Porosity Logs Research Articles

Acoustic televiewer amplitude data for porosity estimation with application to porewater conversion

The ability to detect and quantify rock matrix porosity is explored using acoustic televiewer images and dolostone rock core samples. In particular, an empirical relationship between the acoustic televiewer amplitude response and core-derived total porosity is developed. The performance of the acoustic televiewer amplitude relationship was compared to neutron and gamma-gamma logs, which are traditional nuclear wireline tools commonly used for estimating porosity in dolostones. Acoustic televiewer amplitude results are consistent with the gamma-gamma (density) and neutron (porosity) logs, provide increased vertical resolution, do not require downhole radioactive sources and also provide oriented fracture information. Sample specific porosity and bulk density values from the calibrated acoustic televiewer datasets are then used to improve porewater concentration estimates from rock concentrations.

Acoustic televiewer amplitude data for porosity estimation with application to porewater conversion

Abstract The ability to detect and quantify rock matrix porosity is explored using acoustic televiewer images and dolostone rock core samples. In particular, an empirical relationship between the acoustic televiewer amplitude response and core-derived total porosity is developed. The performance of the acoustic televiewer amplitude relationship was compared to neutron and gamma-gamma logs, which are traditional nuclear wireline tools commonly used for estimating porosity in dolostones. Acoustic televiewer amplitude results are consistent with the gamma-gamma (density) and neutron (porosity) logs, provide increased vertical resolution, do not require downhole radioactive sources and also provide oriented fracture information. Sample specific porosity and bulk density values from the calibrated acoustic televiewer datasets are then used to improve porewater concentration estimates from rock concentrations.

The integration of elastic wave properties and machine learning for the distribution of petrophysical properties in reservoir modeling

Conventional reservoir modeling employs variograms to predict the spatial distribution of petrophysical properties. This study aims to improve property distribution by incorporating elastic wave properties. In this study, elastic wave properties obtained from seismic inversion are used as input for an artificial neural network to predict neutron porosity in between well locations. The method employed in this study is supervised learning based on available well logs. This method converts every seismic trace into a pseudo-well log, hence reducing the uncertainty between well locations. By incorporating the seismic response, the reliance on geostatistical methods such as variograms for the distribution of petrophysical properties is reduced drastically. The results of the artificial neural network show good correlation with the neutron porosity log which gives confidence for spatial prediction in areas where well logs are not available.

Hydrocarbon type detection using the synthetic logs: A case study, Baba member, Gulf of Suez, Egypt

Lack of pressure data and/or one curve of the essential well logs such as density, neutron, sonic, resistivity and gamma ray curves in some reservoirs, is not preferred situation for the petrophysict, which will affect the petrophysical evaluation process and the reservoir characterization study.Four wells (A, B, C and D), in the area of interest, penetrated Baba sandstone zone in Badri field. The main challenge in the studied zone is the absence of pressure gradients and the density-neutron separation is not clear enough to judge the hydrocarbon type in all wells at the level of baba sandstone reservoir, where they are needed for the estimation of the hydrocarbon type (oil or gas). So it is preferred to use the three porosity logs to investigate the hydrocarbon type. The second challenge is constructing synthetic logs in the absence of the three main porosity logs.In this paper, all available data were used to estimate the hyocarbon type in two studied wells at the level of Baba sandstone zone, firstly by using well logging relationships between density and sonic in the wells which have sonic logs at the zone of interest, secondly by applying special petrophysical studies to help for estimation of hydrocarbon type.Crossplot technique was used to estimate the hydrocarbon type in the zone of interest using the synthetic sonic curve, original density curve, original neutron curve plus their derived porosity and variables values in the used well logging crossplots including some petrophysical crossplots, which can help in this situation.The results of this study revealed that, 1) the synthetic logs play a great role in solving of petrophysical and geological problems, 2) there is a good relationship between the density and sonic logs in the analogues (A and D) wells, and 3) the obtained synthetic sonic log makes a sense with lithology and other logs (neutron and gamma). Therefore, the synthetic sonic log, in corporation with the density and the neutron logs as well as the crossplots can detect the fluid type in the studied wells. The final result that the most likely hydrocarbon type in well (B) is oil while it has gas signature in well (C).

