No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Understanding reservoir heterogeneity through water-saturation prediction via neural network — A case study from offshore Nile Delta
Abstract
In complex geologic settings with a great degree of heterogeneity in reservoir properties, such as submarine channel complexes as in the Nile Delta province, we face the challenge of characterizing the reservoir based on availability of different seismic attributes. Amplitude variation with offset (AVO) analysis and prestack inversion techniques show impressive results in delineating the gas-bearing reservoirs, especially in clastic systems. However, a shortage of available wells and/or seismic data leads to a challenge in applying AVO and any prestack seismic inversion approaches. In addition, quantitative prediction of water saturation (Sw) from seismic is ambiguous because of its independent nonlinear relationship with conventional seismic attributes and inversion products. Water-saturation prediction away from the well is essential in order to characterize the reservoirs effectively. Therefore, probabilistic neural network (PNN) analysis has been implemented to predict Sw 3D volume using full-stack seismic data and Sw logs. In this case study, we applied the proposed neural network workflow over one of the late-Pliocene gas-sandstone reservoirs, Sequoia Field, in the West Delta Deep Marine (WDDM) concession, offshore Nile Delta, Egypt. The resulting volume then was tested using two blind wells that haven't been used in the analysis. The predicted Sw volume contains fine details that were used with variance and spectral-decomposed volumes to understand the reservoir's internal architecture in terms of sand body geometries and connectivity. The resulting volumes were used to better define the reservoir and optimize a new development well location.
... The Sequoia field, the case study, is one of the major gas fields in both WDDM and Rosetta concessions ( Fig. 1) Samuel et al. (2003). The field is located on the north-western margin of the outer slope of the Nile Delta, approximately 50 km from the nearest shoreline Mohamed et al. (2017). Different inversion methodologies have been proposed for reservoir properties characterization to predict rock/fluid properties from seismic amplitude data. ...
... The reservoir's base is defined by a major incision that represents the base of the canyon (Fig. 4). As presented by Mohamed et al. (2017), the canyon is filled by many smaller channels that are stacked together to form the final shape of the reservoir (Fig.3). The pay gas sand is approximately 77 m. ...
... Locations of the study wells and the blind well are shown. Modified fromMohamed et al. (2017) ...
The Prediction of the reservoir characteristics from seismic amplitude data is a main challenge. Especially in the Nile Delta Basin, where the subsurface geology is complex and the reservoirs are highly heterogeneous. Modern seismic reservoir characterization methodologies are spanning around attributes analysis, deterministic and stochastic inversion methods, Amplitude Variation with Offset (AVO) interpretations, and stack rotations. These methodologies proved good outcomes in detecting the gas sand reservoirs and quantifying the reservoir properties. However, when the pre-stack seismic data is not available, most of the AVO-related inversion methods cannot be implemented. Moreover, there is no direct link between the seismic amplitude data and most of the reservoir properties, such as hydrocarbon saturation, many assumptions are imbedded and the results are questionable. Application of Artificial Neural Network (ANN) algorithms to predict the reservoir characteristics is a new emerging trend. The main advantage of the ANN algorithm over the other seismic reservoir characterization methodologies is the ability to build nonlinear relationships between the petrophysical logs and seismic data. Hence, it can be used to predict various reservoir properties in a 3D space with a reasonable amount of accuracy. We implemented the ANN method on the Sequoia gas field, Offshore Nile Delta, to predict the reservoir petrophysical properties from the seismic amplitude data. The chosen algorithm was the Probabilistic Neural Network (PNN). One well was kept apart from the analysis and used later as blind quality control to test the results.
... In a case study by Mohamed et al. (2017) a PNN analysis was used to predict a water saturation volume using full-stack seismic data and Sw logs for the late-Pliocene gas-sandstone reservoirs of the Sequoia field, also in the WDDM concession. The resulting Sw volume showed fine channel details in good agreement with the wells. ...
... Pre-stack inversion techniques are known to give good results in delineating gas reservoirs, especially in clastic systems (e.g., Mohamed et al., 2017). ...
... However, especially in the case of gas, because of its nonlinear relationship with elastic properties the quantitative prediction of water saturation (Sw) from seismic is unreliable (e.g., Mohamed et al., 2017). The estimation of Sw away from the wells is valuable for an improved reservoir characterization and helps to calculate gas reserves and to optimize new well locations. ...
