Search

Recently, many researchers have focused on reservoir production optimization because it is one of the most essential processes in closed-loop reservoir management. Surrogate-assisted production optimization in particular has received a lot of research attention. This technique applies a simple yet vigorous approximation model to substitute expensive numerical simulation runs. However, almost all the existing methods independently use a single approximation model and neglect the potential synergies between these models. In order to make full use of the potential synergies of these existing approximation models, a novel multi-fidelity (MF) surrogate-assisted multi-objective production optimization (MOPO) method based on transfer stacking (MFTS-MOPO) is proposed. In the MFTS-MOPO method, the radial basis function network and support vector regression surrogate models are applied to approximate the high-fidelity (HF) model as the two additional low-fidelity (LF) models. Then a multi-fidelity surrogate model is adopted to evaluate objectives during the optimization process by transferring the two additional and streamline low-fidelity models to the computationally expensive high-fidelity model. Furthermore, two sampling infill strategies are applied to efficiently improve the quality of the multi-fidelity surrogate model. The uniqueness of the proposed MFTS-MOPO method is that the transfer stacking technique is employed to efficiently use the information from different fidelity models to establish the MF surrogate model and the infill sampling strategy used to improve its performance. In addition, three benchmark problems and two reservoirs with different scales were applied to illustrate the effectiveness and accuracy of the MFTS-MOPO method. It was found that the MFTS-MOPO method had superior performance in convergence and diversity than other conventional methods. [ABSTRACT FROM AUTHOR]

In recent years, the surrogate-assisted method, which is one of the most promising ways for reducing the computational burden of numerical simulation–based water-flooding reservoir production optimization, has been widely applied by researchers worldwide. Meanwhile, there are different fidelity models that can be applied to extract samples for establishing a surrogate model. However, most of the researches neglected the potential synergies between low-fidelity (LF) and high-fidelity (HF) samples. In this study, therefore, a rapid intelligent multi-fidelity support vector regression (SVR) model–assisted multi-objective production optimization method, namely, MFSVR-MOPO, is proposed to lessen the computational burden of the numerical simulation–based production optimization. The uniqueness of this proposed method is that the LF and HF samples are mapped to a high-dimension space by the kernel function of the SVR, and the relationships between variables and objectives are evaluated by a linear model. Moreover, the gray wolf algorithm was applied to search for the optimal hyperparameters of the SVR model so as to improve the evaluation accuracy. Two frequently used production optimization synthetic reservoirs with different scales were studied to illustrate the effectiveness and accuracy of the MFSVR-MOPO method. The results showed that the MFSVR-MOPO method performed comparably with the HF model–based method in terms of convergence and diversity, but reduced the number of simulation runs on the two reservoirs by approximately 108 times and 50 times, respectively. [ABSTRACT FROM AUTHOR]

The resource potential of tight sandstone reservoirs in China is huge, and these reservoirs have typical characteristics such as complex lithology, low sandstone maturity, strong diagenesis, small pore throat, low natural capacity of producing wells, and strong nonhomogeneity. During development, some production wells suffer from low production and rapidly decreasing production of tight sandstone geological sweet spots. Based on the above understanding, unlike geological sweet spots considering the main parameters such as porosity, permeability, and oil saturation, four engineering sweet spot parameters were integrated and optimized in this study, namely, brittleness index, stress difference coefficient, fracture toughness, and natural weak surface. According to the characteristics of each parameter, a method is proposed to combine the probability index of a complex fracture system with the probability index to obtain a higher reservoir stimulation volume, and a method was established to evaluate engineering sweet spots of tight sandstone reservoirs. Using this evaluation method, an evaluation was conducted in this study for engineering sweet spots reservoir stimulation by combining geological and logging interpretation with a tight reservoir production well in Block C of Ordos Basin as the study target; the correlation between engineering sweet spots index FI and daily oil production was 0.67. On this basis, 5 production wells were selected in the area, and a comparative verification of the engineering sweet spots was conducted in this study using the criteria of the evaluation method. The evaluation test oil results of the evaluation method are considered to be more reliable and realistic, and have a certain guiding role on the field development of production wells in tight sandstone reservoirs. [ABSTRACT FROM AUTHOR]

