Use of genetic algorithm to estimate CO2–oil minimum miscibility pressure—a key parameter in design of CO2 miscible flood

A new genetic algorithm (GA)-based correlation has been developed to estimate the CO₂–oil minimum miscibility pressure (MMP)—a key parameter in design of CO₂ miscible flood to enhance oil recovery. In the order of their effects and importance, the correlation uses the following key input parameters: reservoir temperature, molecular weight of C₅₊, and the ratio of volatiles (C₁ and N₂) to intermediates (C₂–C₄, H₂S, and CO₂) and has been validated against experimental data and other commonly used correlations reported in the literature, notably that of Alston et al., Glaso, Yellig and Metcalfe, Cronquist, Lee, and Holm and Josendal early correlation.Based on the Darwinian theory of evolution, it is demonstrated that the GA is particularly suited to problems with non-linearity, variable discontinuity, large search space, and all kinds of objective and constraint functions. It works by exploration and exploitation of the search space (that is, the probability of finding the global optimum increases). Moreover, GA solutions are less likely to be misleading, as the problem does not need to be ”restructured” to fit the solution method. That is, the development of the GA-based correlation does not entail any pre-correlation data manipulation, and as a consequence, all data could be utilized as reported.The GA software developed in this study uses chromosomes coded with real numbers to encode correlation coefficients in an initial random population with 100 chromosomes size. Such encoding technique enhances the GA robustness. For the selection technique, the roulette wheel method has been used. Furthermore, to produce the offspring, one-point crossover and mutation with 100% probability are used. Noteworthy advantage offered by GA-based correlation is that this correlation can be used when there is a lack of experimental data and also to design an optimal laboratory program to estimate MMP. Moreover, the correlation errors could be minimized further through a series of iterative optimisation runs.Our results suggest that a GA-based CO₂–oil MMP correlation could be superior to other correlations commonly used. For example, compared to other correlations, the GA-based correlation yielded a better match with data with an average error of 5.86% and standard deviation of 7.96%, followed by Alston et al. correlation, which yielded an average error of 8.21% with a standard deviation of 10.04%. Among the correlations tested and compared against the GA-based correlation, Yellig and Metcalfe correlation yielded an error of 14%.     

基于BP网络和遗传算法的波阻抗混合反演

针对BP网络和遗传算法（GA）波阻抗反演精度较低且效率不高等问题，提出一种基于BP网络和遗传算法的混合波阻抗反演方法。该方法利用BP算法计算出一定精度的波阻抗初始模型后，再利用改进的遗传算法对该初始模型进行迭代反演，可得到精度更高的反演结果。BP-GA混合波阻抗反演算法简化了二进制的编码解码过程，对BP和GA进行了优化，在保证一定精度的条件下提高了收敛速度，分别用不同子波、不同初始模型和不同噪声试验了该方法的效果，并用实际资料验证了该算法的有效性。     

一种基于BP神经网络的完井液污染类型识别方法

目的 解决油田目的层钻井过程中完井液受盐水、残酸等污染后不能高效识别污染类型的问题。方法 对完井液进行不同质量占比的盐水、残酸污染测定,采用K-means聚类订正不同污染等级数据样本的标签。根据数据样本特征的获取难易度、隐藏层数目,训练不同的BP神经网络模型,并由留一交叉验证法检验模型的分类准确率。结果 数据样本拥有的特征越多,训练的BP神经网络分类准确率越高,隐层数目越多,分类准确率反而越低。选择包含“流变+老化+滤失+井名”4类特征的数据样本建立1隐藏层的BP神经网络模型,其平均分类准确率达到93.18%。结论 由流变、滤失等特征训练的BP神经网络模型可快速应用于试油现场,解决完井液污染类型识别问题,避免了试油现场因缺少大型仪器而无法鉴别完井液污染类型的难题。     

The Development of a Robust ANFIS Model for Predicting Minimum Miscibility Pressure

Miscible gas injection has been considered one of the most important enhanced oil recovery techniques. Minimum miscibility pressure (MMP) is a key parameter in the design of an efficient miscible gas injection project. This parameter is usually determined using a slimtube apparatus in the laboratory. However, many attempts have been made to introduce MMP predicting correlations. In this study an adaptive neuro-fuzzy inference system (ANFIS)–based correlation has been developed to estimate the MMP values. In this model, the MMP of reservoir fluid is correlated with 27 variables containing concentrations of different components in reservoir oil and injecting gas, molecular weight and specific gravity of C₇ ⁺ in reservoir oil and also reservoir temperature. This correlation can be applied to predict the effect of each individual parameter on the MMP values.     

