A general approach for porosity estimation using artificial neural network method: a case study from Kansas gas field

This study aims to design a back-propagation artificial neural network (BP-ANN) to estimate the reliable porosity values from the well log data taken from Kansas gas field in the USA. In order to estimate the porosity, a neural network approach is applied, which uses as input sonic, density and resistivity log data, which are known to affect the porosity. This network easily sets up a relationship between the input data and the output parameters without having prior knowledge of petrophysical properties, such as porefluid type or matrix material type. The results obtained from the empirical relationship are compared with those from the neural network and a good correlation is observed. Thus, the ANN technique could be used to predict the porosity from other well log data.     

Committee neural networks for porosity and permeability prediction from well logs

Neural computing has moved beyond simple demonstration to more significant applications. Encouraged by recent developments in artificial neural network (ANN) modelling techniques, we have developed committee machine (CM) networks for converting well logs to porosity and permeability, and have applied the networks to real well data from the North Sea. Simple three‐layer back‐propagation ANNs constitute the blocks of a modular system where the porosity ANN uses sonic, density and resistivity logs for input. The permeability ANN is slightly more complex, with four inputs (density, gamma ray, neutron porosity and sonic). The optimum size of the hidden layer, the number of training data required, and alternative training techniques have been investigated using synthetic logs. For both networks an optimal number of neurons in the hidden layer is in the range 8–10. With a lower number of hidden units the network fails to represent the problem, and for higher complexity overfitting becomes a problem when data are noisy. A sufficient number of training samples for the porosity ANN is around 150, while the permeability ANN requires twice as many in order to keep network errors well below the errors in core data. For the porosity ANN the overtraining strategy is the suitable technique for bias reduction and an unconstrained optimal linear combination (OLC) is the best method of combining the CM output. For permeability, on the other hand, the combination of overtraining and OLC does not work. Error reduction by validation, simple averaging combined with range‐splitting provides the required accuracy. The accuracy of the resulting CM is restricted only by the accuracy of the real data. The ANN approach is shown to be superior to multiple linear regression techniques even with minor non‐linearity in the background model.     

Estimating thermal conductivity from core and well log data

The aim of the presented work was to introduce a method of estimating thermal conductivity using well log data. Many petrophysical properties of rocks can be determined both by laboratory measurements and well-logs. It is thus possible to apply geophysical data to empirical models based on relationships between laboratory measured parameters and derive continuous thermal conductivity values in well profiles. Laboratory measurements were conducted on 62 core samples of Meso-Paleozoic rocks from the Carpathian Foredeep. Mathematical models were derived using multiple regression and neural network methods. Geophysical data from a set of seven well logs: density, sonic, neutron, gamma ray, spectral gamma ray, caliper and resistivity were applied to the obtained models. Continuous thermal conductivity values were derived in three well profiles. Analysis of the obtained results shows good consistence between laboratory data and values predicted from well log data.     

3D inversion of DC data using artificial neural networks

In this paper, we investigate the applicability of artificial neural networks in inverting three-dimensional DC resistivity imaging data. The model used to produce synthetic data for training the artificial neural network (ANN) system was a homogeneous medium of resistivity 100 Ωm with an embedded anomalous body of resistivity 1000 Ωm. The different sizes for anomalous body were selected and their location was changed to different positions within the homogeneous model mesh elements. The 3D data set was generated using a finite element forward modeling code through standard 3D modeling software. We investigated different learning paradigms in the training process of the neural network. Resilient propagation was more efficient than any other paradigm. We studied the effect of the data type used on neural network inversion and found that the use of location and the apparent resistivity of data points as the input and corresponding true resistivity as the output of networks produces satisfactory results. We also investigated the effect of the training data pool volume on the inversion properties. We created several synthetic data sets to study the interpolation and extrapolation properties of the ANN. The range of 100–1000 Ωm was divided into six resistivity values as the background resistivity and different resistivity values were also used for the anomalous body. Results from numerous neural network tests indicate that the neural network possesses sufficient interpolation and extrapolation abilities with the selected volume of training data. The trained network was also applied on a real field dataset, collected by a pole-pole array using a square grid (8 ×8) with a 2-m electrode spacing. The inversion results demonstrate that the trained network was able to invert three-dimensional electrical resistivity imaging data. The interpreted results of neural network also agree with the known information about the investigation area.     

