Evaluating an adequate computer model representing the characteristics of a subsurface reservoir is of great importance in the oil and gas industry. With recent data acquisition developments, high-resolution geological models can be generated for the subsurface reservoirs; however, uncertainties of geological data and too many details in the model have brought up remarkable limitations for application of high-fidelity geo-models in reservoir management. To generate efficient low-resolution models, which are relatively representative of the reservoir dynamic behaviour, data-driven models are proposed. The data-driven models are trained principally based on the production history of the reservoir with primary knowledge of structural information. The Data-driven Discrete Well Affinity (DiWA) model is designed to generate efficient proxy models that require basic geological information and full production history of the reservoir. Since it benefits the computational power of the Delft Advanced Terra Simulator (DARTS) and adjoint gradient-based optimization, the DiWA model has become an efficient framework in terms of accuracy and computational costs. This research shows that the DiWA model can relatively reconstruct the reference reservoir parameters out of an ensemble of priors, and the final results can be improved utilizing closer prior knowledge to the true response. Additionally, it is demonstrated that the performance of the DiWA model in a complex, realistic oil field can be enhanced using a comprehensive objective function, applying appropriate well controls, understanding the underlying driving mechanisms, using sufficient degrees of freedom, and applying consistent boundary conditions.

In 1949, Toms (Toms B.A., (1949, 1977)) observed that small amounts of a drag reducing agent (DRA) could cause a considerable drag reduction in turbulent pipe flow. In application of polymer enhanced oil recovery, degradation of polymers in the supply lead could cause clogging. It was, however observed that surfactants at sufficiently high concentration also showed drag reduction without the problem clogging. A DRA reduces the energy loss by friction and unstable flow, thus improving injection throughput with the same pressure pump and thereby reducing the exergetic pumping costs. This study investigates experimentally the drag reducing capacity of surfactants and compares it to the drag reducing capacity of polymers. For the experiment, a set-up consisting of a pump, a coiled test tube with a length of 1.48 m and an inner diameter of 0.5 mm and pressure gauges is built. The diameter of the coil is 12.5 cm. We use a pump capable of injection up to 200 ml/min. The pressure drop is measured between the entrance and end of the tube. The injection rate is varied between 1 and 200 ml/min, roughly corresponding to Reynolds numbers between 50 and 10,000. The additives are dissolved in brine with a 33,000 ppm salt concentration. The viscosity of the solution is dependent on the concentration of the DRA. The ratio of the measured pressure drop with only brine and the pressure drop with the DRA solution was used to calculate the drag reduction (DR) factor, as from a technical point of view we are only interested whether adding DRA reduces the drag with respect to the original brine solution. From an academic point of view, we remark that for low concentrations the viscosity enhancement due to the presence of the DRA is negligible. As polymers we use xanthan (a biopolymer), and a synthetic emulsion polymer based on polyacrylamide. Maximum DR factors are 23% for xanthan at 90ppm and 32% at 90ppm for the synthetic emulsion polymer. DR only occurs at turbulent conditions. Three types of surfactants, each from a different branch of surfactants are used in this study. The surfactants used are AOS {훼-Olefin Sulfonate}, CTAB {hexadeCylTrimethylAmmonium Bromide} and APG {Alkyl PolyGlucoside} which are a cationic, anionic and a nonionic surfactant respectively. The surfactants did not show any DR at (for DRA applications) high concentrations up to 20.000ppm. Addition of Sodium Salicylate (NaSaL) to CTAB with a 1:1 ratio led to a maximum DR of 33% at 2500 ppm concentration. Several pressure gauges have been installed along the test tube in order to observe how the pressure drops along the tube, how the DRAs affect these pressure drops and at what location of the test tube the DR factor is the highest. It is found that xanthan has the same DR factor at each location of the test tube, the emulsion polymer has a decreasing DR factor as the distance from the inlet of the test tube increases and the CTAB+NaSaL DRA has an increasing DR factor as the distance from the inlet increases. The DRAs are sheared using a constriction in the flow loop while the degradation is monitored. It is observed that xanthan is less susceptible to degradation in comparison to the emulsion polymer due to its more rigid chemical structure. But xanthan and the emulsion polymer would be inefficient to use in looped flow systems as they are affected by degradation. The CTAB+NaSaL DRA on the other hand shows no degradation meaning that the micellar rod-like structures that give the DR effect are being repaired when the shear force is being removed. However, for surfactants higher concentrations (1000-2500 ppm) are required.

