A novel MLP-ANN approach to predict solution gas-oil ratio

Abstract

The better understanding and estimation of reservoir fluids properties have straight effects on accuracy of different processes such as simulation, well testing, and material balance calculations, so importance of accurate estimation of PVT properties such as solution gas-oil ratio is obvious. To this end, in the present paper, multilayer perceptron artificial neural network (MLP-ANN) is used as a novel predictive tool to estimate solution gas-oil ratio in terms of temperature, bubble point pressure, oil American Petroleum Institute gravity API, and gas specific gravity. Therefore, a total number of 1,137 experimental solution gas-oil ratios has been gathered from reliable databank for evaluation of model outputs. The different graphical and statistical analyses such as determination of average absolute relative deviation (AARD), the coefficient of determination (R²) and root mean square error (RMSE) were used to evaluate the performance of MLP-ANN algorithm. The comparisons show that MLP-ANN algorithm has great accuracy in prediction of solution gas-oil ratio, so it can be used as a simple tool to predict phase behavior of reservoir fluids.     

Applying ANFIS-PSO algorithm as a novel accurate approach for prediction of gas density

The accurate estimations of processes in gas engineering need a high degree of accuracy in calculations of gas properties. One of these properties is gas density which is straightly affected by pressure and temperature. In the present work, the Adaptive neuro fuzzy inference system (ANFIS) algorithm joined with Particle Swarm Optimization (PSO) to estimate gas density in terms of pressure, temperature, molecular weight, critical pressure and critical temperature of gas. In order to training and testing of ANFIS-PSO algorithm a total number of 1240 experimental data were extracted from the literature. The statistical parameters, Root mean square error (RMSE), coefficient of determination (R²) and average absolute relative deviation (AARD) were determined for overall process as 0.14, 1 and 0.039 respectively. The determined statistical parameters and graphical comparisons expressed that predicting mode is a robust and accurate model for prediction of gas density. Also the predicting model was compared with three correlations and obtained results showed the better performance of the proposed model respect to the others.     

Application of RBF-ANN in prediction of natural gas density in different operational conditions

Abstract

The degree of accuracy in prediction of different processes of gas engineering is extensively dependent on gas properties. One of the dominant properties which has straight effects on calculation and performance of different parts of gas industries is natural gas density. Due to this fact, in this paper, radial basis function artificial neural network (RBF-ANN) was used as novel approach to estimate gas density in terms of molecular weight, critical pressure and critical temperature of gas, pressure and temperature. To prepare and validate RBF-ANN model, a large and reliable experimental data bank was gathered from literature. A comprehensive analysis which include statistical and different graphical analysis were done to evaluate the performance. The coefficients of determination (R²) were determined as 0.99995 and 0.99993 for training and testing phases respectively. The comparisons illustrate that the RBF-ANN has great potential in prediction of natural gas density at different operational conditions.     

Utilization of LSSVM algorithm for estimating synthetic natural gas density

In the gas engineering the accurate calculation for pipeline and gas reservoirs requires great accuracy in estimation of gas properties. The gas density is one of major properties which are dependent to pressure, temperature and composition of gas. In this work, the Least squares support vector machine (LSSVM) algorithm was utilized as novel predictive tool to predict natural gas density as function of temperature, pressure and molecular weight of gas. A total number of 1240 experimental densities were gathered from the literature for training and validation of LSSVM algorithm. The statistical indexes, Root mean square error (RMSE), coefficient of determination (R²) and average absolute relative deviation (AARD) were determined for total dataset as 0.033466, 1 and 0.0025686 respectively. The graphical comparisons and calculated indexes showed that LSSVM can be considered as a powerful and accurate tool for prediction of gas density.     

Development of novel method for prediction of gas density in operational conditions

The performance of gas industries is extensively function of gas properties such as gas density. Due to this importance in the present work, a novel grid partitioning based fuzzy inference system method applied to predict gas density base on pressure, temperature and molecular weight of gas. To this end, the required experimental data are collected from reliable sources. Different comparison scenarios are used to evaluate the ability of model. The coefficients of determination (R²) for training and testing phases are calculated as 0.9985 and 0.9980 respectively. The determined indexes and graphical evaluations show that predicting model can estimate gas density in high degree of accuracy. According to the obtained results, the predicting model can be used as a simple and powerful software in gas industries to predict different processes.     