Multi-objective well control optimization, which considers simultaneous optimization of more than one objective, has become an essential tool for efficient intelligent decision-making about well control parameters. In general, multi-objective well control parameter optimization involves numerical simulation, which is computationally expensive. Surrogate-assisted methods that use a simple yet rigorous approximation model have shown great potential in reducing the need for expensive numerical simulations. In order to strike a balance between reducing computational burden and enhancing accuracy of a surrogate model, we introduce here a semi-supervised learning method called tri-training surrogate-assisted multi-objective optimization (MOO–TTSA), which is based on radial basis function network (RBFN) and non-dominated sorting genetic algorithm-II (NSGA-II) to assist the optimization process. The uniqueness of this method is that a tri-training strategy is applied to enrich the training data set by selecting solutions with high-confidence fitness to renew the surrogate model at each generation so as to obtain better accuracy and lessen the computational burden. In this way, the number of simulation runs is reduced while the accuracy of the surrogate is guaranteed. To the best of our knowledge, this is the first time that a combination of tri-training technique and MOO is used for reservoir well control problems. The MOO–TTSA method was applied to two benchmark problems and two synthetic reservoirs. The results showed that the MOO–TTSA method reduced the number of simulation runs on the two reservoirs, one for about 160 times and the other for 50 times. Besides, the hypervolumes of both the benchmark and reservoir problems of MOO–TTSA were similar with those of the simulation-based NSGA-II method (NSGA-II), but superior to those of the single static surrogate model-based NSGA-II method (RBFN–NSGA-II). It was shown that the MOO–TTSA method was more efficient, superior and converged faster compared to the other conventional optimization models. [ABSTRACT FROM AUTHOR]

Studies on stress sensitivity of relative permeability in the laboratory were mostly carried out by varying overburden pressure at constant pore pressure (PP). However, in a real oilfield situation, changes in net stress are induced by changes in PP rather than overburden pressure. This research presents the effect of PP variation at a confining stress of 25 MPa on water–oil and silica nanofluid–oil relative permeability curves. Results showed that, at low PP variations of 0.1, 2, and 5 MPa, two-phase flow parameters exhibited distinct trends. Thus, for water–oil relative permeability, initial water (Swi) and residual oil (Sor) saturations both decreased as PP increased. End-point oil Kro (Swi) and water Krw (Sor) relative permeabilities both increased for water flooding. Similar trends were observed for nanofluid flooding. End-point mobility ratio and displacement efficiency both increased with increase in PP within this range for water and nanofluid flooding. At 10 MPa PP, Swi decreased compared to low PP variations, while Sor increased significantly. Similar observation was made for nanofluid flooding except for Sor which had moderate increment. End-point mobility ratio increased significantly for water flooding, but had a slight increase for nanofluid flooding compared to low PP variations. Displacement efficiency decreased for both water and nanofluid flooding compared to low PP variations. At the same PP, water–oil and nanofluid–oil relative permeabilities showed similar Swi and Kro (Swi) but differing Sor and Krw (Sor). The findings provide insights into the effect PP variation on relative permeability and would be important in flooding design considerations. [ABSTRACT FROM AUTHOR]

Silica nanofluids have proven to be successful in improving hydrocarbon recovery in the petroleum industry, and increased demand for hydrocarbons has necessitated its application in low-permeability reservoirs. In recent times, surface coating of nanoparticles has been employed to reduce its retention in porous media, but this does not entirely eliminate nanoparticle attachment to pore walls. Knowledge of changes that occur in pore wall and structure such as specific surface area (SSA), pore size distribution and total pore volume (TPV) would be useful in understanding retention mechanisms. This study used nitrogen adsorption technique in studying changes in pore structure due to silica nanofluid treatment. The Brunauer–Emmett–Teller theory and Barrett–Joyner–Halenda adsorption model were used in determining SSAs and TPVs, respectively. SSA, adsorbability and TPV increased in treated samples compared to untreated samples and the rates of change increased with treatment time due to extra pores induced by nanoparticle coagulation. Percent changes in TPV matched closely with SSA and was responsible for increments in the latter. Scanning electron micrographs confirmed coagulation of nanoparticles which increased with treatment time and introduced pseudo-pores on pore walls, resulting in increase in TPV. Increase in differential pore volume was observed for the entire studied range of 2–100 nm except for 3–4 nm which showed no changes in all samples. Severity of differential pore volume increased with treatment time. This study provides insights into nanoscopic changes that occur on pore walls and structure when employing silica nanoparticles in improving hydrocarbon recovery in low-permeability hydrocarbon reservoirs. [ABSTRACT FROM AUTHOR]