查看更多>>摘要：Mixed nanoparticle stabilized Pickering emulsion is a novel research topic, which has been attracted more attention. In the present study, the novel amphiphilic GO-Janus SiO2-Cn nanoparticles (GOJS-Cn, n represents the number of carbon atoms in the modified alkyl chain) were successfully fabricated for the first time and investigated their ability in stabilizing Pickering emulsions for enhancing oil recovery. Fourier transform infrared spectroscopy (FTIR), transmission electron microscopy (TEM), and X-ray photoelectron spectroscopy (XPS) were employed to determine the successful fabrication of the GOJS-Cn. The optical microscopy was utilized to directly observe the GOJS-Cn (n = 0, 3, 8, and 12) stabilized emulsions and the static multiple light scattering was employed to further evaluate its stability quantitatively. It was found that the GOJS-C12 displayed the greatest ability to stabilize emulsions, even in high salinity (10~4 mg/L NaCl/MgCl2/CaCl2/AlCl3) and temperature (90 °C). Then, the interfacial tension, water contact angle, and dilatational measurements were conducted to reveal the mechanism of the GOJS-C12 superior emulsifying ability. The amphiphilic GOJS-C12 with great amphiphilicity spontaneously adsorbed at oil-water interface to promote the formation of rigid interfacial film, which was essential for the long-term stability of emulsion. Moreover, the core flooding tests reflected the GOJS-C12 stabilized emulsion could significantly enhance oil recovery (～14.8%), then based on microscopic visualization experiments the potential EOR mechanism was proposed.
查看更多>>摘要：Deep learning technology can fit the non-linear and non-stationary characteristics in geological statistics. It has become an important tool for predictive modeling. However, the classical deep neural network cannot integrate the spatial dependence directly from UTM_X, UTM_Y and depth when applied to spatial prediction and lacks the reliability analysis of the results. This study proposed a flexible spatial prediction method based on gated recurrent unit (GRU) neural network with Kriging technology. Spatial dependent deep learning, spatial constraint weights and related reservoir information are used to complete the prediction of lithology spatial distribution. Furthermore, Bayesian theory is integrated to quantify the reliability of the model. Seismic information is used as the spatial constraint of GRU neural network to realize the fusion of cross-domain knowledge, which can improve the accuracy and reliability of the prediction. Compared with the traditional Kriging method and deep learning method, the prediction accuracy (R~2 = 95.071%) of the proposed method is improved by 8.642% and 3.034% in the field data. The method can be potentially applied to the spatial distribution prediction of related geological conditions.
查看更多>>摘要：Acoustic logging while drilling (LWD) technology is extensively applied for petroleum prospecting. It utilizes the acoustic characteristics of stratum to deduce the formation properties such as porosity, permeability, and oil saturation, etc., which can help to detect the location of oil reservoirs. In acoustic logging process, the formation longitudinal (P) and transverse (S) wave velocity is the primary factor of interest. Nevertheless, an accurate identification of P/S wave velocity in different formations is hard to be performed due to the influence of the collar waves, which has become an urgent problem in LWD measurements. To study this issue more practically, a small-sized measurement system based on the propagation mechanism of collar waves is established in our laboratory. With this system, we design and perform multiple sets of monopole acoustic LWD experiments with different models. After recording the experimental data, the attenuations of inner and outer collar waves received along the axial direction are compared and analyzed. The result shows that the collar waves have an exponential decay in amplitude by the expansion of source-receiver distance, but the outer collar wave decreases more rapidly. We also conduct experiments at different source frequencies (20-400 kHz) to quantitatively describe the relationship between the amplitude and frequency. The result suggests that the inner/outer collar waves and P/S waves have a similar frequency response, and it's difficult to remove collar waves from the full waveforms of LWD signals by only changing source frequency. Therefore, we present a new data processing method for extracting the formation wave velocities according to the energy distribution of the componential waves, and then apply it in the analysis of two borehole measurements. The inner and outer collar waves are almost removed from the processed time domain waveforms, which leads to a successful verification of formation P/S wave velocities. These experimental results prove the feasibility and repeatability of our method, and propose an alternative method for the LWD data explanations.
查看更多>>摘要：Currently, the common step of intelligent learning application in reservoir characterization is to study its performance in intelligent computing directly. When the specific task is relatively simple and the adequate learning samples are not beyond reach, the model established by the intelligent learning algorithm can often achieve satisfying results. However, small labeled samples are usually the majority in reservoir characterization, making the method with stronger learning ability more likely to fall into the trap of noise in data, which further results in the poor or unstable performance of the model in practice. In this paper, we propose a learning method that aggregates fc-Nearest Neighbors (fcNN), Decision Tree (DT), Random Forest (RF), and XGBoost by stacked generalization (Stacking) to obtain a more robust model. We adopted this model to classify the lithofacies of a real well log data and compared it with the models established by the above four learning methods and the Soft Voting ensemble, through a statistical test measured by Fi score. The results show that compared with other methods, the proposed method can establish a more robust model with higher prediction accuracy for logging lithofacies classification with limited data.
