Fig 2 - uploaded by Ronghao Cui
Content may be subject to copyright.
Function curves of the polynomial kernel (a), the RBF kernel (b), and the mixed kernel (c). x 5 0.2 is the test point in three types of kernels. As an example of mixed kernel functions, d 5 1 and c 5 10 are set.
Source publication
Diffusion coefficient of carbon dioxide (CO2), a significant parameter describing the mass transfer process, exerts a profound influence on the safety of CO2 storage in depleted reservoirs, saline aquifers, and marine ecosystems. However, experimental determination of diffusion coefficient in CO2-brine system is time-consuming and complex because t...
Contexts in source publication
Context 1
... polynomial kernel in Eq. (6) and the radial basis function (RBF) kernel in Eq. (7) are typical examples of global and local kernels, respectively [76]. Figure 2 depicts the difference between them. As shown in Fig. 2(a), data points far from each other are able to affect kernel values effec- tively in global kernels, while local kernels in Fig. 2(b) only allow data points close to each other to exert an impact on kernel values ...
Context 2
... the curse of dimensionality [67,75]. In fact, the kernel function is divided into two types, i.e., global kernels and local kernels. The polynomial kernel in Eq. (6) and the radial basis function (RBF) kernel in Eq. (7) are typical examples of global and local kernels, respectively [76]. Figure 2 depicts the difference between them. As shown in Fig. 2(a), data points far from each other are able to affect kernel values effec- tively in global kernels, while local kernels in Fig. 2(b) only allow data points close to each other to exert an impact on kernel ...
Context 3
... The polynomial kernel in Eq. (6) and the radial basis function (RBF) kernel in Eq. (7) are typical examples of global and local kernels, respectively [76]. Figure 2 depicts the difference between them. As shown in Fig. 2(a), data points far from each other are able to affect kernel values effec- tively in global kernels, while local kernels in Fig. 2(b) only allow data points close to each other to exert an impact on kernel ...
Context 4
... m is the fraction coefficient. As shown in Fig. 2(c), mixed kernels not only receive strong response around the test point but also guarantee that the value of response far from the test point wouldn't be attenuated rapidly. However, we can see from Fig. 2 that all of these free coefficients, i.e., d, c, and m influence the performance of SVM. In Sec. 2.3, we will find their optimal ...
Context 5
... m is the fraction coefficient. As shown in Fig. 2(c), mixed kernels not only receive strong response around the test point but also guarantee that the value of response far from the test point wouldn't be attenuated rapidly. However, we can see from Fig. 2 that all of these free coefficients, i.e., d, c, and m influence the performance of SVM. In Sec. 2.3, we will find their optimal values. Journal of Energy Resources Technology APRIL 2019, Vol. 141 / 041001-3 2.2.3 Genetic Algorithm. We use GA to obtain the optimal values of c, d, c, and m. Introduced by imitating the evolutionary ...
Citations
... Generally, CO 2 diffusion coefficient in brine is known as a function of pressure, temperature, salinity, and also the composition of salt. However, several researches showed that CO 2 diffusion coefficient in liquid may be correlated to the properties of the liquid phase (such as viscosity, and density) [49,51,52]. Eq. (7.2) also shows that the effective diffusivity (D eff ;CO 2 2water ) in the presence of porous media could be reliant on the porosity ([) and tortuosity [26]. ...
Carbonated water injection (CWI) is considered as a feasible enhanced oil recovery (EOR) method that its mechanisms, which are responsible for oil recovery, mainly initiate with the partitioning of dissolved carbon dioxide (CO2) from carbonated water to the adjacent oil phase. Alongside the EOR, CWI is considered as a promising method for geological CO2 storage beneath earth layers in depleted oil reservoirs. Countless researches have been performed on in-depth analysis of the role of CWI in improving oil recovery at the core scale. A better understanding of the primary mechanisms leading to additional oil recovery is gained through pore-scale visual investigations. The role of operational parameters such as injection rate, CO2 content, pressure, and temperature, which influence the performance of CWI, are discussed in this chapter. Challenges of CWI application are presented and discussed technically and economically. Finally, some important research findings and gaps in present understanding are highlighted.