Predicting the Bubble-Point Pressure and Formation-Volume-Factor of Worldwide Crude Oil Systems

Abstract

This paper presents models for predicting the bubble-point pressure (P _b ) and oil formation-volume-factor at bubble-point (B _ob) for crude oil samples collected from several regions around the world. The regions include major producing oil fields in North and South America, North Sea, South East Asia, Middle East, and Africa. The model was developed using artificial neural networks with 5200 experimentally obtained PVT data sets. This represents the largest data set ever collected to be used in developing P _b and B _ob models. An additional 234 PVT data sets were used to investigate the effectiveness of the neural network models to predict outputs from inputs that were not used during the training process. The network model is able to predict the bubble-point pressure and the oil formation-volume-factor as a function of the solution gas–oil ratio, the gas relative density, the oil specific gravity, and the reservoir temperature. In order to obtain a generalized accurate model, back propagation with momentum for error minimization was used. The accuracy of the models developed in this study was compared in details with several published correlations. This study shows that if artificial neural networks are successfully trained, they can be excellent reliable predictive tools to estimate crude oil properties better than available correlations. The network models can be easily incorporated into any reservoir simulators and/or production optimization software.     

Accurate prediction of properties of carbon dioxide for carbon capture and sequestration operations

Development of robust predictive models to estimate the transport properties of gases (namely viscosity and thermal conductivity) is of immense help in many engineering applications. This study highlights the application of the artificial neural network (ANN) and least squares support vector machine (LSSVM) modeling approaches to estimate the viscosity and thermal conductivity of CO₂. To propose the machine learning methods, a total of 800 data gathered from the literature covering a wide temperature range of 200–1000 K and a wide pressure range of 0.1–100 MPa were used. Particle swarm optimization (PSO) and genetic algorithm (GA) as population-based stochastic search algorithms were applied for training of ANNs and to achieve the optimum LSSVM model variables. For the purpose of predicting viscosity, the PSO-ANN and GA-LSSVM methods yielded the mean absolute error (MAE) and coefficient of determination (R²) values of 1.736 and 0.995 as well as 0.51930 and 0.99934, respectively for the whole data set, while for the purpose of predicting thermal conductivity, the PSO-ANN and GA-LSSVM models yielded the MAE and R² values of 1.43044 and 0.99704 as well as 0.72140 and 0.99857, respectively for the whole data set. Both methods provide properly capable method for predicting the thermal conductivity and viscosity of CO₂.     

GA辅助BP神经网络预测催化裂化装置汽油产率

催化裂化反应-再生系统是一个高度非线性和强耦合的操作系统,用传统建模方法很难描述。鉴于人工神经网络（ANN）非线性预测和自学习自适应能力强,而遗传算法（GA）全局寻优能力强的特点,将两者结合,先通过GA寻得BP神经网络最优的权值和阈值初值,再赋予BP,从而改善BP模型随机不确定选择初值的方法,提高其映射精度。以某炼油厂2.8 Mt/a MIP装置反应-再生系统为研究对象,选取第一反应器温度、第二反应器温度、第一再生器温度、第二再生器温度、反应器压力、再生器压力等6个变量为神经网络的输入变量,汽油产率为输出变量,建立6-11-1的BP神经网络,并采用GA来对BP神经网络的权值和阈值进行优化。结果表明,未经GA优化时BP神经网络对催化裂化汽油产率的预测数据的均方误差为5.16,而经GA优化后预测数据的均方误差为4.92。     