松辽盆地营城组玄武岩流动单元测井响应特征

流动单元是玄武岩地层的最基本组成单元,其内部分带性控制储层的储集性能和有效储层的分布位置.运用钻井岩心资料建立了玄武岩流动单元分带地质模式,单个流动单元由上而下依次为上部气孔带、中部致密带和下部气孔带.依据自然伽马(GR)、声波时差(DT)、补偿密度(RHOB)、深侧向电阻率(LLD)和中子孔隙度(NPHI)分析流动单...     

A non-linear regularized constrained impedance inversion

Inversion for seismic impedance is an inherently complicated problem. It is ill‐posed and band‐limited. Thus the inversion results are non‐unique and the process is unstable. Combining regularization with constraints using sonic and density log data can help to reduce these problems. To achieve this, we developed an inversion method by constructing a new objective function, including edge‐preserving regularization and a soft constraint based on a Markov random field. The method includes the selection of proper initial values of the regularization parameters by a statistical method, and it adaptively adjusts the regularization parameters by the maximum likelihood method in a fast simulated‐annealing procedure to improve the inversion result and the convergence speed. Moreover, the method uses two kinds of regularization parameter: a ‘weighting factor’λ and a ‘scaling parameter’δ. We tested the method on both synthetic and field data examples. Tests on 2D synthetic data indicate that the inversion results, especially the aspects of the discontinuity, are significantly different for different regularization functions. The initial values of the regularization parameters are either too large or too small to avoid either an unstable or an over‐smoothed result, and they affect the convergence speed. When selecting the initial values of λ, the type of the regularization function should be considered. The results obtained by constant regularization parameters are smoother than those obtained by adaptively adjusting the regularization parameters. The inversion results of the field data provide more detailed information about the layers, and they match the impedance curves calculated from the well logs at the three wells, over most portions of the curves.     

Porosity and permeability prediction from wireline logs using artificial neural networks: a North Sea case study

Estimations of porosity and permeability from well logs are important yet difficult tasks encountered in geophysical formation evaluation and reservoir engineering. Motivated by recent results of artificial neural network (ANN) modelling offshore eastern Canada, we have developed neural nets for converting well logs in the North Sea to porosity and permeability. We use two separate back-propagation ANNs (BP-ANNs) to model porosity and permeability. The porosity ANN is a simple three-layer network using sonic, density and resistivity logs for input. The permeability ANN is slightly more complex with four inputs (density, gamma ray, neutron porosity and sonic) and more neurons in the hidden layer to account for the increased complexity in the relationships. The networks, initially developed for basin-scale problems, perform sufficiently accurately to meet normal requirements in reservoir engineering when applied to Jurassic reservoirs in the Viking Graben area. The mean difference between the predicted porosity and helium porosity from core plugs is less than 0.01 fractional units. For the permeability network a mean difference of approximately 400 mD is mainly due to minor core-log depth mismatch in the heterogeneous parts of the reservoir and lack of adequate overburden corrections to the core permeability. A major advantage is that no a priori knowledge of the rock material and pore fluids is required. Real-time conversion based on measurements while drilling (MWD) is thus an obvious application.     

Deriving effects of pressure depletion on elastic framework moduli from sonic logs

We have developed a method to determine the effect of pore pressure depletion on elastic framework moduli, by utilizing sonic logs from wells drilled at different locations through a reservoir at varying depletion stages. This is done by first inverting the sonic logs for elastic framework bulk and shear moduli, thus carefully removing pressure dependent fluid effects. By crossplotting these elastic framework moduli against an increase in net stress (which is directly related to depletion), we derive the stress sensitivity of the elastic framework moduli. We found that the observed stress sensitivity was consistent with time-lapse seismic results and that the sensitivity derived from the sonic logs was much smaller than predicted by hydrostatic measurements on core samples. This method is applicable to depletion scenarios in mature fields and can be used both for modelling and inverting time-lapse seismic data for effects of pore pressure depletion on seismic data.     

谐振Q理论井震匹配法及品质因子反演

储层勘测中地震波速度一般低于声波测井速度,撇开观测系统与人为因素造成的误差,岩石本征的黏弹性是造成这一现象的重要原因.本文在井、震匹配问题中引入了黏弹性岩石的杨氏谐振品质因子(Q)模型,对储层岩石进行了井震匹配与Q反演.文章阐述了如何通过谐振Q模型合理的校正声波速度,从而实现井震匹配.在波形匹配的基础上,进行地层品质因子的反演,合理的计算出目的层位的品质因子值.     