... In [31], the effectiveness of the LSSVM in estimating the CO2 solubility is studied and the influence of salinity, pressure, and temperature is analyzed. In [32], a model is developed by the use of GA and support vector machine (SVM), and its proficiency is examined in versus of NN-based approaches. In [33][34][35][36], decision making techniques are used for modeling problems in engineering applications. ...
The modeling problem is one of the important topics in engineering applications. In various applications, it is required to find a mathematical model to represent the relationship between output and the associated input variables. In this study, an approach on basis of a new deep learned type-3 (T3) fuzzy logic system (FLS) is introduced. The modeling of CO2 solubility on basis of temperature, molality of NaCl, and pressure is considered as an application. The monitoring of carbon dioxide (CO2) solubility in brine is one of the effective approaches in carbon capture and sequestration technique to reduce it in the atmosphere. A new hybrid learning method is presented to optimize the suggested model. The new adaptation laws are carry-out to tune the rule parameters and centers of membership functions (MFs). The values of horizontal slices and α-cuts are learned by the unscented Kalman filter (UKF). By the real-world experimental data sets, several statistical examinations, and comparison with conventional well-known fuzzy neural networks (NNs) and learning methods, the reliability and good performance of the suggested method are demonstrated. Also, the sensitivity of the input variables is analyzed by the use of the Sobol approach. © 2022, Budapest Tech Polytechnical Institution. All rights reserved.
... However, the hybrid kernel complicates the structure of SVM, which necessitates the adjustment and selection of more hyperparameters. To address this Research Article problem, numerous heuristic optimization methods such as the particle swarm optimization (PSO) [22], the genetic algorithm (GA) [23], the gravitational search algorithm (GSA) [24], the cuckoo search algorithm (CSA) [11], and the beetle antennae search algorithm (BAS) [25] have been applied to search the optimal hyperparameters of SVM. PSO and GA suffer the local minimum problem while CSA performs poorly in local searches [26,27]. ...
... Here, ω and b are the model parameters to be trained, C is the penalty coefficient, ε is the insensitive function, and ξ andξ are the slack variables [23]. The Lagrange method is used to solve Eq. (5), and then the SVM mode can be written as ...
... The most commonly used global and local kernel functions are the polynomial kernel and RBF kernel, as listed in Eqs. (7) and (8), respectively [23]. The polynomial kernel in Eq. (7) has a strong extrapolation ability, while the RBF kernel in Eq. (8) has strong interpolation ability. ...
In this paper, the optimal hybrid kernel support vector machine is employed to propose a compensation strategy intended for the temperature drift of a fiber optical gyroscope (FOG). First, the mode of the hybrid kernel with an interpolation and extrapolation capability is constructed, which consists of the radial basis function and the polynomial kernel function. Second, the combination model of the beetle antennae search algorithm and gravitational search algorithm that has both local and global search capability is proposed to optimize the structure-related parameters of a hybrid kernel support vector machine (HKSVM). Finally, the proposed approach is trained and tested using the experimental data of temperature drift at two different rates of temperature change (10°C/min and 5°C/min). In addition, the proposed method is validated against those conventional compensation algorithms. According to the research results, the compensation error (mean squared error) of the proposed approach is reduced by 92% compared to the traditional support vector machine based on the radial basis function.
... Amooie et al. conducted a comprehensive data-driven study on the interfacial tension of pure and impure gas-brine mixtures within saline aquifers, then, they developed seven machine learningbased models for predicting the IFT [19]. Feng et al. used a hybrid technique of support vector machine, mixed kernels, and genetic algorithm to realize an efficient and accurate prediction of the CO 2 diffusivity in brine at reservoir conditions [20]. Sahour et al. developed a methodology combining the statistical methods and machine learning techniques based on the available hydrogeology and hydrometeorology data, to map the groundwater salinity in the southern coastal aquifer of the Caspian Sea. ...