海上油气田水下立式采油树生产通道流动规律及结构优化

水下采油树中油气流动的通道很多,流体对生产通道壁面的冲蚀会导致其失效。借助CFD(计算流体力学)软件研究生产通道模型的冲蚀特征,并应用BP神经网络对CFD模拟得出的数据进行训练和预测,再将上述BP神经网络模型并入遗传算法GA中,从而建立起基于GA–BP的过流通道结构优化模型,以最大冲蚀率的极小值为优化目标对生产通道进行结构优化,以延长其使用寿命。     

Modeling of CO2-brine interfacial tension: Application to enhanced oil recovery

Development of reliable and accurate models to estimate carbon dioxide–brine interfacial tension (IFT) is necessary, since its experimental measurement is time-consuming and requires expensive experimental apparatus as well as complicated interpretation procedure. In the current study, feed forward artificial neural network is used for estimation of CO₂–brine IFT based on data from published literature which consists of a number of carbon dioxide–brine interfacial tension data covering broad ranges of temperature, total salinity, mole fractions of impure components and pressure. Trial-and-error method is utilized to optimize the artificial neural network topology in order to enhance its capability of generalization. The results showed that there is good agreement between experimental values and modeling results. Comparison of the empirical correlations with the proposed model suggests that the current model can predict the CO₂–brine IFT more accurately and robustly.     

Auto-Design of Neural Network–Based GAs for Manipulating the Khangiran Gas Refinery Sweetening Absorption Column Outputs

Abstract

In this work, a new approach for the auto-design of neural network-based genetic algorithm (GA) has been adopted to manipulate the products of an absorption column in the Khangiran gas refinery located in northeastern Iran. The experimental input data included gas flow rate, gas pressure, gas temperature, amine temperature, amine flow rate. In order to construct a GA–artificial neural network (ANN)-based model, the H₂S flow rate and dirty amine flow rate were selected for the output. The proposed method was assessed by the data taken from a case study in the Khangiran gas refinery. Design of topology and parameters of the neural networks as decision variables was first achieved by a trial-and-error procedure followed by genetic algorithms, which enhances the effectiveness of the forecasting scheme. The results reveal that the testing results from the model were in good agreement with the experimental data.     

基于BP神经网络和遗传算法的天然气脱水装置能耗优化

为了降低天然气脱水过程中的能量消耗和操作成本,应用工业流程模拟软件ProMax建立了天然气脱水装置工艺模型,并应用BP （Back Propagation）神经网络对ProMax模拟得出的数据进行训练和预测,再将上述BP神经网络模型并入遗传算法（Genetic Algorithms,GA）中,从而建立起了基于BP和GA的天然气脱水装置能耗优化模型。应用该能耗优化模型对川渝地区某净化厂内600×10⁴ m³/d天然气脱水装置进行了操作参数优化。结果表明：在保证净化气产品质量的前提下,该脱水装置的能耗降低了12.23%,经济效益明显提高。该方法通用性强,也可用于其他过程系统的操作参数优化。     

BP神经网络预测最小混相压力

应用BP神经网络预测CO2最小混相压力,选择C5＋分子量、油藏温度、挥发油（CH4和氮气）的摩尔分数、中间油（C2-C10）的摩尔分数作为参数,用相关文献的实验结果作为样本进行训练,选取网络模型各层函数、隐含层节点数和算法得出适合的BP神经网络,结合实际细管实验的数据及相关参数修改网络输入参数应用于实际油藏,预测最小混相压力并分析相关的影响因素,指导生产和相应理论研究。     

Modeling Asphaltene Precipitation Under Gas Injection and Pressure Depletion Conditions