... Those channels are different in shape and behavior and include laterally amalgamated channels, sinuous channels, channels with frontal splays, and leveed channels. Following the work of Cross et al. (2009) and the application of seismic variance and spectral decomposition on 3D seismic data presented by Mohamed et al. (2017), we are able to map and understand the geometry of the incised channels. The Sequoia story started with the initial erosion of the channel (stage I), followed by the deposition of several laterally amalgamated channels (stage II), then another deposition of sinuous channels with aggradational stacking patterns (stage III). ...
... The seismic-derived attributes are grouped into instantaneous, windowed frequency, filter slice, derivative, integrated, and time attributes. Moreover, all of the attributes are increased by applying nonlinear transformsnatural log, exponential, square, inverse, and square root (Mohamed et al., 2017). We try all of these attributes and then determine statistically the best order and number of attributes that give the lowest prediction error and save them for the subsequent network training. ...
... We have applied the same PNN procedure that was proposed by Mohamed et al. (2017), using Hampson-Russell software, to predict S w and ϕ in two different separate runs using different sets of attributes. First, we filtered, resampled, and smoothed the logs to match the seismic data vertical scale (4 ms). ...
In highly heterogeneous basins with complex subsurface geology, such as the Nile Delta Basin, accurate prediction of reservoir modeling has been a challenge. Reservoir modeling is a continuous process that begins with field discovery and ends with the last phases of production and abandonment. Currently, the stochastic reservoir modeling method is widely used instead of the traditional deterministic modeling method to consider spatial statistics and uncertainties. However, the modeling workflow is demanding and slow, typically requiring months from the initial model concept to flow simulation. In addition, errors from early model stages become cumulative and are difficult to change retroactively. To overcome these limitations, a new workflow is proposed that implements probabilistic neural network inversion to predict reservoir properties. First, well-log data were conditioned properly to match the seismic data scale. Then, the networks were trained and validated, using the conditioned well-log data and seismic internal/external attributes, to predict water saturation and effective porosity 3D volumes. The resulting volumes were sampled in simulation 3D grids and tested using a blind well test. Subsequently, the permeability was calculated from a porosity-permeability relationship inside the reservoir. Finally, a dynamic simulation project of the field was performed in which the historical field production and pressures were compared to the predicted values. One of the Pliocene deepwater turbidite reservoirs in the offshore Nile Delta was used to demonstrate the proposed approach. The results proved the accuracy of the model in predicting the reservoir properties and honoring the heterogeneity of the reservoir. The new approach represents a shortcut for the seismic-to-simulation process, providing a reliable and fast way of constructing a reservoir model and making the seismic-to-simulation process easier.
... AVO and inversion attributes identify DHI features (Roden et al. 2014) and demonstrate remarkable results in defining gas-bearing reservoirs. However, a lack of usable wells and seismic data complicates the implementation of AVO and other pre-stack seismic inversion procedures (Mohamed et al. 2017). Attributes such as energy, similarity, signal-to-noise ratio, dip-variance, and average frequency can be used to decode gas chimney paths accurately. ...
The prediction of highly heterogeneous reservoir parameters from seismic amplitude data is a major challenge. Seismic attribute analysis can enhance the tracking of subtle stratigraphic features. It is challenging to investigate these subtle features, including channel systems, with conventional-amplitude seismic data. Over the past few years, the use of machine learning (ML) to analyze multiple seismic attributes has enhanced the facies analysis by mapping patterns in seismic data. The purpose of this research was to assess the efficiency of an unsupervised self-organizing map (SOM) approach supported by multi-attribute analysis that could improve gas channel detection and seismic facies classification in Serpent Field, offshore Nile Delta, Egypt. As well as evaluates the importance of several available seismic attributes in reservoir characterization rather than analyzing individual attribute volumes. In this study, the single attribute (spectral decomposition attribute) highlighted the gas channel spatial distribution using three distinct frequency magnitude values. Subsequently, we employ principal component analysis (PCA) as an attribute selection method, discovering that combining seismic attributes such as sweetness, envelope, spectral magnitude, and spectral voice as input for SOM reflects an effective method to determine facies. The clustering results distinguish between shale, shaly sand, wet sand, and gas-saturated sand and identify gas–water contact on a 2D topological map (SOM), where each pattern indicates certain facies. This is achieved by associating the SOM outputs with lithofacies determined from petrophysical logs. Reducing exploration and development risk and empowering the geoscientist to generate a more precise interpretation are the ultimate objectives of this multi-attribute analysis.
... These properties include porosity, volume of shale, net pay, and water saturation. Understanding these properties is essential for optimizing oil production and comprehending reservoir heterogeneity (Mohamed et al., 2017;Ren and Duncan, 2021;Paramo et al., 2023). ...