Deriving TDS values in coarse sediments from long normal and electromagnetic logs

This paper presents an empirical relationship of quantitatively linked electromagnetic (EM) borehole recordings of the total dissolved solids (TDS) in pore water in the Quaternary deposits of the Belgian coastal plain. First, the long normal (LN) logs are linked to EM logs, then the already developed relationships between LN resistivity measurements and the TDS values are rewritten for EM recordings. The main parameter in these equations is the formation factor, which is derived from ground water analyses and LN logs through Archie's law. The EM recording has several advantages compared to the LN logs. The EM analysis allows measuring in PVC-cased wells and is not hindered by the invasion zone around the well. Furthermore, it has a high vertical resolution. LN logs can be measured only once, after drilling a well; EM recordings can be repeated several times in monitoring wells, which allows the gathering of time-dependent data over a complete vertical cross section. Such data could be obtained with LN logs only in wells with screens over the full-depth interval, which causes a hydraulic short circuit. This short circuit can result in a large artificial flow through the well between different levels, resulting in a salinity profile, which is no longer representative for the studied site. Remediation against short circuiting is a reduction of the screened interval, which strongly reduces the gathered information. The application of the derived equations is one of setting up a monitoring network along the Belgian coast to monitor the trend in salinity levels and comparing present salinity levels with older LN recordings to investigate the salinity changes in the last 30 years. Deep wells already present in the Belgiancoastal plain can then be used to monitor both the fresh water head changes and the salt water evolution. The technique has also been used for parameter identification for which real concentration measurements were needed.     

Deep neural network application for 4D seismic inversion to changes in pressure and saturation: Optimizing the use of synthetic training datasets

In this work, we tackle the challenge of quantitative estimation of reservoir dynamic property variations during a period of production, directly from four-dimensional seismic data in the amplitude domain. We employ a deep neural network to invert four-dimensional seismic amplitude maps to the simultaneous changes in pressure, water and gas saturations. The method is applied to a real field data case, where, as is common in such applications, the data measured at the wells are insufficient for properly training deep neural networks, thus, the network is trained on synthetic data. Training on synthetic data offers much freedom in designing a training dataset, therefore, it is important to understand the impact of the data distribution on the inversion results. To define the best way to construct a synthetic training dataset, we perform a study on four different approaches to populating the training set making remarks on data sizes, network generality and the impact of physics-based constraints. Using the results of a reservoir simulation model to populate our training datasets, we demonstrate the benefits of restricting training samples to fluid flow consistent combinations in the dynamic reservoir property domain. With this the network learns the physical correlations present in the training set, incorporating this information into the inference process, which allows it to make inferences on properties to which the seismic data are most uncertain. Additionally, we demonstrate the importance of applying regularization techniques such as adding noise to the synthetic data for training and show a possibility of estimating uncertainties in the inversion results by training multiple networks.     

Research Note: Transverse isotropy estimation from dipole sonic logs acquired in pilot and production wells

Dipole sonic logs acquired in near‐vertical pilot wells and over the build section of nearby horizontal production wells are inverted to determine the five elastic constants characterizing a transversely isotropic formation, under the assumption of lateral homogeneity. Slowness values from a single depth in the vertical well are combined with data from the corresponding depth in the deviated well; these data are then inverted using nonlinear optimization to derive the five elastic constants. The technique is demonstrated on data from the Haynesville Shale in Texas. Estimates of the anisotropy are in line with a priori expectations; the Thomsen ε and γ parameters are well correlated and generally possess positive anellipticity, with Thomsen's ε greater than Thomsen's δ.     

