracture corridors. The extraction of seismic lineaments was performed using edge-enhancing and automated interpretation techniques. A simple approach was introduced to identify the most persistent lineaments and increase the confidence in their structural nature. The distribution and properties of small scale fractures was constrained by a combination of image logs, cores, drilling logs, seismic attributes and petrophysical model. Outside the corridors, the vertical variability of the fracture density was derived from the interpretation of electrical images, whereas its lateral variations could be correlated to petrophysical changes. Two consistent sets of orientations could be identified from image logs, while a detailed comparison between seismic attributes and flowing fractures from well data showed that the most open fractures were located in the vicinity of the faults. Multi-point statistics (MPS) were then used to constrain the location of the most effective fractured zones while retaining spatial consistency and maximum accuracy at the well locations. The final discrete fracture network (DFN) honored all measurements and computed trends and was upscaled to effective fracture properties (porosity, permeability and shape factor) in the reservoir grid. A dual porosity simulation model was generated and the fracture properties were calibrated against dynamic data. The mismatch between observed and simulated dynamic data was ascribed to the uncertainty in the input to the DFN modeling process. To avoid local modifications while preserving the original fracture property distribution and contrast in the model, history matching was performed by iteratively conditioning fracture intensity to dynamic data and subsequently recomputing the effective fracture properties based on the new modified fracture intensity distribution.