Estimation of Petroleum Reservoir Parameters Using an Integrated Approach Neural Network,Principal Component Analysis and Fisher Discriminant Analysis

A new predictive methodology is introduced, based on a combined principal component analysis (PCA), Fisher discriminant analysis (FDA), and artificial neural network (ANN) methodologies for parameters estimation of a petroleum reservoir. Prediction of continuous petrophysical parameters is often time consuming and complicated because of geological variability such as facies changes due to sedimentary and structural changes. The petrophysical parameters, however, are usually difficult to measure due to reliability considerations, limitations insights on cost, inappropriate instrument maintenance, and sensor failures, evaluated by crude diagrams of reservoir parameters valuably. PCA and FDA provides an optimal lower dimensional representation in terms of discriminating among classes of data and are developed utilizing the reservoir historical data to incorporate reliability and prediction capabilities of ANN. The developed soft sensors are applied to predict the parameters of Marun reservoir located in Ahwaz, Iran, by utilizing the available geophysical well log data. The resulting outcomes demonstrate the promising capabilities of the proposed hybrid PCA-FDA-NN methodology than the conventional back-propagation NN, FDA, and PCA algorithms.     

The Estimation of Formation Permeability in a Carbonate Reservoir Using an Artificial Neural Network

Abstract

Reservoir permeability is an important parameter that its reliable prediction is necessary for reservoir performance assessment and management. Although many empirical formulas are derived regarding permeability and porosity in sandstone reservoirs, these correlations cannot be accurately depicted in carbonate reservoir for the wells that are not cored and for which there are no welltest data. Therefore, having a framework for estimation of these parameters in reservoirs with neither coring samples nor welltest data is crucial. Rock properties are characterized by using different well logs. However, there is no specific petrophysical log for estimating rock permeability; thus, new methods need to be developed to predict permeability from well logs. One of the most powerful tools that we applied by the authors is artificial neural network (ANN), whose advantages and disadvantages have been discussed by several authors. In particular, 767 data sets were used from five wells of Bangestan reservoir in a southwestern field of Iran. Depth, Neutron (NPHI), Density (RHOB), Sonic (DT) logs, and evaluated total porosity (PHIT) from log data were used as the input data and horizontal permeability obtained by coring was as target data. Sixty percent of these data points were used for training and the remaining for predicting the permeability (i.e., validation and testing). An appropriate ANN was developed and a correlation coefficient (R) of 0.965 was obtained by comparing permeability predictions and the actual measurements. As a result, the neural science can be used effectively to estimate formation permeability from well log data.     

应用神经网络计算储层参数

以岩心分析资料及多种测井信息为依据，首先利用样本信息的神经元模型（ＣＵＳＩ）解决了储层参数的计算问题，并利用改进后的自适应神经元模型（ＡＣＵＳＩ）提高了分析精度。最后利用前馈神经网络的误差反向传播模型（ＢＰ）网络的外延和信息表达能力解决了非储层的定量识别。应用上述方法对辽河油田四口井进行了逐步参数分析，分析结果与实际情况吻合很好。     

Permeability Estimation of the Reservoir Based on Particle Swarm Optimization Coupled with Artificial Neural Networks

Abstract

In this work we investigate how the integration of back-propagation (BP) with particle swarm optimization (PSO) improves the reliability and prediction capability of PSO. This strategy is applied to predict permeability in Mansuri Bangestan reservoir located in Ahwaz, Iran, utilizing available geophysical well log data. Our methodology utilizes a hybrid PSO–BP. The particle swarm optimization algorithm was shown to converge rapidly during the initial stages of a global search, but around global optimum, the search process will become very slow. On the contrary, the gradient descending method can achieve faster convergence speed around global optimum and with greater accuracy. The proposed algorithm combines the local search ability of the gradient-based BP strategy with the global search ability of particle swarm optimization. PSO is used to decide the initial weights of the gradient decent methods so that all of the initial weights can be searched intelligently. The experimental results show that the proposed hybrid PSO–BP algorithm is better than the PSO algorithm in convergence speed and accuracy.     

基于神经网络的柳林地区煤储层渗透率预测方法

分析了煤储层渗透率预测中存在的问题,提出基于测井信息的GA-BP神经网络预测煤储层渗透率方法,分析其机理及测井参数标准化处理方法。以柳林地区56口井的试井和测井资料为基础,利用灰色关联分析法优选6个测井参数作为输入变量,建立了GA-BP神经网络渗透率预测模型。将渗透率模型的预测结果与实测结果比较,两者具有较高的吻合度,证明该方法在煤储层参数预测的实践中具有较好的适应性。基于所建立的数学模型,对研究区的渗透率进行了预测,完成了渗透率平面分布图,为柳林地区煤层气的勘探开发提供了依据。     