查看更多>>摘要：In oil reservoirs, the sweet spot is found that the well could be positioned quickly and accurately, the drilling rate and the oil-gas production are increased, development cost is reduced. Among them, sorting, granularity and porosity are important factors to evaluate whether the exploration area is a sweet spot. A low permeability oil reservoir is taken as the research object, this paper mainly focuses on the prediction of the above evaluation parameters. In order to solve this problem, this paper proposed a new deep learning hybrid model. The model is constructed based on temporal convolutional network (TCN) and long short-term memory network (LSTM). Firstly, the influence of logging parameters on the prediction of evaluation indexes is analyzed. Secondly, the model is used to predict the screened sequence data. In the process of model construction, particle swarm optimization (PSO) is used to optimize the global hyperparameters, and finally the prediction model is obtained. The model is compared with TCN algorithm, traditional machine learning and empirical formula. This model improves the prediction accuracy of reservoir evaluation parameters in low permeability oilfield.
查看更多>>摘要：Hydraulic fracturing flow-back fluid (FFF) poses a significant challenge to environmental security, and it also can provide valuable information for geological and engineering applications. In order to reveal the source and composition of excess FFF in Jimsar tight oil reservoir, Junngar Basin, Northwest of China, and determine the stage of the hydraulic fracturing flow-back process, this study systematically collected the samples of surface water (SW), hydraulic fracturing fluid (HFF), and FFF from three horizontal wells in the study area. After measuring and analyzing the thousand deviation values of diplogen (δD) and oxygen (δ~(18)O) isotopes from the collected samples, we obtained the geochemical information and the temporal variation characteristics. By analyzing the influence degree of isotope exchange of Water Rock Interaction (WRI) and nonlinear fitting the field measured data of the FFF, the convergent model of δ~(18)O with flow-back time of the FFF was constructed. Based on the law of mass isotope conservation, we proposed a method to calculate the content of formation water in the FFF. The results showed that the SW (δD_(SW) = -73.35‰, δ~(18)O_(SW) = -11.30‰) and HFF (δD_(HFF) = -73.03‰, δ~(18)O_(HFF) = -11.20‰) samples conform to the isotopic characteristics of atmospheric precipitation in the study area. In contrast, the FFF samples are characterized by a positive drift in δ~(18)O and a negative drift in δD, clearly deviating from the atmospheric waterline distribution. The water composition of FFF has changed. The temporal variation characteristics of the δ~(18)O in FFF indicated that the flow-back process can be divided into three stages: The first stage was about 15 days after the flow-back process. As the formation water began to mix into the flow-back fluid, the δ~(18)O of FFF increased rapidly. The second stage was 15-40 days after the flow-back process, and the increase rate of δ~(18)O was smaller than that in the first stage. The isotopic composition of FFF gradually shifted to formation water dominated. The third stage was 40 days after the flow-back process, and the δ~(18)O tended to be stable. To be specific, as formation water was continuously mixed into the fracturing fluid, the composition of FFF gradually changed from HFF dominated to formation water. It was calculated from the field measured data that the original δ~(18)O of formation water was -6.902‰; on the 60th day of the flow-back process, the content of formation water in FFF was about 84% and 81% in well JHW43 and JHW44, respectively. Conclusively, the FFF is mainly composed of formation water.
查看更多>>摘要：Vertical seismic profile (VSP) data have the advantage of higher data quality than surface seismic data. Using VSP data has become a powerful technology to achieve accurate recognition of fine structures. Conventional VSP seismic migration methods mainly use up-going waves (reflection waves) for imaging and often treat down-going waves (multiple waves) as a kind of noise to be suppressed and eliminated. The up-going waves (reflection waves) only provide a narrow subsurface coverage near the receiver well. However, the down-going waves (multiple waves) also carry underground medium information and can provide much wider subsurface illumination than the up-going waves (reflection waves). The imaging quality can be improved by using reflection and multiple waves simultaneously. Therefore, we developed a new joint reverse time migration (JRTM) method to use both the up-going and down-going waves simultaneously based on the mirror principle. The up- and down-going data can be simultaneously placed at the reflection positions and mirror reflection positions as the boundary conditions and initial conditions for wavefield reverse time extrapolation. Some numerical examples from the graben, the modified Marmousi, and the modified HESS models show that the proposed new JRTM method can jointly use and image multiple and reflection waves simultaneously. In addition, the newly proposed JRTM method also has a higher imaging quality than reflection wave migration and multiple wave migration from VSP data in complex structures. High-quality imaging results from our proposed JRTM can be further used to improve the accuracy of subsequent oil and gas reservoir characterization and seismic inversion of VSP data.