Compositional simulation is computationally intensive for high-fidelity models due to thermodynamic equilibrium relations and the coupling of flow, transport and mass transfer. In this report, two methods for accelerated compositional simulation are outlined and demonstrated for a gas vaporization problem. The first method uses a proxy model that reduces the number of components and the second method reduces the number of grid blocks (i.e. upscaling). Both methods are implemented within the operator-based linearization framework of the Delft Advanced Research Terra Simulator. Lebesgue integration is applied in the loss function of a neural network allowing the neural network to discover the operator space of the reference model in reduced dimensions. Training is carried out for a one-dimensional homogeneous reservoir and minimizes the misfit of the leading and trailing shocks of a compressible pseudo-binary model with respect to observations of the reference model. The operator space of the pseudo-binary model is initially approximated with the method of multiscale reconstruction of physics, a numerical representation of the method of characteristics. Training is carried out in a two-stage transfer learning scheme to increase computational efficiency. In the first stage, neural networks are trained to approximate the analytical reconstruction. In the second stage, a solver is embedded in the loss function of the neural network and the forward solution is used to calculate the Lebesgue integral. The transfer training scheme minimizes the misfit of the leading and trailing shocks for 10 discrete time steps in a one-dimensional homogeneous reservoir. The misfit of the trained model shows a significant improvement in the location of the trailing shock and a modest improvement in the estimation of the leading shock. The trained proxy is applied to the top and bottom 15 layers of the SPE10 model and the estimation of the first and last breakthrough is assessed in conjunction with the error of the phase-state classification. The phase-state classification is significantly improved through time which is also expressed in improvements of the estimation of breakthrough times. The average difference in breakthrough time for the trained and untrained models with respect to the reference model is 293days versus 570days for the trailing shocks and 15days versus 16days for the leading shock. The established training framework enables the development of proxies with increased complexity. Rigorous upscaling defines the upscaled operator space with dynamic and non-equilibrium thermodynamic upscaling functions. These functions combined, define the upscaled operator space for the three-dimensional compositional space and are inferred from data points gathered from a limited, characteristic portion of the full-size model. Gathered data points are interpreted with an interpolation function or neural networks to construct structured OBL meshes for implementation within DARTS. This upscaled operator space can effectively be used for different boundary conditions without reevaluating the upscaling functions.

Water injection is a core element of geothermal, petroleum and waste water management industries where water is injected into the subsurface after treatment and filtration to comply with environmental regulations and water quality requirements. Some impurities still remain in the water even after the filtration process. It has been widely reported that water injection operations encounter severe injectivity decline which is of substantial concern for field management. Total dissolved and suspended solids in injection water are one of the significant reasons for injectivity decline. These solids are then filtered by the porous media and cause formation damage which can result in significant injectivity decline. Major past studies have assumed typical filtration techniques (micro-filtration) where the size of remaining dispersed particles in the water are reduced to ~2 microns. Hence, the majority of the experimental studies have used suspended particles of ~1-5 microns to investigate injectivity decline. These studies show that initially internal filter cake starts to develop and after some transition time external filter cake is formed, after which injection face is almost completely plugged. However, not many studies have been performed with ultrafiltration where remaining dispersed particle size in injection water is reduced to nano size range. In this study, injectivity decline by ultra-filtered water injection was investigated experimentally. To mimic ultra-filtered water, spherical silica nanoparticles of 120 nm diameter were used as dispersed particles in the injected water. First, stability study of nanoparticle colloid was carried out by varying nanoparticle concentration, brine compositions and pH. Hydrodynamic size and zeta potential measurements showed that there exists a salinity and pH range in which nanoparticle colloid remain within the expected size range. Core flood experiments were conducted on Bentheimer sandstone core plugs. Pressure measurements along the core and influent/effluent analysis were used to study the transport and retention of nanoparticles in porous media. Experimental results showed about 50 to 70 percent less injectivity decline compared to micron size suspended particles. Furthermore, results showed that external filter cake does not form by nanoparticle flow through porous media if the injection fluid’s pH and salinity are kept within a defined range obtained from stability study. Only deep bed filtration takes place where three main retention mechanisms dominate i.e. surface deposition, plugging and entrainment. Finally, a numerical model is presented in this study that describes deep bed filtration taking into account observed retention mechanisms. Model results are found to be in good agreement with experimental results.