Accurate reservoir characterization is essential for successful hydrocarbon extraction, especially in complex fields such as the Garraf Oil Field. This study aims to enhance reservoir characterization by integrating different petrophysical assessments and rock typing methodologies. Density, neutron, and sonic porosity evaluations were used to assess porosity, while gamma-ray logs and resistivity measurements were used to determine shale volume. The Archie equation was employed to estimate water saturation and sensitivity analysis was used to determine the cutoff values. The study also utilized rock typing techniques, including hydraulic flow unit assessment and Rock fabric number cross-plots, to categorize reservoir rocks into flow units and identify unique rock types. The combination of these approaches led to the precise identification of reservoir heterogeneities and optimal oil production zones. The results showed that the Gamma-ray log is the best method for determining shale volume, and the closest method for porosity determination is the density log. The water resistivity value was estimated at 0.016, while the Archie parameters (a,m,n) were 1.1, 2.1, and 3.7, respectively, with cutoff values of 0.22 for shale volume, 0.11 for porosity, and 0.56 for water saturation. The study identified five rock types ranging from packstone, pack to wackstone, wackstone, wack to mudstone, and mudstone. Overall, the integration of petrophysical evaluations and rock typing techniques facilitates the accurate delineation of oil-rich zones with enhanced reservoir connectivity.
... The area of interest includes Saffron field, which lies in the West Delta Deep Marine (WDDM) concession, 60-120 km offshore in the deepwater of the present-day Nile Delta [1] (Figure 1). Saffron field is a Pliocene submarine delta slope canyon system, with complex turbiditic channel-levee reservoirs [2]. ...
Extended Elastic Impedance (EEI) is a very useful seismic reconnaissance attribute. EEI logs can directly correspond to the petrophysical properties of the reservoir and the seismic. EEI reflectivity volumes can be obtained directly from the pre-stack seismic data. Better discrimination between the seismic anomaly caused by either lithology or fluid content can be utilized by applying this approach. The concept of extended elastic impedance is used to derive the petrophysical properties and distribute the reservoir facies. The study area was a Pliocene gas field, that lies in the deep marine, Offshore Nile Delta, Egypt. The workflow is simple, efficient, and uses very few inputs. We started with the fluid/ lithology logs and investigated the optimum projection in the intercept/gradient domain. Then, we used the conditioned angle stacks, to calculate the intercept/ gradient volumes, using Shuey’s two-term Approximation. The intercept and gradient volumes are converted directly to the fluid and lithology 3D volumes, without any of the pre-stack inversion constraints. The outputs were tested using a blind well and the correlation exceeds 80%. The results show that the EEI is a worthy effort to highlight the difference between the reservoir and nonreservoir sections, to identify the hydrocarbon area.
... X channel system shows evidence for faulting, including widening of the channel and small-scale channel diversions and intra-slope ponding of flows. Moreover, Mohamed et al. (2017) discussed that the X field is a submarine delta slope canyon system with complex turbiditic channel-levee reservoirs. Othman et al. (2021) studied the application of artificial neural network in seismic reservoir characterization with a case study from Offshore Nile Delta. ...
Putting a new concerning to each of the Eastern Mediterranean, offshore plays, gas detection using the most advanced tools, and the zones between gas zones which may not be easily recognized. This was done by adding the total combinable magnetic resonance (TCMR), to density porosity (PHIT-D), and density magnetic resonance porosity (DMRP). X-gas reservoir in the offshore Eastern Mediterranean used as case study. PHIT-D highest result values exist against gas bearing zones with about 29% in X-D5 and lowers at zones not bearing gas with about 6% in X-D3. TCMR highest result values exist against zones not bearing gas with about 49% in X-D5 and lower value zones bearing gas with about 32% in X-1. DMRP in the gas bearing zones does not increase like PHIT-D or decrease like TCMR so it is considered the best gas detector against gas bearing reservoirs and that is very useful. By comparing PHIT-D with DMRP, the gas zones are best detected, but in between zones, some gas zones not detected. By adding TCMR to the comparison, some new gas zones are detected which are lying in-between normal detected gas zones; so, this is a new solution. Interpretation is that PHIT-D is a tool, but DMRP is not a tool and is just integration between two tools: density and NMR. This makes a blind zone between density and DMRP which could be cleared by adding TCMR tools at the same time. The new method will change the producibility from gas reservoirs for the future operations in Eastern Mediterranean plus the previous reservoir production. Also, still applicable to any other gas reservoir.