由声测井资料计算反射系数序列中多分辨逼近的选择

介绍了由声测井资料计算反射系数序列中的多分辨率逼近理论,详细分析了由高分辨率的声测井资料计算低分辨率的反射系数序列的地质要求,进而在理论上讨论了此类问题中多分辨逼近的选择,并比较了由Ｈａａｒ系与由三次样条函数构造的多分辨率逼近这两种典型的多分辨逼近。最后利用理论模型及实际资料算例验证了选择恰当的多分辨率逼近在解决这类问题中的有效性。     

Well tie,fluid substitution and AVO modelling: a North Sea example

Accurate well ties are essential to practical seismic lithological interpretation. As long as the geology in the vicinity of the reservoir is not unduly complex, the main factors controlling this accuracy are the processing of the seismic data and the construction of the seismic model from well logs. This case study illustrates how seismic data processing to a near-offset stack, quality control of logs and petrophysical modelling improved a well tie at an oil reservoir. We demonstrate the application of a predictive petrophysical model in the preparation and integration of the logs before building the seismic model and we quantify our improvements in well-tie accuracy. The data for the study consisted of seismic field data from a 3D sail line through a well in a North Sea oilfield and a suite of standard logs at the well. A swathe of fully processed 3D data through the well was available for comparison. The well tie in the shallow section from first-pass seismic data processing and a routinely edited sonic log was excellent. The tie in a deeper interval containing the reservoir was less satisfactory: the phase errors within the bandwidth of the seismic wavelet were of the order of 20°, which we consider too large for subsequent transformation of the data to seismic impedance. Reprocessing the seismic data and revision of the well-log model reduced these phase errors to less than 10° and improved the consistency of the deep and shallow well ties. The reprocessing included densely picked iterative velocity analysis, prestack migration, beam-forming multiple attenuation, stacking the near-offset traces and demigration and remigration of the near-offset data. The petrophysical model was used to monitor and, where necessary, replace the P-wave sonic log with predictions consistent with other logs and to correct the sonic log for mud-filtrate invasion in the hydrocarbon-bearing sand. This editing and correction of the P-wave transit times improved the normal-incidence well tie significantly. The recordings from a monopole source severely underestimated the S-wave transit times in soft shale formations, including the reservoir seal, where the S-wave velocity was lower than the P-wave velocity in the drilling mud. The petrophysical model predicted an S-wave log that matched the valid recordings and interpolated between them. The subsequent seismic modelling from the predicted S-wave log produced a class II AVO anomaly seen on the CDP gathers around the well.     

Artificial neural networks for removal of couplings in airborne transient electromagnetic data

Modern airborne transient electromagnetic surveys typically produce datasets of thousands of line kilometres, requiring careful data processing in order to extract as much and as reliable information as possible. When surveys are flown in populated areas, data processing becomes particularly time consuming since the acquired data are contaminated by couplings to man‐made conductors (power lines, fences, pipes, etc.). Coupled soundings must be removed from the dataset prior to inversion, and this is a process that is difficult to automate. The signature of couplings can be both subtle and difficult to describe in mathematical terms, rendering removal of couplings mostly an expensive manual task for an experienced geophysicist. Here, we try to automate the process of removing couplings by means of an artificial neural network. We train an artificial neural network to recognize coupled soundings in manually processed reference data, and we use this network to identify couplings in other data. The approach provides a significant reduction in the time required for data processing since one can directly apply the network to the raw data. We describe the neural network put to use and present the inputs and normalizations required for maximizing its effectiveness. We further demonstrate and assess the training state and performance of the network before finally comparing inversions based on unprocessed data, manually processed data, and artificial neural network automatically processed data. The results show that a well‐trained network can produce high‐quality processing of airborne transient electromagnetic data, which is either ready for inversion or in need of minimal manual processing. We conclude that the use of artificial neural network scan significantly reduce the processing time and its costs by as much as 50%.     

DETECTION OF THIN BEDS WITH THE PSEUDO-ACOUSTIC IMPEDANCE SECTION*

The conventional seismic response of a thin bed approximates the time derivative of the incident wavelet, whereas the pseudo-impedance response approximates the incident wavelet. Consequently the pseudo-impedance response of a geological sequence composed of thin beds is simpler and easier to interpret than the conventional response. By calibrating the sonic log data with check-shot data and performing zero-phase seismic processing, the fit of the sonic log and pseudo-velocity section is improved. Discrepancies in amplitude and phase, however, generally remain. A five-step processing and interpretation procedure, which benefits from multichannel interpretation along the model seismic section generated from the sonic logs, is described. The method has been tested with field data. In the test the detection of thin beds and the estimation of the natural gas content was more reliable with the proposed procedure than with the conventional method.     