Application of Fuzzy C-means algorithm as a novel approach to forecast gas density in different conditions

One of the dominant parameters in accurate calculation and forecasting processes gas industries is accurate estimation of gas properties. The gas density is known as an effective parameter in gas processes calculations which affected by pressure and temperature. In the present paper, the Fuzzy c-means (FCM) algorithm is utilized as a novel predictive tool to estimate gas density as function of molecular weight, critical pressure and critical temperature of gas, pressure and temperature. In the purpose of training and testing of proposed FCM algorithm, a total number of 1240 measured data were gathered from reliable sources. The outputs of model and experimental data comparisons showed the great agreement between them such that the coefficients of determination for training and testing datasets were determined as 0.9982 and 0.9903 respectively. According to the obtained results from the graphical and statistical comparisons it can be concluded that the FCM algorithm has great ability and enough accuracy in prediction of gas density.     

Proposing a novel and accurate model for forecasting dynamic viscosity of n-alkane in operational conditions

Abstract

The viscosity of fluid is known as resistivity of fluid to flow and straightly affected by temperature and pressure. As it is obvious, the viscosity of reservoir fluid is known as one of the critical parameters which extensively effect on production. Therefore, in the present paper, multilayer perceptron artificial neural network (MLP-ANN) is used as a novel and accurate model to predict dynamic viscosity of normal alkanes in the operational conditions. To this end, 228 dynamic viscosity points as function of carbon number of n-alkane, temperature, and pressure were collected from a reliable paper. The comparison between MLP-ANN outputs and experimental dynamic viscosities is performed in graphical and statistical manners. The calculated coefficients of determination 0.99739 and 0.99051 for training and testing phases express the great ability of MLP-ANN algorithm in prediction of dynamic viscosity of n-alkane. According to the analysis, MLP-ANN has enough accuracy and potential to be used as software for which applicable in petroleum industry.     

Prediction of catalytic hydro conversion of normal heptane over catalysts using multi-layer perceptron artificial neural network (ANN-MLP)

Abstract

Catalytic reforming in the presence of metal-acid bifunctional catalysts is a widely used reaction in refinery industry to improve some properties of products like temperature performance of diesel and octane number of gasoline. So the ability of the prediction of Iso-C₇ selectivity during n-heptane hyroconversion is a key issue. In this study, a data set which was collected from previous publications are put in an artificial neural network-multi layer perceptron (MLP-ANN) model. Properties used as input parameters are: temperature, pressure, WHSV (weight hourly space velocity), catalysts acidity and pore volume of the catalysts, and Iso-C₇ selectivity used as the output parameter. Based on results, the MLP-ANN has great ability to estimate n-heptane hydroconversion. Root mean squared error (RMSE) and R-squared (R²) error were calculated for training, test and total set of data. For training set, test set and total set RMSE are 97915, 5.1607 and 3.9441, respectively and corresponding R² are 0.97915, 0.9334 and 0.9746 respectively.     

Quick estimation of gas pipeline inventory

Quick estimation of the pipeline inventory at the end of fiscal periods is common practice in companies dealing with the transmission of natural gas. The accuracy of line pack estimation depends on the accuracy of estimating the mean value of the gas density (or specific volume) which depends on composition and the distribution of pressure, temperature, and compressibility factor along the pipeline. An improved line pack estimation formula is developed for natural gas flowing in a horizontal pipeline. The derived formula takes into consideration the variation of velocity, pressure, temperature, and compressibility factor along the pipeline. Numerical results comparing the errors associated with line pack estimation on the basis of alternative approximations to the mean pressure, temperature, and compressibility factor are presented. The errors entailed on using different approximations depend on the gas composition, and the magnitude of the pressure and temperature ranges existing at the extremities of the pipe segment.     

Application of MLP-ANN as novel tool for estimation of effect of inhibitors on asphaltene precipitation reduction

Sedimentation of heavy fractions of oil such as asphaltene is the main concern in different parts of petroleum industries like production and transportation. Due to this fact, the inhibition of asphaltene precipitation becomes one of the great interests in the petroleum industry. In the present investigation, multi-layer perceptron artificial neural network (MLP-ANN) was developed to estimate asphaltene precipitation reduction as a function of concentration and kind of inhibitors and oil properties. To this end, a total number of 75 data points were extracted from reliable source for training and validation of predicting algorithm. The outputs of MLP-ANN were compared with experimental data graphically and statistically, the determined coefficients of determination (R²) for training and testing are 0.996522 and 0.995239 respectively. These comparisons expressed the high quality of this algorithm in the prediction of asphaltene precipitation reduction. so the MLP-ANN can be used as a powerful machine for estimation of different processes in petroleum industries.     