... Channels are not the same as canyons. The latter are filled with a complex sequence of turbidites, including transgressive sandstones, slumps, crevasse splays, and numerous overbank deposits (Mohamed et al., 2017). The key elements of Saffron canyon fill are shown in Figure 3. ...
... Scarab Field, with Pliocene slope-channels, is located within West Delta Deep Marine (WDDM) concession ( Fig. 1) in the Nile Delta Basin (Abdel Aal et al., 2000;Samuel et al., 2003 (Mokhtar et al., 2016), covers an area of 6150 km 2 (Samuel et al., 2003). It approximately lies fifty to a hundred kilometers, offshore, northwest the Nile cone margin (Samuel et al., 2003;Mohamed et al., 2017). Pliocene sediment supply from the River Nile built the deepwater slope-channel system, found in big canyon fill pattern (Mokhtar et al., 2016), within the West Delta Deep Marine concession (Garziglia et al., 2008;Cross et al., 2009). ...
Static reservoir modeling adequately and precisely defines the reservoir framework (geometry) and architecture (property). So as to substantially reduce the expected risk while defining the hydrocarbon-bearing reservoir boundaries, static geocellular model was constructed, in this study, to define lithology and fluid stochastic distribution, and volume of hydrocarbons as well, within the Late Pliocene hydrocarbon-bearing El Wastani Formation, (sandstone channel-1 and channel-2), in Scarab Field (offshore Nile Delta, Egypt). Twenty 2D seismic lines and well log data, of four wells, were used in this study. Stratigraphic, structural, facies, and petrophysical models were applied not only to estimate the volume of hydrocarbons within the reservoir but also to identify the best location for proposing new development wells that are of crucial value for oil industry development. In accordance with the constructed reservoir geocellular model, we found that the northwestern part of Scarab field is the best location for running further oilfield development plans.
... The training process is carried out until at least one of the following conditions is met: (i) a minimization of a mean square error (MSE) goal is achieved; (ii) occurrence of three consecutive nonimprovements in the MSE for the validation subset; or (iii) a maximum number of iterations is completed. The test subset is used only to estimate the prediction power of the PNN by performing a blind test and it is not used for building the NN model (Leiphart and Hart, 2001;Mohamed et al., 2017). ...
The acoustic impedance (AI) inversion aims to obtain a high-resolution impedance volume
by integrating well-log and band-limited seismic data. Two AI inversion schemes the colored
inversion (CI) and the model-based inversion (MBI) are utilized to characterize possible sand
channel from the post-stack seismic section and log data from 13 wells from the Blackfoot
region, Alberta, Canada. The results from analyses indicate that both the model-based and
colored inversion methods provide mutually consistent impedance volumes with an average
correlation coefficient of 0.986 and 0.886 for MBI and CI, respectively. Both inversions show
low-impedances ranging from 6750-7350m/s*g/cc between 1060ms and 1065ms time interval
which is interpreted as a sand channel. The slice of the acoustic impedance variation along all
cross line and inline validates the presence of low impedances along the interpreted sand
channel. Thereafter, the multivariate regression and the Probabilistic Neural Network (PNN) are
employed to predict porosity volumes using CI and MBI inverted impedance as external
attributes. The cross-plots between predicted porosities and actual porosities using multivariate
regression and PNN algorithms indicate that PNN produces better statistical estimates of porosity
distribution compared to those predicted from the multivariate regression. Both methods show
high porosity values along the sand channel. The maximum porosity in the sand channel is 18%
when MBI derived impedance is used as an external attribute while it is 16% in the case of CI.
The results suggest that given seismic and well log data for a region, a combination of modelbased
inversion and PNN can produce a more reliable estimate of the petrophysical properties of
the subsurface.
Seismic inversion is used routinely in hydrocarbon exploration and development as an effective geophysical tool. Inversion methodologies vary from relative to absolute poststack, prestack, and stochastic seismic inversion. Most of the inversion algorithms suffer from the “nonuniqueness” problem, which means there is more than one possible geologic model that fits the seismic data. Because of this, other information must be provided in the form of a low-frequency initial model. Construction of an accurate initial model requires honoring the subsurface geology of the field, adequate well control, and accurately picked horizons. The lack of all or some of these items makes it hard to build a reliable initial model, and the subsequent inversion will be subject to cumulative errors and accordingly questionable results. To mitigate these limitations, we implement the multiple linear regression algorithm to blend selected seismic attributes with seismic velocities to build a robust 3D initial model. First, we analyze the elastic properties of a given well against the seismic internal attributes and the available processing velocities. Next, we build a set of multilinear equations to convert the chosen attributes into 3D elastic properties. Finally, we smooth the results and produce the elastic initial models. No horizons are needed in this process. To prove its validity, the proposed approach was applied in the offshore Nile Delta, Egypt. The new initial models for the case studies show better consistency with the geology of the area and contain very fine details compared with the conventional initial models. Hence, the proposed approach suits exploration of deep targets as it provides a reliable link between seismic internal attributes, velocities, and elastic logs for the construction of an accurate 3D initial model even though the data may be insufficient.