Monitoring Well Completion Evaluation with Borehole Geophysical Density Logging

Grout continuity and the location of the bentonite seal and sand pack in PVC-cased monitoring wells can be evaluated with cased-hole geophysical density logs. This method relies upon density contrasts among various completion conditions and annular materials. Notably, the lack of annular material behind pipe (i.e., void space) creates a low-density zone that is readily detected by borehole density measurements.
Acoustic cement bond logging has typically been applied to the evaluation of cement in the annular space of completed oil and gas production wells, and in some cases to ground water monitoring wells. These logs, however, can only be obtained in the fluid-filled portion of the borehole, and their interpretation is severely hindered by the presence of the micro-annulus between casing and cement. The influence of the micro-annulus on cement bond logs can be mitigated in steel-cased wells by pressurizing the wellbore during acquisition of the log, but this procedure is not feasible in PVC-cased monitoring wells. The micro-annulus does not affect cased-hole density logs or their interpretation.
Empirical measurements made in the laboratory with density probes provide information on their depths of investigation and response to specific completion conditions. These empirical data, and general knowledge of the density of annular completion materials (sand, bentonite, cement), are used to support interpretations of cased-hole density logs acquired in the field. Three field examples demonstrate the applicability of geophysical density logs to the evaluation of PVC-cased monitoring well completions.     

NMR Parameters Determination through ACE Committee Machine with Genetic Implanted Fuzzy Logic and Genetic Implanted Neural Network

Free fluid porosity and rock permeability, undoubtedly the most critical parameters of hydrocarbon reservoir, could be obtained by processing of nuclear magnetic resonance (NMR) log. Despite conventional well logs (CWLs), NMR logging is very expensive and time-consuming. Therefore, idea of synthesizing NMR log from CWLs would be of a great appeal among reservoir engineers. For this purpose, three optimization strategies are followed. Firstly, artificial neural network (ANN) is optimized by virtue of hybrid genetic algorithm-pattern search (GA-PS) technique, then fuzzy logic (FL) is optimized by means of GA-PS, and eventually an alternative condition expectation (ACE) model is constructed using the concept of committee machine to combine outputs of optimized and non-optimized FL and ANN models. Results indicated that optimization of traditional ANN and FL model using GA-PS technique significantly enhances their performances. Furthermore, the ACE committee of aforementioned models produces more accurate and reliable results compared with a singular model performing alone.     

Nonlinear inversion of electrical resistivity imaging using pruning Bayesian neural networks

Conventional artificial neural networks used to solve electrical resistivity imaging (ERI) inversion problem suffer from overfitting and local minima. To solve these problems, we propose to use a pruning Bayesian neural network (PBNN) nonlinear inversion method and a sample design method based on the K-medoids clustering algorithm. In the sample design method, the training samples of the neural network are designed according to the prior information provided by the K-medoids clustering results; thus, the training process of the neural network is well guided. The proposed PBNN, based on Bayesian regularization, is used to select the hidden layer structure by assessing the effect of each hidden neuron to the inversion results. Then, the hyperparameter α _k , which is based on the generalized mean, is chosen to guide the pruning process according to the prior distribution of the training samples under the small-sample condition. The proposed algorithm is more efficient than other common adaptive regularization methods in geophysics. The inversion of synthetic data and field data suggests that the proposed method suppresses the noise in the neural network training stage and enhances the generalization. The inversion results with the proposed method are better than those of the BPNN, RBFNN, and RRBFNN inversion methods as well as the conventional least squares inversion.     

塔河油田碳酸盐岩缝洞型储层的测井识别与评价方法研究

塔河油田奥陶系以碳酸盐岩为主，油气的主要储渗空间为裂缝和溶蚀孔洞，具有很强的非均质性．本文利用常规及成像测井资料，对碳酸盐岩缝洞型储层的识别与评价方法进行研究．为了综合各种测井方法识别裂缝，建立了综合裂缝概率模型，计算综合裂缝概率指示裂缝的发育程度．利用地层微电阻率扫描成像测井资料进行裂缝和溶蚀孔洞的定性、定量解释．定量计算的裂缝参数为：裂缝密度、裂缝长度、裂缝平均宽度、平均水动力宽度、裂缝视孔隙度；定量计算的溶蚀孔洞参数有：面孔率、孔洞密度．根据缝洞型储层孔隙空间类型及其中子孔隙度、补偿密度、声波、双侧向电阻率的测井响应物性特征，建立缝洞型碳酸盐岩储层复杂孔隙介质解释模型，用于确定裂缝、溶蚀孔洞孔隙度和评价储层．