New insights into porosity determination using artificial intelligence techniques for carbonate reservoirs

Abstract The porosity of the petroleum reservoirs is considered one of the most important parameters in reserve estimation because it determines the effective volume of the hydrocarbon that is stored in the reservoir. Based on the reserve estimation, the development plan can be set and managed. Porosity can be determined in the laboratory which is the most expensive methods. Porosity also can be determined from the logs such as density, neutron, sonic, and NMR logs. There are a lot of uncertainties in the porosity estimation from wireline logs because it depends on many statistical analysis and also is affected by the logging environment and logging tools. The prediction of the porosity from different porosity logs using artificial intelligence (AI) methods validated with the laboratory measured values is the best method to determine an accurate value of the rock porosity. The objective of this research is to evaluate AI tools such as artificial neural network (ANN), Support vector machine (SVM) and Adaptive neuro fuzzy inference system (ANFIS) to predict the reservoir porosity based on wireline log data. More than 1700 field measurements of porosity with logs data were used for training and testing the AI techniques. The results obtained showed that ANN and ANFIS can be used to estimate the reservoir porosity based on log data with a high correlation coefficient (R) and low average absolute percentage error (AAPE). The main inputs required for porosity estimation are bulk density, neutron porosity, and sonic compressional time. The developed mathematical equation based on the weights and bias of the ANN model can be used to predict the reservoir porosity based on log data with a correlation coefficient of 0.98 and an AAPE less than 8%. The advantage of this work is that we extracted the mathematical model from the ANN that can be used directly to determine the porosity without the need for training and testing the data. The porosity estimation from the neutron-density crossplots, which is the current technique used by the industry, yielded 14.7% error.

A Statistically Driven Spectral Method for Deriving Reservoir Properties Using 3D Seismic Data and Well Log Suites


 
 
 
 A statistically driven spectral method was carried out on 3D seismic data and well logs in ‘’VIC’’ Field within the Niger Delta with the aim of deriving reservoir properties and delineating stratigraphic features using edge detection attributes like coherence so as to have a better and clearer view of subsurface structure of a reservoir interval that possesses hydrocarbon using Spectral method.
 A suite of data consisting of seismic sections and composite logs comprising Gamma-ray, Resistivity, Spontaneous Potential, Sonic Time and Porosity logs (density and Neutron) were utilized to identify reservoir interval on log signature across wells 4 and 5 and the reservoir interval obtained was between 11,164 feet and 11,196 feet. Edge detection attribute like coherence was computed from the amplitude data in time domain and transformed to frequency domain using Fourier Transform tool in MATLAB. In order to display well log in time, well to seismic tie was carried out using check shot data which was used as time to depth relationship.
 The analysis of the spectral domain shows distinct bright spots that vary with measured depth due to variation in fluid and formation properties. The results led to an enhancement of seismic data interpretation in the field of study due to a spectral technique method that was applied to calculate the frequency slices. The results indicate that the spectral domain in coherence attributes revealed better geological features and the reservoir character such as faults and fractures.
 Frequency domain gives better geological maps as it is used to filter data, which means it is an enhancement of hidden features in time domain and gives a smoother variation of the features that has low frequency values. A reservoir with low frequency values is a sandy environment showing stratigraphy features. Hence, the reservoir is suspected to be a channel fill reservoir. This implies that Spectral domain (frequency) defines major geological areas of the ‘’VIC’’ field and gives much clearer image of the reservoir features within the field than in time domain.
 
 
 
 

Dielectric constant measurements of sweep frequency and its effect from 20 MHz to 1000 MHz

The array dielectric tool (ADT) can be used in situations where the formation water resistivity is variable or unknown and allows for obtaining Archie's cementation exponent (m) in combination with other porosity logging methods. However, few laboratory measurements of the dielectric dispersion properties at wide frequencies (20 MHz–1000 MHz) have been shown till now, which hinders our understanding of the petrophysical mechanisms. In this study, we introduce the dielectric constant measurements of a sweep frequency in the range of 20–1000 MHz based on the parallel-disk capacitor method. In order to easier control the variables, such as the m and the porosity, artificial cores with changeable m values were manufactured from glass beads. The relationships between the core dielectric constant and the frequency, saturation, salinity, and pore texture were investigated in the MHz to GHz range. It is shown that the frequency, water saturation, salinity, and pore texture affect the dielectric dispersion of cores. Additionally, a modified formula for predicting m with a high-frequency dielectric constant was developed based on core sample data. This study provides a better understanding of the dielectric dispersion mechanism and the modified formula for calculating m can improve the analysis of well log data.