Predicting petrophysical parameters, particularly saturation, is a common challenge due to the lack of direct relationships with seismic elastic attributes. Therefore, in this paper, we used one of the Artificial Intelligence (AI) algorithms for seismic reservoir characterization to overcome this challenge. A new trend is the use of Artificial Neural Network (ANN) algorithms to predict reservoir characteristics. The capacity to construct nonlinear relationships between petrophysical logs and seismic data is the fundamental advantage of the ANN algorithm. A probabilistic neural network (PNN) was the selected algorithm. The input data includes seismic full-stack and log data for four wells. Multi-attribute linear regression and PNN were applied to the Nader field north western desert to predict water saturation (Sw) and effective porosity (PHIE) for the Bahariya reservoir. Stepwise regression, which derives subsets of attributes from internal and external attributes, is used in a multi-attribute analysis. External attributes are represented in acoustic impedance generated from seismic inversion, whereas internal attributes are represented in well logs and seismic data. The seismic inversion technique has several types, but model-based inversion is the focus of this study. The training of a neural network (NN) using three attributes in Sw prediction and using six attributes in PHIE prediction showed a substantial number of correlations. The actual and predicted water saturations and porosities of the PNN have correlation coefficients of 97.5% and 91%, respectively. Then, Sw and PHIE parameters were extended over the Bahariya reservoir to describe the distribution of these parameters in Nader field. The results prove the validity of the workflow to accurately predict Sw and PHIE with a higher accuracy than ever before in the north western desert. Hence, this workflow can be implemented in the analogue basins in the future.
In a complex reservoir with a significant degree of heterogeneity, it is a challenge to characterize the reservoir using different seismic attributes based on available data within certain time constraints. Prestack seismic inversion and amplitude variation with offset are among the techniques that give excellent results, particularly for gas-bearing clastic reservoir delineation because of the remarkable contrast between the latter and the surrounding rocks. Challenges arise when a shortage of seismic or well data presents an obstacle in applying these techniques. A further challenge arises if it is necessary to predict water saturation (Sw) using the seismic data because of the independent nonlinear relationship between Sw and seismic attributes and inversion products. Prediction of Sw is necessary not only for characterizing pay from nonpay reservoirs but also for economic reasons. Therefore, extended elastic impedance has been performed to produce a 3D volume of Sw over the reservoir interval. Then, a 3D sweetness volume and spectral decomposition volumes were used to grasp the geometry of the sand bodies that have been charged with gas in addition to their connectivity. This could help illustrate the different stages in the evolution of the Saffron channel system and the sand bodies distribution, both vertically and spatially, and consequently increase production and decrease development risk.
This work deals with a comparative study between density porosity and density magnetic resonance porosity in late Pliocene El Wastani gas reservoir, Sequoia field, West Delta Deep Marine (WDDM). In this study, the available well logging data by collecting, gathering, uploading, analyzing, and interpreting are used. Porosity determination, from the petrophysical parameters, routinely considered the most important process. The determined porosity by the two techniques is compared. Density resulting in density porosity (PHIT-D) showed results more than 23% in gas-bearing reservoir zones and less than 22% in non-gas reservoir zones. The porosity determined from integrating nuclear magnetic resonance (NMR) with conventional density porosity resulting in density magnetic resonance porosity (DMRP) showed results less than 33% in gas-bearing reservoir zones and more than 37% non-bearing gas zones. Comparison between the results of the two techniques in gas-bearing zones, PHIT-D is increasing and DMRP is not affected. DMRP considered the best and most true porosity against gas reservoir. This comparison is valid in any gas-bearing formations by using the proposed technique.
