Discrete Fracture Network Modeling of Saline Lacustrine Tight Oil Reservoirs Severely Deformed by Tibetan Plateau Neotectonism

Abstract The study area is located in Qaidam Basin on Tibetan Plateau, which has very complex geological background sedimentologically and structurally. Building the discrete fracture network (DFN) of study zone is a critical work step for subsequent dual porosity / dual permeability (Dp/Dp) dynamic simulation which further enables the optimized design of field development plan (FDP). The saline lacustrine tight oil reservoirs in study zone were severely deformed by the Neotectonism coupling with the far-field collision between the Indian and Eurasian plates. There were relatively abundant fracture interpretations from image well logs with core calibration and relatively amplitude preserving seismic cube. First, statistics of well fractures were performed in terms of their types and attitudes, which enabled proposing their genetic mechanism together with information of other references and dividing fracture sets more reasonably for subsequent DFN modeling. Then, many efforts were made on fracture driver attribute screening with the aid of neural network (NN) and principal component analysis (PCA) functions. Dozens of possible fracture driver attributes were examined, which were mainly geometric attributes, edge-detection attributes, amplitude/frequency related attributes and lithology/porosity related attributes. Meanwhile, microseismic monitoring (MSM) data were also used to help screen the pertinent fracture drivers. Three fracture driver attributes were finally screened out and they were combined together to produce a fracture intensity trend which was finally used as co-kriging secondary variable together with well fracture intensity logs for modeling the 3D-grids of fracture intensity based on gaussian random function simulation algorithm. Besides the existed fault model (as large-scale fractures) generated when building matrix geomodels, a medium-scale fracture set was built deterministically with "Fault Patch" extraction method based on the seismic edge-detection Ant Cube. Then for small-scale fractures, totally four fracture sets were built stochastically based on split 3D fracture intensity and their own parameters of geometry, orientation, aperture. These five fracture sets were then upscaled to produce the fracture properties with porosity and permeability constraints from wells, i.e., usually less than 0.1% and hundreds of millidarcies respectively. Then these fracture properties were further calibrated and used in Dp/Dp dynamic simulation process. Genuine discrete fracture network and Dp/Dp dynamic simulation were built up and performed for the first time for the complex oil reservoirs on Tibetan Plateau, based on organic multi-disciplinary integration. But more importantly, many useful conclusions and decisions were reached based on the Dp/Dp dynamic simulation which made the field development plan more robust and cost-effective.