Rock type based poroperm and continuous permeability predictions in a tight gas formation

Porosity from log response such as density provides a continuous representation of pore volume as function of depth in a well, which can be calibrated with core analysis data. Obtaining a continuous log of permeability is not as straightforward as there is yet no means of logging permeability. While DST-derived permeability values are useful in calibrating dynamic models, they only represent an average value over the radius investigated by the test and will not readily correlate to permeability values derived from core especially where there are lateral and vertical permeability variations within the reservoir. It is, however, possible to obtain a depth-continuous permeability estimate by deriving a free regression algorithm known as the poroperm transform function, which defines how the permeability varies as a function of porosity. Such correlations are typically derived empirically from overburden corrected core-derived porosity and permeability data. General porosity-permeability trends are far too scattered to be of use. However, far tighter porosity-permeability trends can be obtained by use of rock typing to identify suitable analogues. The following study highlights how microstructural rock-tying can be used to improve permeability prediction in a set of tight gas sandstone wells. Scanning electron microscopy (SEM) and quantitative X-ray diffraction (QXRD) data were obtained from >200 tight gas sandstone samples from the Southern North Sea for which porosity and permeability measurements had previously been conducted. The SEM and QXRD data were used to derive microstructural and mineralogical rock types respectively. Samples from each rock-types occupy different but overlapping positions on porosity-permeability cross plots. Exponential functions were fitted to porosity-permeability data for each rock type and then applied to the porosity values from wire-line log data to derive continuous permeability estimates. The log porosity curves-being the independent variables of the respective functions, were validated by core observation to avoid error propagation. Continuous permeability curves that honour mineralogical variation were obtained by the use of microstructural rock typing.Key words: Porosity, Permeability, Rock-typing, Poroperm, Microstructural, Mineralogical.

Integrated approach using core analysis and wireline measurement to estimate rock mechanical properties of the Zubair Reservoir, Southern Iraq

The Zubair Reservoir is a regionally extensive oil-producing sandstone sequence which is part of the petroleum system in Iraq, Kuwait, Syria, Iran, and Saudi Arabia. Rock mechanical properties are extremely useful in optimizing drilling, production, and reservoir compaction. Hence, an accurate technique for estimating rock mechanical properties may significantly improve the economic revenues derivable from the reservoir. However, information about rock mechanical properties along the depth is often discontinuous and limited to core samples taken within the reservoir interval. The present study aims to predict a continuous profile of rock mechanical properties throughout the depth of the Zubair Reservoir directly from wireline porosity logs. Special characteristics of Zubair sandstone were evaluated and described using thin-section photographs, X-ray diffraction analysis, and scanning electron microscope imaging to identify and characterize the mineralogy, texture, and structure of the sandstone unit. Thereafter, both triaxial and triaxial multistage testing techniques were carried out on 130 plug samples from the Zubair sandstone formation to measure the rock's mechanical properties. Valid local empirical correlations were established between the porosity and these mechanical parameters. The obtained results revealed that the rock mechanical properties are primarily functions of porosity and the direct linear expression is more reliable than power functions and exponential functions. The best relationship obtained between UCS, internal friction angle, Young's modulus, bulk modulus, shear modulus, and Poisson's ratio with porosity had a correlation coefficient (R2) of 0.84, 0.75, 0.80, 0.59, 0.83 and 0.70, respectively. The accuracy of the newly suggested empirical correlations was subjected to statistical analysis. Moreover, these relationships were compared with the existing commonly used correlations reported in the literature on real field data from Zubair sandstone. The results showed that the suggested empirical correlations have a high accuracy and reliability, while the previous correlations could not adequately cover the Zubair sandstone data. This emphasizes the importance of using local correlations to estimate rock mechanical properties. The predicted continuous rock mechanical profile provides a good indication of the strength and stability of the formation around the wellbore. Consequently, it can be used in solving wellbore instability problems; preventing sand production; optimizing drilling processes, which includes the selection of the bit type and drilling parameters; and fracturing operations across the Zubair sandstone reservoir.