One of the main challenges that we face is the accurate prediction of pore-fluid properties with the highest possible resolution. The seismic resolution is the most limiting factor, especially in our case, in which the main reservoirs are deepwater turbidite channels and their thin beds typically fall below the seismic tuning thickness. Therefore, we designed a new workflow that combines the geostatistical inversion and the neural network analysis with the aim of predicting a 3D high-resolution water saturation (sampled every 1 ms), overcoming the limitation of seismic detectability and providing better reservoir characterization. The power of the geostatistical inversion is that it provides multiple model realizations, and each realization honors the well data (statistical information and logs) and the seismic data. These realizations are more reliable and high-resolution versions of the elastic parameters. On the other hand, the main advantage of the neural network is that it establishes a stable nonlinear link between the input seismic and inversion results and the target water saturation. The available data set for this study includes three angle stacks and seven wells from Scarab field, offshore Nile Delta. The resulted high-resolution saturation volume was tested using blind-well analysis and revisit post the drilling of a new well later on. It gave spectacular results in both cases. The normalized correlations between the predicted saturation volume and the real saturation logs are 0.87 and 0.89, respectively. The results prove the validity of the workflow to accurately predict water saturation with a higher resolution than ever before.
Building an accurate static model is critical to understand the reservoir heterogeneity, maintain the production, and optimize new wells locations. The stochastic modeling approach proved significant efficiency as a new and accurate modeling tool. In this case study, we applied the stochastic modeling approach to Sequoia Field. The Sequoia Field is a Pliocene gas field, offshore Nile Delta, Egypt. The field is a delta slope, multi-stacked canyon systems with complex turbidity channel-levee deposits. The canyon-fill consists of sandy channels, levees, crevasse splays, overbank deposits and slumps with multiple fills and incision episodes. The reservoir architecture commonly is the first priority in a stochastic reservoir model and is usually described in terms of different facies to rock types. The Geomodel grid layout was made considering the geological realism. It was constructed from the beginning not to be upscale at the end, on a scale grid design of increment 200x200x2 m mean, with around 4 hundred thousand cells. This increment was selected in such a big field to preserve the heterogeneity of the field with respecting to run time and the maximum number of the cell could be run in a dynamic model. The new model built has been used to calculate in-place volumes for Sequoia Field. The integrated structural framework of the model was made using the time and depth converted seismic horizons which used to create horizon model in time and depth domain. And the fault sticks were used to create the fault model and eventually the fault surfaces within the reservoir. The facies volume fractions were calculated from wells and considering the gross rock volumes from magnitude maps and inversion volumes. The channel trends were generated out of the voxels created from the inversion products. The reservoir properties like porosity water saturation volumes were modeled stochastically and co-simulated using correlation coefficients biasing to the facies property. Thin bed corrections were made. Hence the in-place volumes were calculated. The Stochastic geomodel optimizing on the grid resolution, incorporating interpretations from a new interpretation of seismic and inversion data and all well log analysis to match history, provide better water predictions and planning of additional wells if needed. This model will be the basis for dynamic modeling and will help in any further field development planning.
In highly heterogeneous basins with complex subsurface geology, as the Nile Delta basin, the accurate prediction of reservoir characteristics is a must. The reservoir characterization is a continuous process that begins with the field discovery and ends with the last phases of production and abandonment. Reservoir static modeling is the final step in the reservoir characterization process and consists of building an upscaled geologic model to be an input to the fluid simulations. The geostatistical reservoir modeling (stochastic modeling) methods are widely used instead of the traditional deterministic modeling methods to consider the spatial statistics and uncertainties. However, the modeling workflows are slow, requiring months from initial model concept to flow simulation or other outputs; Moreover, the early stages errors become cumulative and are difficult to retrospectively change. The neural network inversion gained popularity over the last decades for its ability to establish nonlinear relationships between the petrophysical logs and seismic data. It has been used to predict various reservoir properties with a reasonable amount of accuracy. Its main limitation resides at seismic resolution, and to overcome this problem a resolution-enhancing workflow has been adopted. This case study is from a Pliocene turbidite field in the offshore Nile Delta to illustrate the proposed modeling workflow. As a beginning, the resolution enhancement of seismic data is accomplished using derivative attributes and structural smoothing. Then, after proper well-log data conditioning, the training and cross-validation of Probabilistic Neural Network (PNN) are performed to produce shale volume (Vsh), porosity (φ), and water saturation (Sw) 3D volumes. The permeability (k) is calculated from poro-perm relationship inside the reservoir. The results are then sampled in 3D grids and tested using dynamic simulation method to assimilate production history. After the initial history match process, PNN parameters are adjusted to improve the match. The final model represents the best match to original field measurements and production data, which is then used in drilling decisions and production planning. The proposed neural network workflow reduces the reservoir modeling construction time by 80-90%, mitigates the cumulative error problems, and decreases the statistical uncertainty as it depends purely on seismic data to distribute the reservoir properties.