地层真电阻率恢复神经网络法

水驱油田进入高含水阶段后,利用测井信息恢复地层真电经是一个关键问题。文中首先利用自回归线性预测技术消除地层厚度对测井曲线值的影响,然后从多条测井曲线提取有效特征,利用取心井训练神经网络,进而预测未知井,达到地层真电经恢复的目的。实验资料的处理结果表明,此法具有好的应用前景。     

人工神经网络在油藏数据训练中的应用

由于人工神经网络（ANN）较传统的模式识别方法更具优越性,可以应用到油藏数据的训练中。采用3种神经网络训练函数,即,网络的权值和阈值训练函数、利用Levenberg-Marquardt规则训练函数和阶梯变化梯度算法,分别对油藏数据进行训练,得到成功的训练结果,从而把预测的油藏参数和三维（3D）地震属性联系起来。实验表明,把ANN用于训练油藏数据是可行的。     

基于BP神经网络的岩石可钻性测井计算研究

利用测井资料求取岩石的可钻性是一种简便可行的途径.常规的岩石可钻性测井预测模型都是基于回归分析而建立起来的,形式简单,精度不高.介绍了一种基于BP神经网络利用测井资料求取岩石可钻性的实用方法.该方法从测井信息与岩石可钻性的内在关系出发,选用与岩石可钻性密切相关的多个测井参数,通过BP神经网络技术建立利用测井资料准确求取岩石可钻性的数学模型;将该方法用于SC油田DU4井等多口井的测井资料处理中,为邻井及时提供了较为准确的地层岩石可钻性剖面,也为该区的新井钻头选型提供了较好的依据.     

Flow Zone Indicator Estimation Based on Petrophysical Studies Using an Artificial Neural Network in a Southern Iran Reservoir

Porosity is a volumetric parameter whereas permeability is a measure of a rock's flow properties and depends on pore distribution and connectivity. Thus zonation of a reservoir using flow zone indicator (FZI) can be used to evaluate reservoir quality based on porosity-permeability relationships. The objective of this study was to develop an accurate reservoir FZI with the aid of artificial neural network (ANN) utilizing available geophysical well log data and dipole sonic imager (DSI) derived body wave velocities. The efficiency of utilizing shear wave and compressional wave velocities (V_p and V_s ) in improving estimation accuracy has been evaluated as well. It is the core data were used for ANN training that involves the calculations of Reservoir Quality Index, normalized porosity (? _z ), and FZI. Correlation between FZI calculated from core data and that obtained from well log data showed that ANN model were successful for estimation of FZI from conventional well log data. The compressional wave velocity was more effective than shear ones in delivering more accurate responses to estimate FZI. On the other hand, in association with other logs, utilizing compressional and shear wave velocities caused the responses to be closer to the reality and decrease the estimation error.     

相似度-遗传神经网络在储层物性预测中的应用

传统的测井解释需要建立精确的数学模型,并常伴有严格的条件限制,因此很难得到真实反映储层特性的结果.采用遗传算法与BP神经网络相结合,利用遗传算法的全局寻优特点,优化神经网络的连接权值和阀值,提高网络的训练精度和预测精度,避免了BP算法易陷入局部极小的缺点,提高运算速度.将相似度的概念引入到测井中,定义相似度在测井中的计...     

基于主成分分析的SOM神经网络在火山岩岩性识别中的应用

针对火山岩储层岩性识别难的问题,提出一种将主成分分析和SOM神经网络相结合对测井资料进行处理的岩性识别方法.主成分分析能较好地提取表征样本的少数几个独它的综合指标,从而能够消除神经网络输入间的相关性,降低神经网络的输入维数,简化网络结构,加速网络收敛速度,从整体上提高网络的性能.针对松辽盆地徐家围子地区内有薄片分析及全岩分析的325块岩样,单独使用主成分分析方法的岩性识别正确率为79.38%,单独使用自组织神经网络方法的岩性识别正确率为82.15%,结合上述2种方法的岩性识别正确率为87.38%.由此在实际处理20口井火山岩层段时,将原始测井数据通过主成分分析进行精简处理,然后再通过SOM神经网络进行识别分类,最终厚度符合率为85.2%,从而为利用常规测井资料识别火山岩岩性又提供一种有效方法.     

新型Elman网络在乙烯精馏塔软测量建模中的应用

针对前馈神经网络在石油化工软测量建模中存在的不足，采用递归型Elman网络实现动态石化生产过程的建模。提出了一种新型的Elman网络——HF Elman网络，将其应用在某石化厂600kt／a乙烯生产装置乙烯质量指标的软测量建模，并与前馈网络和Elman网络的建模性能进行了比较，仿真结果表明，HF Elman网络具有良好的动态建模能力，能够更好地实现乙烯精馏塔出口成分含量的软测量建模。     