Asphaltene precipitation is a major problem during primary oil production and enhanced oil recovery in the petroleum industry. In this work, a series of experiments was carried to determine the asphaltene precipitation of bottom hole live oil during gas injection and pressure depletion condition with Iranian bottom hole live oil sample, which is close to reservoir conditions using high pressure-high temperature equilibrium cell. In the majority of previous works, the mixture of recombined oil (mixture dead oil and associated gas) was used which is far from reservoir conditions. The used pressure ranges in this work covers wide ranges from 3 to 35 MPa for natural depletion processes and 24–45 MPa for gas injection processes. Also, a new approach based on the artificial neural network (ANN) method has been developed to account the asphaltene precipitation under pressure depletion/gas injection conditions and the proposed model was verified using experimental data reported in the literature and in this work. A three-layer feed-forward ANN by using the Levenberg-Marquardt back-propagation optimization algorithm for network training has been used in proposed artificial neural network model. The maximum mean square error of 0.001191 has been found. In order to compare the performance of the proposed model based on artificial neural network method, the asphaltene precipitation experimental data under pressure depletion/gas injection conditions were correlated using Solid and Flory-Huggins models. The results show that the proposed model based on artificial neural network method predicts more accurately the asphaltene precipitation experimental data in comparison to other models with deviation of less than 5%. Also, the number of parameters required for the ANN model is less than the studied thermodynamic models. It should be noted that the Flory and solid models can correlate accurately the asphaltene precipitation during methane injection in comparison with CO₂ injection.     

The Application of an Artificial Neural Network (ANN) and a Genetic Programming Neural Network (GPNN) for the Modeling of Experimental Data of Slim Tube Permeability Reduction by Asphaltene Precipitation in Iranian Crude Oil Reservoirs

Abstract

The authors introduce a new implementation of the neural network (ANN), genetic programming neural network (GPNN), and neuro-fuzzy (NF) technology in petroleum engineering. An intelligent framework is developed for calculating the amount of permeability reduction by asphaltene precipitation in Iranian crude oil reservoirs over a wide pressure, temperature, and solvent mole fraction range. Theoretical results and practical experience indicate that a feed-forward network can approximate a wide class of function relationships very well. In this work, a conventional feed-forward multilayer ANN, GPNN, and NF approach have been proposed to predict the amount of permeability reduction. The accuracy of the method is evaluated by predicting the amount of permeability reduction of various reservoir fluids not used in the development of the models. One of the ways in modeling such systems is using intelligent techniques, which need information about the systems, so, based on some intelligent learning methods, it can provide a suitable model. Furthermore, the performance of the model is compared with the performance of a simple model for permeability reduction prediction, a new correlation, and experimental data. Results of this comparison show that the proposed GPNN method first and then NF method is superior both in accuracy and generality, over the other models.     

Phase equilibrium modeling of semi-clathrate hydrates of the CO2+H2/CH4/N2+TBAB aqueous solution system

In this study, the semi-clathrate hydrate dissociation pressure for the CO₂+N₂, CO₂+H₂, CO₂+CH₄, and pure CO₂ systems in the presence of different concentrations of TBAB aqueous solutions is predicted using a strong machine learning technique of multi-layer perceptron neural network (MLP-NN). The developed model, with an overall correlation coefficient (R²) of 0.9961 and mean square error (MSE) of 5.96E?02, presented an excellent accuracy in prognosticating experimental data. A complete statistical evaluation performed to promise the strength and generality of the multi-layer perceptron artificial neural network (MLP-ANN). In addition, the applicability of the proposed network and quality of experimental data was assessed through the Leverage approach.     

利用BP人工神经网络预测天然气中重组分对净化装置的影响

由于天然气中的重组分会对脱硫装置的运行效果及产品气气质造成影响,这一实际生产问题需要得到有效的解决。在MDEA溶液吸收性能评价装置上测定了不同条件下的MDEA溶液吸收性能,系统地研究了不同重组分对MDEA溶液吸收性能的作用规律,采用多因素方差分析筛选了关键因素,以判定其影响程度的大小,并采用人工神经网络建立了天然气中重组分不利影响的预测模型。结果表明:天然气中的重组分i-C_5、C_6、C_7、C_8和C_(10)对MDEA溶液吸收能力具有十分显著的影响,它们均属于BP神经网络预测模型的有效输入信号,模型预测值与真实值较为近似,BP人工神经网络表现出良好的准确性和稳定性。因此,利用BP人工神经网络能够准确、可靠地预测天然气中重组分对MDEA溶液吸收性能的不利影响。     

Precise model for estimating CO<Subscript>2</Subscript>—oil minimum miscibility pressure