Coal seam gas (CSG) content is an important factor affecting the safety related to gas explosions, and it is also an indicator of natural gas resource assessment. The conventional method of establishing a network model for predicting gas content is mainly to predict the gas content of an individual location point, and achieve gas content prediction in the target area through an interpolation algorithm, which cannot accurately control the interpolation beyond the prediction point. Therefore, we selected a certain study area as an example, conducted a comprehensive analysis of CSG enrichment logging and seismic response characteristics, to determine the relative sensitive parameters, and used a support vector machine (SVM) network for sensitive parameters training based on genetic constraints. The algorithm (GA) optimizes the penalty parameters and kernel functions of the prediction model, completes the prediction of the gas content at the drill site, and uses neural network inversion to train and predict the target gas curve. The distribution of the gas content in the three-dimensional volume of the coal seam was obtained. A variety of gas content influencing factors and sensitive parameters of gas enrichment were integrated to establish a set of prediction methods for the volume gas content, which provided a theoretical basis for accurately predicting coal seam gas content. Comparison with subsequent measured data verified that this prediction method has a higher accuracy in this study area.
In deepwater depositional systems such as the slope canyon-turbidite channel system encountered on the continental slope of the offshore West Nile Delta Basin, deterministic prestack seismic inversion followed by facies classification using formation microimager facies logs delineated thick-bedded and thin-bedded gas-sand reservoirs encased in bypassed stratigraphic traps. Prestack seismic inversion was applied over the Scarab Field in the West Delta Deep Marine concession to evaluate the hydrocarbon potential of newly identified stratigraphic traps. Three angle stacks were inverted using a simultaneous inversion approach to estimate the elastic properties (P-impedance and V P /V S ratio). Using the elastic volumes produced from the inversion, Bayesian facies classification was applied to separate thin- and thick-bedded gas-sand facies from shale. Facies classification was focused on two prospective bypassed stratigraphic traps: the Upper Scarab Channel remnant levees (remnant middle levee [RML] and southern remnant levee) and the Lower Scarab Channel lateral accretion packages (LAPs). A development well location is proposed to validate the interpreted gas-sand reservoirs in the shallower RML and the deeper LAPs stratigraphic traps. The results suggest that application of the prestack seismic inversion and facies classification led to reliable inferences likely to have a positive impact on field development, potential reserves growth, and consequently boosting gas production.
Drilling wells in the oil and gas industry is a substantial process, whether they are appraisal wells drilled for reservoir-characteristic assessments at the exploration stage or production wells drilled following prior assessments. The challenge has always been to reduce drilling-related expenses and natural/environmental hazards by reducing the number of wells drilled, and to evaluate reservoir characteristics with as few calibration wells as possible. Physical and mathematical modeling of seismic data can help us understand the geologic and structural formations with minimal wells, and interpolate reservoir characteristics across large areas between a few drilled wells. In a new comparative approach, simultaneous prestack inversion and artificial neural network (ANN) methods are used to create 3D Poisson's ratio (PR) models built upon low-frequency initial models (IMs). Training the ANN on IMs similar to those used in the inversion has improved its performance while creating a valid base of comparison between the two methods. The inversion method was able to model the PR around four wells that had been used in creating the IMs. The generalized regression neural network that was trained on a PR IM, along with other seismic attributes, gave results that were consistent with the existing wells. The results of both methods confirm the existence of a strong relationship between PR and known hydrocarbon presence in these wells. However, examining the results with a blind well showed that the ANN was notably more successful than inversion in extrapolating the results beyond the logged sections in the wells and away from control wells. While this particular conclusion cannot be generalized, and the results obtained from the same methodology may vary from one reservoir to another, such results suggest that this procedure can become a robust part of a predrilling reservoir-evaluation phase in developing hydrocarbon fields.