Multivariate statistical applications for quantifying contribution of seismic data to production prediction in the Wolfcamp, Midland Basin

A workflow with the objective of quantifying the added value of seismic-based information to improved prediction of oil production in the Wolfcamp section of the Midland Basin is described. A variety of multivariate statistical algorithms are employed in the workflow: unsupervised and supervised classification, categorical-based nonlinear regression, and spatial regression for well log estimation. Porosity is estimated from elastic inversion volumes within the context of lithotype classification originating with well logs. Porosity feet becomes the primary driver in production prediction. Two parallel analyses are performed. In the first, only well logs are used to predict porosity along all vertical wellbores. As a proxy for the impact of well spacing, different minimum search distances are investigated for porosity estimation error and associated error in prediction of cumulative oil production at six months. The second analysis is similar except, along each wellbore, porosity is extracted from the seismic porosity volume and then used as soft data. These two analysis flows are then applied to horizontal wellbores, which have no original porosity logs, to compare how well each predicts known cumulative production. The incorporation of seismic-based information does show significant error reduction.

Porosity estimation in kerogen-bearing shale gas reservoirs

Porosity is a fundamental petrophysical parameter in shale gas reservoirs that governs the space for hydrocarbon storage, and directly determines the free gas amount and absorbed gas capability. Technically, in kerogen-bearing shales, well-log derived porosity may yield inaccurate results as the porosity tools in response to both the kerogen and the liquid-filled pore spaces that are often undifferentiated. In this paper, we propose a new method for porosity estimation in kerogen-bearing shales, where porosity is assumed to be composed of both matrix porosity and kerogen porosity. The kerogen responses of density, sonic and neutron logs are first calculated from experimental data to calibrate porosity logs with the kerogen volume. The matrix porosity is then computed using the improved Wyllie equation based on the calibrated logging data with the kerogen influence removed. The kerogen porosity is estimated by a mass-balance relation based on the original total organic carbon (TOCo) and kerogen maturity characterized by the percentage of convertible organic carbon (Cc) and the transformation ratio (TR). Application of the new method to a shale gas reservoir in the Ordos Basin, China shows that the estimated porosity matches the core derived porosity satisfactorily well. Furthermore, the results also indicate that the shale kerogen porosity is relatively higher than the shale matrix porosity when the kerogen amount and maturity are high. The results of the study lead to a better understanding of the shale formation and thus contribute towards the better evaluation of shale gas reservoirs.

CHARACTERIZATION OF THE HYDROCARBON POTENTIAL AND NON-POTENTIAL ZONES USING WAVELET-BASED FRACTAL ANALYSIS

The identification of prospective and dry zone is of major importance from well log data. Truthfulness in the identification of potential zone is a very crucial issue in hydrocarbon exploration. In this line, the problem has received considerable attention and many conventional techniques have been proposed. The purpose of this study is to recognize the hydrocarbon and non-hydrocarbon bearing portion within a reservoir by using the non-conventional technique. The wavelet based fractal analysis (WBFA) has been applied on the wire-line log data in order to obtain the pre-defined hydrocarbon (HC) and non-hydrocarbon (NHC) zones by their self-affine signal nature is demonstrated in this paper. The feasibility of the proposed technique is tested with the help of most commonly used logs, like self-potential, gamma ray, resistivity and porosity log responses. These logs are obtained from the industry to make out several HC and NHC zones of all wells in the study region belonging to the upper Assam basin. The results obtained in this study for a particular log response, where in the case of HC bearing zones, it is found that they are mainly situated in a variety of sandstones lithology which leads to the higher Hurst exponent. Further, the NHC zones found to be analogous to lithology with higher shale content having lower Hurst exponent. The above proposed technique can overcome the chance of miss interpretation in conventional reservoir characterization.

Rock typing in the Upper Devonian-Lower Mississippian Woodford Shale Formation, Oklahoma, USA

Most U.S. shale plays are spatially extensive with regions of different thermal maturity and varying production prospects. With increasing understanding of the heterogeneity, microstructure, and anisotropy of shales, efforts are now directed to identifying the sweet spots and optimum completion zones in any shale play. Rock typing is a step in this direction. We have developed an integrated workflow for rock typing using laboratory-petrophysical measurements on core samples and well logs. A total of seven wells with core data were considered for rock typing in the Woodford Shale. The integrated workflow has been applied in the Woodford Shale in a series of steps. In the first step, unsupervised clustering algorithms such as [Formula: see text]-means and self-organizing maps were used to define the rock types. Rock type 1 is generally characterized by high porosity and total organic carbon (TOC). Rock type 2 had intermediate values of porosity and TOC and thus, moderate source potential and storage. Rock type 3 had the highest carbonate content, poor storage, and source rock potential. In the next step, a classification algorithm, support vector machines (SVM), was used to extend the rock types from the cores to the logs. A logging suite with gamma ray, resistivity, neutron porosity, and density logs was used for extending the rock types. These logs were used because they are commonly available and adequate to differentiate different rock types. The rock types were populated in the uncored sections of the seven cored wells and additionally in 12 wells (taken from Drilling Info) using a trained SVM model. Additional wells were taken to have sufficient data for production correlation. In the final step, a rock-type ratio (RTR) was defined based on the fraction of rock type 1 over the gross thickness. RTR was found to positively correlate with normalized oil equivalent production.