查看更多>>摘要：Liquid accumulation is one of the most common challenges associated with natural gas gathering processes particularly in pipelines where mechanical pigging is infeasible and there is a risk of terrain slugging. Aqueous foam drainage technology is a potential method to remove liquid accumulation owing to its low risk and cost and lack of constraints. An experimental study of gas-liquid two-phase flow with foam was conducted to clarify the effects of foam on the slug generation mechanisms at the elbow of upward-inclined pipelines. The experiments were conducted in horizontal-upward pipelines at different inclination angles (5°, 10°, and 20°). The flow pattern was captured by a high-speed camera and the real-time differential pressures along the pipeline were recorded. The results indicate that the slug generation mechanisms at the elbow mainly consist of an interfacial wave growth mechanism at low gas velocities and an interfacial wave aggregation mechanism at high gas velocities. In the presence of aqueous foam, the interfacial wave growth mechanism still exists. In contrast, the interfacial wave aggregation mechanism disappears, and the flow pattern transforms into wavy foam flow under the same conditions. In addition, the increase in the inclination angles intensified the liquid backflow in the upward-inclined section and contributed to the liquid accumulation at the elbow. Furthermore, owing to the Bernoulli effect, the increase also enhanced the liquid-carrying capacity of the gas in the divergent section of the elbow, resulting in increased interfacial fluctuations and promoting the generation of liquid slugs. Meanwhile, the eigenvalues extracted from the fast Fourier transform (FFT) of the differential pressures correlate well with the flow characteristics at the elbow, and the average values at separated flow conditions are all below 0.05, which proves that the method enables an accurate prediction of whether the liquid slug is generated at the elbow of the upward-inclined pipe.
查看更多>>摘要：History matching can estimate the parameter of spatially varying geological properties and provide reliable numerical models for reservoir development and management. However, in practice, high-dimension, multiple-solutions and computational cost are key issues that restrict the application of history matching methods. Recendy, the combination of deep-learning-based surrogate model and sampling algorithm has been widely studied in history matching to overcome the limitations. Considering that real-world large-scale reservoirs often have hundreds of thousands or even millions of grid-based uncertain parameters, extracting spatial features using convolutional neural networks requires a lot of computational cost and storage requirements. Therefore, in this work, we mainly study how to use the recurrent neural network (RNN) to construct the surrogate model for history matching. Specifically, we propose a multilayer RNN surrogate model based on a vector-to-sequence modeling framework. The multilayer RNN surrogate model with gated recurrent unit (GRU), termed MLGRU, is developed to approximate the mapping from feature vector of geological realizations to the production data. The feature vector is the low-dimensional representation of geological parameter fields after using the re-parameterization method, while production data are the simulation results of historical period. In addition, we design a log-transformation-based windowed normalization (LTWN) method for the production data, which can enhance the learnability and features of production data. The MLGRU model is incorporated into a multimodal estimation of distribution algorithm (MEDA) to formulate a history matching workflow. The hyper-parameters and performance of the proposed MLGRU model are analyzed by numerical experiments on a 2D reservoir model. Furthermore, numerical experiments performed on the Brugge benchmark model, a large-scale 3D reservoir model, demonstrated the performance of the proposed surrogate model and history matching method.
查看更多>>摘要：Many fractured reservoirs are dual porosity reservoirs including matrix pores and fractures, which both contribute to the porosity. However, previous studies either ignored the effect of pores or fractures on dual laterolog, or had special application conditions. Considering the joint contribution of pores and fractures, a response equation of dual laterolog in dual porosity reservoirs is established. First, based on a borehole model with a vertical fracture in dual porosity reservoirs, fracture porosity equations are established in the investigation depths of deep and shallow laterologs, respectively. Using Archie's equations, the response equations of deep and shallow laterolog resistivities are derived, respectively. Then, the response equation, which is similar to the traditional equation of dual laterolog in fractured tight reservoirs, is established. It contains three unknown parameters, i.e. fracture width, fracture cementation exponent and matrix cementation exponent. To solve the equation, a virtual core model is proposed to estimate the fracture cementation exponent by fracture width. The estimation method of matrix cementation exponents is established by empirical formulas. Then, the response equation can be transformed into a nonlinear equation with one unknown parameter of fracture width, and conventional iterative methods can be employed to solve it and obtain the fracture width. Using the response equation and the solution scheme, the change characteristics of fracture cementation exponents in different fractured reservoirs are first analyzed. It shows that the fracture cementation exponents increase with the fracture widths, and they decrease with the increase of the ratio of formation water resistivity to unfractured reservoir resistivity. The applications indicate that the equation and solution scheme can obtain relatively accurate fracture widths quickly in dual porosity reservoirs, and they are recommended for dual porosity reservoirs with high fracture angel (>70°) and low fracture density (<l/3.78 m~(-1)). The findings of this study can help better understanding of the different contributions of fractures and pores to dual laterolog responses in dual porosity reservoirs, and they can provide an alternative solution for evaluating dual porosity reservoirs.
NSTL