Viscosity and Density Correlations for Hydrocarbon Gases and Pure and Impure Gas Mixtures

Abstract

In this work, newly developed correlations for hydrocarbon gas viscosity and density are presented. The models were built and tested using a large database of experimental measurements collected through extensive literature search. The database covers gas composition, viscosity, density, temperature, pressure, pseudoreduced pressure and temperature and compressibility factor for different gases, and pure and impure gas mixtures containing high amount of pentane plus and small concentration of nonhydrocarbon components. Gas viscosity and gas density models were built with 800 randomly selected data points extracted from the large database. The models were developed using the Alternating Conditional Expectations (ACE) algorithm. The models' accuracy was validated using the rest of the database, and their efficiency was tested against some commonly used correlations. The developed models seemed very efficient and they accurately predicted the experimental viscosity and density measurements, overcoming several constraints limiting the other correlations' accuracy with average absolute errors of 3.95% and 4.93% for the gas viscosity and gas density models, respectively. Sensitivity analysis of the proposed gas viscosity model indicated the positive impact of density and pseudoreduced temperature and the trivial impact of pseudoreduced pressure. The gas density model was found to be sensitive to all input parameters of pseudoreduced temperature, apparent molecular weight, and pseudoreduced pressure listed on the order of their impact. Negative impact was predicted for reduced temperature, whereas positive ones werenoticed for the pseudoreduced pressure and gas apparent molecular weight.     

Forecasting gas density using artificial intelligence

Proper calculations of gas engineering require precise determination of gas properties and its associated variations with pressure and temperature. These properties can be determined by conducting experimental tests on gathered fluid samples from the bottom of the wellbore or at the surface as well as using equations of state and empirical correlations. This work is concentrated to develop a robust and quick model based on artificial network trained with teaching learning based optimization (ANN-TLBO) using 693 data sets at a wide range of pressure and temperature for gas density prediction. Comparing gas density from the predictive method and experimental results describe that the proposed ANN-TLBO model is of reliable accuracy for determining gas density. Sensitivity analysis also showed the extreme effect of temperature and pressure on gas density.     

A novel ANFIS model to prediction of the density of n-alkane in different operational condition

Abstract

The density has an important role in the oil and gas industries calculation. In this study, an adaptive neuro-fuzzy interference system (ANFIS) model was employed to predict the density of n-alkane. The result obtained by the ANFIS model analyzed with the statistical parameters (i.e., MSE, RMSE, and R²) and graphical method. According to the result obtained the ANFIS has the highest accuracy with R² = 0.999, MSE = 0.1438, and RMSE = 0.3792.     

Applying of LSSVM approach as a novel tool for accurate prediction of asphaltene inhibition efficiency

Asphaltene precipitation is one of critical problems for petroleum industries. There are different methods for inhibition of asphaltene precipitation. One of the common and effective methods for inhibition of asphaltene precipitation is utilizing asphaltene inhibitors. In this work, Least squares support vector machine (LSSVM) algorithm was coupled with simplex optimizer to create a novel and accurate tool for estimation of effect of inhibitors on asphaltene precipitation as function of concentration and structure of inhibitors and crude oil properties. To this end a total number of 75 measured data was extracted from the literature for training and testing of predicting model. The average absolute relative deviation (AARD), the coefficient of determination (R²) and root mean square error (RMSE) of total data for prediction algorithm were determined as 1.1479, 0.99406 and 0.61039. According to these parameters and graphical comparisons the LSSVM algorithm has potential to predict asphaltene precipitation in high degree of accuracy.     

Prediction of thermodynamic properties of natural gas mixtures using 10 equations of state including a new cubic two-constant equation of state

In this contribution, 10 equations of state (EoSs) are used to predict the thermo-physical properties of natural gas mixtures. One of the EoSs is proposed in this work. This EoS is obtained by matching the critical fugacity coefficient of the EoS to the critical fugacity coefficient of methane. Special attention is given to the supercritical behavior of methane as it is the major component of natural gas mixtures and almost always supercritical at reservoir and surface conditions. As a result, the proposed EoS accurately predicts the supercritical fugacity of methane for wide ranges of temperature and pressure. Using the van der Waals mixing rules with zero binary interaction parameters, the proposed EoS predicts the compressibility factors and speeds of sound data of natural gas mixtures with best accuracy among the other EoSs. The average absolute error was found to be 0.47% for predicting the compressibility factors and 0.70% for the speeds of sound data. The proposed EoS was also used to predict thermal and equilibrium properties. For predicting isobaric heat capacity, Joule–Thomson coefficient, dew points and flash yields of natural gas mixtures, the predictive accuracy of the EoS is comparable to the predictive accuracy of the Redlich–Kwong–Soave (RKS) EoS or one of its variants. For predicting saturated liquid density of LNG mixtures, however, the accuracy of predictions is between the RKS and Peng–Robinson (PR) EoSs.     