This paper constructs a prediction model based on Multilayer Perceptron (MLP) to explore the formation mechanism of brine (seawater evaporation or freezing). Four brine sets are extracted from the published, real-world data, and the simulation test with the same six chemical substance features. After integrated comparative experiments on six evaluation metrics, the results show that this model outperforms the other baseline prediction algorithms, achieving the highest precision 0.9625 and at least 8.45% improvement. Furthermore, for predicting the real-world test set, the results confirm the existence of freezing brine for the first time in Laizhou Bay area, China. This model is also used to analyze the mixed simulation results for brine and fresh groundwater. The experimental results indicate that only two out of twenty-nine samples of various concentrations change formation mechanisms after mixing. Overall, the model can effectively distinguish the evaporation and freezing brine, and discover the seawater concentration pathway.
... By modifying the structure of the SVM, Suykens and Vandewalle proposed least-squares support vector machines [38]. LS-SVM is also categorized as a supervised learning paradigm suitable for classification as well as regression tasks [38,39]. LS-SVM interprets experimental datasets using the least-squares methodology for solving a system of linear equations [40]. ...
Bubble point pressure (BPP) not only is a basic pressure-volume-temperature (PVT) parameter for calculation nearly all of crude oil characteristics, but it also determines phase type of oil reservoirs, gas to oil ratio, oil formation volume factor, inflow performance relationship, and so on. Since measurement of BPP of crude oil is an expensive and time-consuming experiment, this study develops a committee machine-ensemble (CME) paradigm for accurate estimation of this parameter from solution gas oil ratio, reservoir temperature, gas specific gravity, and stock-tank oil gravity. Our CME approach is designed using linear combination of predictions of four different expert systems. Unknown coefficients of this combination are adjusted through minimizing deviation between actual BPPs and their associated predictions using differential evolution and genetic algorithm. Our proposed CME paradigm is developed using 380 PVT datasets for crude oils from different geological regions. This novel intelligent paradigm estimates available experimental databank with an excellent accuracy i.e. absolute average relative deviation (AARD) of 6.06% and regression coefficient (R2) of 0.98777. Accurate prediction of BPP using our CME paradigm decreases the risk of producing from two-phase region of oil reservoirs.
... The capability to utilize CO2 for enhanced oil recovery in mature reservoirs as well as store it in depleted reservoirs to reduce greenhouse emissions has continued to make CO2 flooding an attractive process to several governments and energy companies around the world [1][2][3][4]. CO2 gas flooding is a popular method of enhanced oil recovery that aids microscopic fluid displacement by mechanical piston-like displacement, viscosity reduction, swelling of the oil and development of miscibility. Miscibility is one of the essential mechanisms, specifically for CO2-EOR processes that need to be carefully studied [5]. ...
The minimum miscibility pressure (MMP) is one of the critical parameters needed in the successful design of a miscible gas injection for enhanced oil recovery purposes. In this study, we explore the capability of using the Gaussian Process Machine Learning (GPML) Approach, for accurate prediction of this vital property in both pure and impure CO2-injection streams. We first performed a sensitivity analysis of different kernels and then a comparative analysis with other techniques. The new GPML model, when compared to previously published predictive models, including both correlations and other machine learning/intelligent models, showed superior performance with the highest correlation coefficient and the lowest error metrics.
... An in-depth review of the available correlations for predicting the diffusivity coefficient of CO 2 in brine reveals the limitations of these techniques from the applicability and accuracy perspectives (Feng et al., 2019). ...
... In the context of the affecting variables, salinity affects the solubility, interfacial tension and phase equilibria, thus influencing the diffusivity. In addition, salinity of the solvents affects brine's viscosities (Cadogan, 2015;Feng et al., 2019); therefore, the salinity effect on diffusivity of CO 2 in brine is emulated by considering the brine's viscosities as an input parameter while establishing the correlations and the paradigms. The data points collected from previous experimental studies were obtained using various techniques and equipments such as Taylor dispersion, a modified version of Ringborm's apparatus, laminar jet apparatus, laminar falling film, laser-induced fluorescence (LIF), 13C pulsed-field gradient NMR, physical absorption experiments in a stirred vessel operated with a horizontal gas-liquid interface, optical capillary cell via time-dependent Raman spectroscopy, wetted sphere apparatus, Taylor Aris dispersion method and see-through windowed high-pressure cell. ...