Evaluation of bitumen intervals in the carbonate reservoir, case study in Y oilfield of Iran

ABSTRACTThe bitumen intervals, which were drilled at Y oilfield of Iran, often reduce drilling efficiency. The characteristics of high natural gamma radioactive(GR), low uranium-free gamma(KTH), and poor GR-KTH relationship have been found to detect the bitumen by logging. Conventional porosity logs, Poisson's ratio, the P & S-wave velocity ratio and the wave impedance, derived from logging, can be used to identify the heavy oil bitumen. The bitumen distribution can be predicted by the log interpretation. Implementation of log evaluation allows better solving the bitumen intervals identify problems, and also to helps in enhancing development efficiency in the similar petroleum reservoirs.

A Diffusion Filter Based Scheme to Denoise Seismic Attributes and Improve Predicted Porosity Volume

This paper proposes a diffusion filter based scheme to denoise seismic attributes and to improve the porosity volume, which is predicted from seismic attributes. We compare the performances of multiple diffusion [such as Perona–Malik diffusion filter, complex diffusion filter, improved complex adaptive diffusion filter (ICADF)] and nondiffusion (such as two-dimensional (2-D) median, 3-D median, smoothing, and bilateral filter) based filters in terms of four metrics such as root mean square error (RMSE), normalized RMSE, signal to noise ratio (SNR), and peak SNR (PSNR). In our earlier publication, we used an artificial neural network (ANN) to predict a lithological property (sand fraction) over a study area. We trained the ANN using an integrated dataset of low-resolution seismic attributes and a limited number of high-resolution well logs. In this paper, we generate the porosity volume from the seismic attributes using an ANN. The predicted porosity logs contain irregularities and artifacts due to the nonlinear mapping of the learning algorithm (e.g., ANN). We apply a set of filters to the output of the ANN to regularize the predicted porosity volume. The filtered porosity logs are compared with the generated log. The ICADF has been found to be most suitable for denoising the seismic data and the porosity volume. Generation of porosity maps from seismic inputs would be helpful to petroleum engineers for reservoir characterization.

Deterministic estimation of gas-hydrate resource volume in a small area of the Ulleung Basin, East Sea (Japan Sea) from rock physics modeling and pre-stack inversion

We made deterministic estimations of the gas-hydrate and in-place gas resource volumes in a small area in the northwestern Ulleung Basin, East Sea (Japan Sea) from 3-D pre-stack seismic data and well-log and core data from the UBGH2-6 well. We modeled the P-impedance (Ip) logs at the well for 0%–100% pore-space gas-hydrate saturation from the P-wave velocity (Vp) and density logs modeled by the simplified three-phase Biot-type equation (STPBE). Then, the Ip volume for the gas-hydrate-bearing zone (GHBZ) was constructed by pre-stack inversion and divided into 28 layers. The porosity and mineralogy along these layers were assumed to be uniform, respectively, to the porosity log upscaled to the layers and the sediment constituents at the well determined from the core samples. Next, the pore-space gas-hydrate saturation at every time sample of each layer was found by matching the Ip value of the time sample to the modeled Ip logs upscaled to the layers. The gas-hydrate saturation volume with a cell size of 25 m × 6.25 m × 1 ms was obtained from the product of the pore-space gas-hydrate saturation volume and the porosity volume. The gas-hydrate saturation volume was converted into the depth volume based on the Vp value at each cell found by matching the pore-space gas-hydrate saturation of the cell to the modeled Vp logs. The estimated total gas-hydrate and gas resource volumes are about 8.43 × 108 m3 and about 1.38 × 1011 m3, respectively.