联合神经网络在储层参数预测中的研究与应用

地质储层参数在建立地质模型中起着关键作用,储层参数通过井资料获得。常规测井解释中多通过经验公式或简化地质条件建立模型计算储层参数。提出了新的神经网络模型,基于BP神经网络、RBF神经网络、支持向量回归并通过单层感知器共同构成联合神经网络模型。该网络模型在储层参数预测过程中能针对单一神经网络的不足而自适应调节网络结构,使预测效果达到最优,避免了单一网络在参数预测时的缺点,提高了预测的准确性。选取了同一地区的3口油井进行训练和验证实验,实验结果表明,联合神经网络模型优于单一的人工神经网络模型。     

基于ICA、PCA与SVM方法的沉积微相定量识别

利用测井资料快速准确地确定沉积微相是油田勘探开发中急需研究解决的问题.沉积微相特征在测井曲线上有所反映,将常规测井资料及其解释成果中的地质资料同岩心资料相结合,分别通过主成分分析(PCA)和独立分量分析(ICA)提取反映沉积微相变化的特征,利用支持向量机(SVM)建立沉积微相的判别模型,根据该模型对未取心井段的沉积微相...     

神经网络及模式识别技术在测井解释中的应用

针对四川碳酸盐岩储层实际情况，用测井资料与岩心分析结果组成孔隙度、含水饱和度样本，经BP网络训练后得到模型参数。使用模型参数进行孔隙度、含水饱和度计算、计算结果精度高。BP网络在产能评价方面也有明显优势，能较好地表达储层参数与产能之间的关系，提高预测结果精度。对汉明网络结构作了适当的调整，使其适用于输入为连续值的模式识别问题。在储层流体性质判别方面，气层、水层的判别符合率达93%；该网络在多个地区的测井相分析中应用表明，能提高沉积微相识别率和预测符合率。实际应用证实，神经网络技术能提高测井解释中的数值计算精度和模式识别符合率，已在测井精细解释及储量计算中应用。     

模糊神经网络推理系统定量计算储层参数

储层参数是地层评价的基础。自适应模糊神经网络推理系统是综合了模糊逻辑与人工神经网络两者优势的一种人工智能方法．该方法既可用于模式分类．又可进行连续计算。这是一种基于数据的建模方法,模糊隶属度函数及模糊规则是通过大量已知数据的学习得到的．而不是按经验或直觉信息给定的。这对于那些特性还不为人们所了解或者特性非常复杂的系统尤其重要．因此特别适于复杂储层评价中的定性解释与定量计算。研究结果表明,这种方法具有较高的计算精度。     

波阻抗约束反演及储层物性参数计算方法

简要地介绍了宽带约束反演的基本实现思路，并着重介绍应用人工神经网络方法将约束反演得到的波阻抗转换成储层物性参数的方法。在储层参数计算中，应用了ＲＢＦＮ网络求取预测函烽，并对预测函数进行显著性估计。该方法在塔里木盆地塔北牙哈下第三系储层分析中应用效果良好。     

A Hybrid Symbiotic Evolution–Neuro Fuzzy Strategy for Permeability Estimation of the Reservoir

A key parameter for reservoir characterization is permeability distribution. Most well log data and core permeability values are corrupted by noise (such as uncertain depth-matching, core testing conditions, and thin beddings). In this work, the authors first used wavelet as a new powerful tool for de-noising data points and then they investigated how the integration of back propagation with group based symbiotic evolution improves the reliability and prediction capability of neuro-fuzzy systems for predicting permeability of real reservoir data.     

基于模糊神经网络的甲醇合成塔转化率软测量模型

以某甲醇合成塔为研究对象,综合考虑甲醇生产过程各因素对甲醇转化率的影响,采用模糊神经网络技术,建立了甲醇转化率的软测量模型。为了提高模糊神经网络建模的精度,提出了一种改进的差分进化算法优化模糊神经网络的参数,从而获得了具有较高精度的软测量模型。以某甲醇合成塔的实际运行数据为样本进行甲醇转化率的软测量建模,结果表明该模型具有较高的精度,能够实现甲醇转化率的实时监测。     

Body Wave Velocities Estimation From Wireline Log Data Utilizing an Artificial Neural Network for a Carbonate Reservoir,South Iran

Reservoir characterization is a prerequisite study for oil and gas field development. Body wave velocities are important parameters for reservoir characterization studies. In this research, a back-propagation artificial neural network (BP-ANN) including the Levenberg-Marquardt training algorithm was used as an intelligent tool to estimate compressional and shear wave velocities. The efficiency of utilizing density log and photoelectric effect (PEF) in improving estimation accuracy have been evaluated as well. The petrophysical data from three wells were used for constructing intelligent models in the South Pars field, Southern Iran. The fourth and fifth wells from the field were used to evaluate the reliability of the model. The results showed that a BP-ANN was successful in estimating body wave velocities and so when just gamma ray, neutron, deep resistivity (lateral log deep) were used as net work inputs, the net exactness ware comparatively low but using PEF effects increased this exactness. By using density log the net exactness noticeably grew and in this manner using both PEF and density log beside other mentioned logs as inputs approached to more real results.