... The comparison includes the empirical models, namely those of Othmer and Thakar (1953), Wilke and Chang (1955) and Cadogan et al. (2015). In addition to the empirical models, the implemented GEP correlation was compared with one of the most recent intelligent paradigms proposed by Feng et al. (2019) based on hybrid genetic algorithm and mixed Kernels-based support vector machine. It is worth mentioning that when performing the comparison with the pre-existing approaches, we included only the points that satisfy the applicability conditions in each correlation. ...
Accurate knowledge of the diffusivity coefficient of CO2 in brine has a significant effect on the design success and monitoring of CO2 storage in saline aquifers, which is a part of carbon capture and sequestration (CCS). Frequently applied experimental approaches for determining this parameter are expensive and time-consuming, and empirical models cannot ensure accurate predictions. Therefore, there is a need to establish cutting-edge correlations for prediction of the diffusivity coefficient of CO2 in brine under various operating conditions. In this work, two white-box machine learning techniques, namely group method of data handling (GMDH) and gene expression programming (GEP) were implemented for correlating the diffusivity coefficient of CO2 in brine with pressure, temperature and the viscosity of the solvent. The obtained results demonstrated the accuracy of the proposed correlations. In addition, statistical and graphical analysis of the performances revealed that GEP correlation outperforms the GMDH correlation, decision trees (DTs), random forest (RF) and all the previous predictive models. GEP correlation exhibited an overall average absolute relative deviation (AARD) of 4.3014% and coefficient of determination (R²) of 0.9979. Finally, by performing the outliers detection, the validity of the GEP correlation was confirmed and only two experimental data points were identified as outliers.
... It is an elite strategy non-dominated sorting genetic algorithm. Based on NSGA, the elite strategy, the density estimation method, and fast non-domination ranking strategy are added [39][40][41]. NSGA-II is one of the most popular multi-objective genetic algorithms. It reduces the complexity of the non-inferior ranking genetic algorithm, has the advantages of fast running speed and good convergence of solution set, and becomes the benchmark of other multi-objective optimization algorithms [42,43]. ...
Variable valve timing technologies for internal combustion engines are used to improve power, torque, reduce emissions and increase fuel efficiency. Firstly, the paper presents a new electro-hydraulic FVVA system which can control the seating velocity of engine valve flexibly. Secondly, based on the NSGA-II genetic algorithm, outlines multi-objective optimization strategy, the paper designs the parameters of FVVA system to make the system easier to implement. Thirdly, the paper builds the combined FVVA engine simulation model. The combined simulation and experimental are executed to validate the designed FVVA engine. Simulation results show brake power is improved between 1.31% and 4.48% and torque is improved by 1.32% to 4.47%. Brake thermal efficiency and volumetric efficiency also show improvement. Experimental results have good agreement with simulation results. The research results can provide a basis for engine modification design.
Undersaturated oil viscosity is an important physical property for reservoir simulation, enhanced oil recovery, and optimal production. There are two distinct methods for undersaturated oil viscosity determination: the first one is experimental measurements which are usually expensive or unavailable; whereas the second one is empirical correlations which frequently have appropriate accuracy. Accordingly, searching for a high reliability undersaturated oil viscosity model is vital. This paper presents a new undersaturated crude oil viscosity model by using multi-gene genetic programming (MGGP). This model was built by using 528 experimental measurements data points that presents broad range of reservoir pressure and oil properties. Another, 276 points were used for validating and testing the new model against eleven published correlations. The results indicated that the new MGGP-based model yields a precise prediction of undersaturated oil viscosity.