We describe a new method for predicting well-log properties from seismic data. The analysis data consist of a series of target logs from wells which tie a 3-D seismic volume. The target logs theoretically may be of any type; however, the greatest success to date has been in predicting porosity logs. From the 3-D seismic volume a series of sample-based attributes is calculated. The objective is to derive a multiattribute transform, which is a linear or nonlinear transform between a subset of the attributes and the target log values. The selected subset is determined by a process of forward stepwise regression, which derives increasingly larger subsets of attributes. An extension of conventional crossplotting involves the use of a convolutional operator to resolve frequency differences between the target logs and the seismic data. In the linear mode, the transform consists of a series of weights derived by least-squares minimization. In the nonlinear mode, a neural network is trained, using the selected attributes as inputs. Two types of neural networks have been evaluated: the multilayer feedforward network (MLFN) and the probabilistic neural network (PNN). Because of its mathematical simplicity, the PNN appears to be the network of choice. To estimate the reliability of the derived multiattribute transform, crossvalidation is used. In this process, each well is systematically removed from the training set, and the transform is rederived from the remaining wells. The prediction error for the hidden well is then calculated. The validation error, which is the average error for all hidden wells, is used as a measure of the likely prediction error when the transform is applied to the seismic volume. The method is applied to two real data sets. In each power as we progress from single-attribute regression to linear multiattribute prediction to neural network prediction. This improvement is evident not only on the training data but, more importantly, on the validation data. In addition, the neural network shows a significant improvement in resolution over that from linear regression.
Within the Nile Delta gas province, reservoirs are dominated by Pliocene slope-channel systems, which are spectacularly imaged on high-quality three-dimensional seismic data. This article deals with the detailed seismic geomorphology of the Sequoia channel system, focusing on the geometry and distribution of its component sand bodies and the impact they have on reservoir heterogeneity. The Sequoia reservoir serves as a potential analog for similar but less well-imaged, deep-water slope systems. The reservoir consists of a succession of sandstones and mudstones organized into a composite upward-fining profile. Sand bodies include laterally amalgamated channels, sinuous channels, channels with frontal splays, and leveed channels and are interpreted to be the products of deep-water gravity-flow processes. Above a major basal incision surface, the reservoir is highly sand prone and made up of laterally amalgamated channels. The medial section of the reservoir is more aggradational and exhibits laterally isolated and sinuous channels. Within the upper part of the reservoir, channels are smaller, straighter, and built of individual channels with associated frontal splay elements and less common leveed channels. The main channel system is buried by a prograding slope succession that includes lobate sand-sheet elements. The stacking of facies within the Sequoia channel system implies a punctuated waning of sediment supply prior to eventual abandonment.
In clastic depositional systems such as those encountered in the Nile Delta Basin, simultaneous prestack seismic-amplitude inversion is an effective method for detecting and appraising gas-bearing sandstone reservoirs. However, the method has limitations concerning the requirement of a reliable set of wavelets, suitable wireline logs, and a sufficiently dense initial model. The neural-network analysis method is an alternative technique which sometimes can provide similar or better results and does not require significant volumes of data. Simultaneous prestack inversion was applied over the Scarab field, West Delta Deep Marine concession, offshore Egypt. The field comprises submarine channel-based gas reservoirs that extend laterally over 20 km2. Six wells were analyzed in a rock-physics study prior to performing inversion. Three angle gathers (near: 0-15°; mid: 15-30°; far: 30-45°) were inverted for P-wave impedance (ZP), S-wave impedance (ZS), P-wave velocity (VP), S-wave velocity (VS), VP/V S, and density (ρ) using the prestack inversion method. Neural-network analysis was performed using full-stack seismic data along with well logs in the training stage, followed by cross-validation of results and rendering of VP, VS, VP/VS, and density volumes. The VP/VS volumes produced from the two methods were used to infer water saturation (Sw). Direct comparisons were made between neural-network and prestack inversion results at a blind-well location to assess the relative quality of each method. Results suggest that application of the proposed neural-network method leads to reliable inferences. Hence, using the neural-network method alone or along with the prestack inversion method has a positive impact on reserves growth and increased production.
The Nile Delta offshore is rapidly emerging as a major gas province. High-quality three-dimensional (3-D) seismic data, coupled with data from 13 consecutive successful deep-water exploration and appraisal wells, have highlighted clear phases of erosion and deposition within the upper Pliocene deep-marine slope channels. The gross reservoir architecture is spectacularly imaged by 3-D seismic techniques, both in time sections and through a variety of amplitude extractions, while an extensive program of core and wire-line log acquisition and analysis has enabled high-resolution definition of the channel-fill sediments. The channels were initiated by the introduction of coarse sediments to the shelf edge possibly at times of relative sea level fall. Initially, there was significant erosion, especially in areas up depositional dip creating what we term "slope, valleys." Subsequent valley infill commonly commenced with debris flows, slumps, and slides, sometimes overlying basal, bypass-related sands, and progressed to amalgamated or stacked channels in packages of upward-decreasing net-to-gross sand ratios. This pattern was commonly repeated following reincision, which may have occurred several times. The different stages of channel development can be considered in terms of slope equilibrium with a reduction in slope gradient promoted by increases in flow size and density and decreases in grain size.