The CO₂—oil minimum miscibility pressure (MMP) is an important parameter for screening and selecting reservoirs for CO₂ injection projects. For the highest recovery, a candidate reservoir must be capable of withstanding an average reservoir pressure greater than the CO₂—oil MMP. Knowledge of the CO₂—oil MMP is also important when selecting a model to predict or simulate reservoir performance as a result of CO₂ injection. This paper, presents a new alternating conditional expectation “ACE”-based model for estimating CO₂—oil MMP. The ACE algorithm estimates the optimal transformation that maximizes the correlation between the transformed dependent variable “CO₂—oil MMP” and the sum of the transformed independent variables that represent reservoir temperature and different components of oil composition. Predicted values of the CO₂—oil MMP from the developed ACE-based model were compared with the experimental and calculated values from the most common correlations reported in the literature for CO₂—oil MMP prediction. The results showed that the ACE-based model is superior to other commonly used correlations. Regarding other correlations, the ACE-based model yielded the highest correlation coefficient (0.9878), the lowest average relative error (0.7428%), and the lowest standard deviation of error (1.2265). The text was submitted by the author in English.     

基于BP神经网络的低温提氦工艺优化

氦气是国家重要性战略物资之一,目前氦气的主要工业来源仍是从天然气中提取。为进一步优化低温提氦工艺,降低工艺能耗水平,对已有低温提氦工艺进行了改进,以一级提氦塔进料温度、压力、回流比、制冷剂高压、低压压力和制冷剂流量6个参数为变量,建立基于BP神经网络算法的综合能耗及提氦浓度预测模型,并对模型进行检验,并运用训练好的BP神经网络对改进工艺的综合能耗及粗氦浓度进行了预测。研究表明:BP模型训练效果较好,可用于综合能耗和粗氦体积分数的预测;通过训练误差分析,确定了模型隐藏层节点数为8时BP模型预测结果最优;利用确定好的BP神经网络预测出最优工艺生产参数,在满足粗氦体积分数不小于63.5%的基础上,综合能耗降低了18.08%。     

An accurate model to predict drilling fluid density at wellbore conditions

Knowledge about rheology of drilling fluid at wellbore conditions (High pressure and High temperature) is a need for avoiding drilling fluid losses through the formation. Unfortunately, lack of a universal model for prediction drilling fluid density at the addressed conditions impressed the performance of drilling fluid loss control. So, the main motivation of this paper is to suggest a rigorous predictive model for estimating drilling fluid density (g/cm³) at wellbore conditions. In this regard, a couple of particle swarm optimization (PSO) and artificial neural network (ANN) was utilized to suggest a high-performance model for predicting the drilling fluid density. Moreover, two competitive machine learning models including fuzzy inference system (FIS) model and a hybrid of genetic algorithm (GA) and FIS (called GA-FIS) method were employed. To construct and examine the predictive models the data samples of the open literature were used. Based on the statistical criteria the PSO-ANN model has reasonable performance in comparison with other intelligent methods used in this study. Therefore, the PSO-ANN model can be employed reliably to estimate the drilling fluid density (g/cm³) at HPHT condition.     

基于不同人工神经网络的催化裂化焦炭产率预测对比

在催化裂化装置（FCC）中,焦炭产率增加不但会使装置的总液收降低,而且会影响装置的热平衡,增加装置的操作难度。控制催化裂化装置焦炭产率十分重要,而其前提是能够准确预测装置的焦炭产率。催化裂化焦炭的生成和烧焦过程是一个连续的过程,影响参数众多且各参数之前互相影响,使用传统的方法建立多参数的预测模型具有一定的难度。本文利用人工神经网络(ANN)结合催化裂化装置的生产数据分别建立了GRNN神经网络预测模型和BP神经网络预测模型。对比分析结果表明,BP神经网络预测结果的准确度和稳定性优于GRNN神经网络。为进一步提高BP神经网络的预测效果,又分别使用了粒子群算法(PSO)和遗传算法(GA)对其进行优化。对比分析两种优化算法表明,两种优化算法均能提高BP神经网络的预测精度,综合考虑预测结果的准确性和稳定性两个方面,经遗传算法优化的BP神经网络预测模型优于经粒子群算法优化的BP神经网络预测模型。