Uncertainty assessment of porosity and permeability by clustering algorithm and fuzzy arithmetic

A hybrid clustering-fuzzy arithmetic algorithm is here proposed, which uses cluster analysis to quantify porosity uncertainty, then the uncertainty is projected to the irreducible water saturation and permeability by the means of fuzzy arithmetic. The proposed method is applied to five wells of the carbonate Sarvak Formation, in an Iranian onshore oil-field. First, cluster analysis is applied to the porosity logs including neutron porosity, bulk density and sonic transit time. The uncertainty range of porosity is defined by the range of neutron porosity in each cluster. In order to estimate the core porosity, neutron porosity is calibrated to the core porosity in each cluster. Due to the average of error, the calibrated clustering-based porosity is at least 33% more accurate than the conventional methods. Based on the generalization ability of porosity estimators, a homogeneous porosity zone is determined northward. Irreducible water saturation, analyzed by the proposed method, has less overestimation, compared to the conventional evaluation of irreducible water saturation. Permeability fuzzy number is compatible with core tests, except in well S1, which is drilled in a location, compressed by two stress regimes (N-S and NW-SE). Two criteria are defined for validating the fuzzy numbers by core data: (i) Crit1 finds an average α-cut of core values; (ii) Crit2 finds the best α-cut to optimize uncertainty interval of the fuzzy number. The α > 0.90 is the most appropriate for porosity and permeability studies.

Fractal analysis of logs to characterize the hydrocarbon and non-hydrocarbon zones of Bhogpara oil field, Northeast India

Identification with accuracy of prospective and dry zone from well log data is of prime importance in reservoir or hydrocarbon studies. This issue has greater stake, where in return many conventional methods have been established. The purpose of this study is to recognize the hydrocarbon and non-hydrocarbon bearing portion within a reservoir by using a non-conventional technique. Application of rescaled range (R/S) analysis and wavelet-based fractal analysis (WBFA) on the wire-line log data to obtain the pre-defined hydrocarbon (HC) and non-hydrocarbon (NHC) zones by their fractal nature is demonstrated in this paper. Among these two techniques, the WBFA tool has provided more prolific results. Applicability of the proposed approach is tested with the help of the most commonly used well log data like self-potential, gamma ray, and porosity log responses. These are used in the industry to distinguish several HC and NHC zones of all wells in the study region belonging to the Upper Assam Basin, India. The results are found to be of lower fractal dimension in this study for a particular log response having HC-bearing zones, which are mainly situated in a variety of sandstone lithology. On the other hand, NHC-bearing zones correspond to lithology with higher shale content categorized with higher fractal dimension. Hence, the WBFA technique can overcome the chance of misinterpretation, which is quite possible in the case of conventional reservoir characterization.

A new method for calculating gas saturation of low-resistivity shale gas reservoirs

The Jiaoshiba shale gas field is located in the Fuling area of the Sichuan Basin, with the Upper Ordovician Wufeng–Lower Silurian Longmaxi Fm as the pay zone. At the bottom of the pay zone, a high-quality shale gas reservoir about 20 m thick is generally developed with high organic contents and gas abundance, but its resistivity is relatively low. Accordingly, the gas saturation calculated by formulas (e.g. Archie) using electric logging data is often much lower than the experiment-derived value. In this paper, a new method was presented for calculating gas saturation more accurately based on non-electric logging data. Firstly, the causes for the low resistivity of shale gas reservoirs in this area were analyzed. Then, the limitation of traditional methods for calculating gas saturation based on electric logging data was diagnosed, and the feasibility of the neutron–density porosity overlay method was illustrated. According to the response characteristics of neutron, density and other porosity logging in shale gas reservoirs, a model for calculating gas saturation of shale gas was established by core experimental calibration based on the density logging value, the density porosity and the difference between density porosity and neutron porosity, by means of multiple methods (e.g. the dual-porosity overlay method by optimizing the best overlay coefficient). This new method avoids the effect of low resistivity, and thus can provide normal calculated gas saturation of high-quality shale gas reservoirs. It works well in practical application. This new method provides a technical support for the calculation of shale gas reserves in this area.

Porosity inversion by Caianiello neural networks with Levenberg-Marquardt optimization

Caianiello neural networks (CNNs) incorporated with the Robinson seismic convolutional model are modified by the Levenberg-Marquardt algorithm to improve convergence. CNNs are extended to the multiattribute domain for reservoir property inversion, with time-varying signal processing by a frequency-domain block implementation using fast Fourier transforms. Optimal inversion can be achieved by applying the Levenberg-Marquardt optimization to multiattribute domain CNNs for convergency improvement due to its ability to swing between the steepest-descent and Gauss-Newton algorithms. The methodology is applied to porosity estimation in an oilfield with six wells in the Bohai Basin of China. Cross-validation results indicate significant correlation between actual porosity logs and predicted porosity logs. Compared with a traditional method, our technique is robust.