稠油注天然气混合密度变化规律研究及模型建立

稠油注气后的密度改变直接影响举升效率,为了提高塔河稠油开采效益,针对塔河油田超稠油开展了稠油注天然气混合密度研究。结果表明:稠油注天然气后混合密度随温度升高、注气比增加而降低,近似呈线性关系。在实验基础上,对应用较广泛的原油注气密度计算Obomanu模型进行了修正,建立了塔河原油注天然气密度模型;Obomanu模型的相关系数R²为0.670 3~0.769 2,修正模型的相关系数R²为0.985 2~0.998 9,修正后的密度模型拟合度提升了30%左右,适用于低压和高压环境;建立了原油注天然气平衡压力-原油密度-平衡气油比关系图版,可指导稠油注天然气开采。      

一种提高气井静压计算精度的方法

以往通过计算气井井口静压求取井底静压的诸多方法,由于未充分考虑随井深的增加,温度、压力的变化以及天然气中硫化氢、二氧化碳等酸性气体对偏差系数的影响,从而降低了其计算精度。文中根据温度、压力的变化范围,采用不同的天然气偏差系数计算模型,对含硫化氢高的酸性气体进行临界参数校正,以提高天然气偏差系数的计算精度,并在此基础上计算气井井底静压。通过实际资料验证,该方法能够比较精确地求取气井井底静压。     

Neural computing approach for predicting vaporization enthalpy of pure hydrocarbons and petroleum fractions

The current collaboration was aimed to approximate the heat of vaporization for petroleum fractions and pure hydrocarbons through using the multi-layer perceptron artificial neural network (MLP-ANN) based on the specific gravity, molecular weight, and boiling point temperature. Furthermore, Levenberg Marquardt algorithm was utilized to train the ANN structure and optimize its tuning parameters. Another comparison was carried out between the outcomes of suggested MLP-ANN model and six well-known correlations. Better results were observed for predicting heat of vaporization by the MLP-ANN model with the obtained value of mean relative error (MRE) and R-squared (R²) equal to 1.31% and 0.9962%, respectively. This computational approach can be applied in the petroleum engineering for a precise determination of heat of vaporization.     

Predicting the conditions for gas hydrate formation

Gas hydrate is a crystalline mixture obtained from gas molecules trapped in the cavity of hydrogen bonding water. To date, an essential step in the development of natural gas industry has been the acquisition of knowledge in the operation and handling of gas under high pressure without hydrate formation. Since there are several ways to predict hydrate formation, this study investigates predicting hydrate formation using the Katz method. In addition, several new models for accurate estimation of gas hydrate formation conditions will be provided. These models are based on artificial neural network (ANN) requirements. To create the model, predictive experimental data published in books and journals, as well as data extracted from Katz graph (Katz chart), estimate the formation conditions of gas hydrate. We validate the model created with the use of various statistical parameters such as mean squared error (MSE) and R²-value. The result of these parameters in models created to predict the formation of hydrates accurately and efficiently is evaluated. In this study, our goals are to use an artificial intelligence neural network to predict the formation of gas hydrates.     

Phase behavior modelling of asphaltene precipitation utilizing MLP-ANN approach

Since the sedimentation of heavy hydrocarbons such as asphaltenes, is the highlighted concern in production and operational, many studies were focused on this challenge in the petroleum industry. Therefore, the petroleum engineers should access to the asphaltene precipitation as an essential factor in order to conquer its problems. In this study, an empirical model for prediction asphaltene precipitation by multi-layer perceptron artificial neural network (MLP-ANN) is offered that takes the effect of the temperature, dilution ratio, and molecular weight for different n-alkanes. The output of this model showed 0.9999 for correlation coefficient (R²) and 0.000495 for mean squared error (MSE). This value illustrates the high quality of this model in compare of other available models. So far, MLP-ANN can offer significant accuracy in predicting asphaltene precipitation of asphaltene and other heavy oil.