Barros, P. M. (Petrobras) | Fonseca, J. S. (Petrobras) | Silva, L. M. (Petrobras) | Maul, A. (Petrobras) | Yamamoto, T. (Petrobras) | Meneguim, T. (Petrobras) | Queiroz, L. E. (Petrobras) | Toribio, T. (Petrobras) | Martini, A. F. (Petrobras) | Gobatto, F. (Petrobras) | González, M. (Paradigm)
Geomechanical models need to incorporate a more reliable geological model concerning the overburden as well as the underburden sections. These aspects are more important when considering the flow simulation model and the injection and/or production of the field due to the surrounding rocks behavior and their responses against those ratios (injection/producing).
Therefore, a 3D model is needed and the oil industry faces a great challenge: how to build this reliable model without seismic information? Or, how to use seismic properties with all their ambiguity to derive this 3D model?
Special attention is necessary when observing the salt-section above the pre-salt reservoir in Santos Basin, Brazil, since the amplitude response is heavily influenced by the velocity model used for the migration process.
The majority of these velocity models are considered non-geological as the main reasoning for them is only to produce a good image. To build a good seismic image, reproducing the geology is not needed: the only condition imposed on the model is that the migration operators applied using that model will focus the image. In other words, we could see a completely alien model regarding geology that focuses the image, and that model would be considered correct.
In this paper, we will illustrate a way to build plausive velocity models fitting both the geology and the mathematics of the migration for a good seismic image. This allows for the use of the amplitude response for many purposes, including (but not limited to) deriving geomechanical properties, predicting the lithologies inside the evaporitic section and recursively building a new, more realistic velocity model.
A new technique is presented which detects electrical anisotropy from conventional resistivity logs in multiple wells, allowing the user to identify laminated pay intervals and create horizontal and vertical resistivity curves (Rh and Rv).
The benefits of tri-axial resistivity data in laminated formations are well known and lead to interpretation results with very low uncertainties. However, many laminated reservoirs were logged before these tools become available and interpretations using conventional measurements are generally pessimistic and prone to very large uncertainty.
The electrical anisotropy of a formation is determined from changes in responses of conventional resistivity measurements in multiple wells with different angles of relative dip. Each analysis is limited to formations which exhibit similar characteristics on other logs and which are above the transition zone. This often means dividing the field into smaller areas where the formation is seen to be consistent. Rv and Rh curves are computed from conventional resistivities and modelled anisotropy, while Rv_sh and Rh_sh values are determined from resistivities in thick shale sequences in multiple wells. These measurements are then used in a Laminated Shaly Sand Analysis.
Case study results are presented from a low-resistivity laminated shaly sand pay interval, along with a shaly sand interval which is not laminated. This illustrates how the technique can be used to differentiate laminated formations from those which may be bioturbated or disturbed in some way. Laminated formations show distinct anisotropic effects in high angle wells while bioturbated formations do not.
In practice, conventional interpretations are run over the entire formation with the analysis from laminated formations overriding the conventional interpretation over specific intervals. Uncertainty in the results is also considered, both in terms of the improvements over conventional interpretations and also improvements based on the number of wells analysed.
To reliably quantify changes in the in-situ stresses due to oilfield exploitation, it is necessary to start from a representative description of the subsurface and simulate both fluid movement and geomechanical effects. For this purpose, a new 3D "hybrid" grid is presented. This grid accurately honors geological features, with no unwarranted simplifications, while being ideally suited for geomechanical simulators and the new generation of flow simulators.
The new 3D Hybrid Grid is dominated by hexahedrons arranged in a structured manner except around faults, where it is made up of tetrahedrons. The grid is constructed from the underlying geological model and the sealed fault network. Both structured (hexahedrons) and unstructured (tetrahedrons) parts follow stratigraphic deposition information. For geomechanical simulations, each compartment is considered as an independent mesh part allowing sliding effects along the faults. At the interfaces between the hexahedrons and tetrahedrons in a fault block, shared nodes are used to ensure stress equilibrium and displacement continuity.
Uncertain states of stress and unforeseen changes in the integrity of the subsurface can have grave economic and environmental consequences. Understanding these helps mitigate development risks, and optimally develop the field. Today, geomechanical studies are not routinely performed and are often based on simplified descriptions of the subsurface. With this new Hybrid Grid, we propose to combine a reliable representation of subsurface with state-of-the art rock mechanics to assess how reservoirs respond to drilling, completion and production. Simplifications in either can lead to incorrect assessment of risks or production forecasts. The grid presented in this paper aims at honoring geology accurately while also being optimal for numerical computations. The zones of tetrahedrons enable the inclusions of even the most complex faulting systems, while the structured hexahedrons precisely follow the stratigraphy and are most efficient for geomechanical simulations.
The gridding technology presented enables a coherent representation of the subsurface for constructing geological models for simulating both flow and geomechanics. Although such meshing schemes exist for modeling manufactured objects, these are difficult to apply to geological formations; our approach now enables them by guiding the meshing using the chronostratigraphic parameterization of the subsurface. It will allow engineers to routinely include the effects of stress changes during production and will build confidence in development plans.
In recent years, the use of statistics, machine learning and artificial intelligence has attracted a lot of interest. However most of the recently developed techniques are not readily applicable to geophysics, while some of the basic statistical methods, which can add significant value to the process of seismic interpretation, remain underutilized. The advent of new seismic structural attributes and technology to improve their visualization has brought about great improvements in the quality of supporting information available to the seismic interpreter. Calculating many attributes introduces a concomitant problem of how to use them efficiently. The use of multivariate analysis methods can provide a good compromise to get optimal information out of several attributes without overwhelming the interpreter with a surfeit of subtly different sources of information. We present an example case study showing the successful application of these techniques to fault-enhancing attributes in a 3D seismic dataset from New Zealand.
Presentation Date: Tuesday, September 26, 2017
Start Time: 9:20 AM
Location: Exhibit Hall C/D
Presentation Type: POSTER
We present a novel method for decomposing different geometrical characteristics of seismic imaged data using principle component analysis (PCA) applied to directional (dip/azimuth) gathers, and then performing deep learning (convolutional neural network) for automatic classification interpretation. The subsurface geometrical objects to be classified are reflectors (continuous structural surfaces) and different types of diffractors (discontinuous objects such as small-scale fractures and faults). Our preliminary results show superiority over other methods involved in geometrical transformation (e.g., Radon) and specular/diffraction weighted stacks.
Presentation Date: Wednesday, September 27, 2017
Start Time: 10:10 AM
Presentation Type: ORAL
Brazil began oil production from its pre-salt carbonate reservoirs in 2008. Recently, these reservoirs have reached an incredible output of 1.53 million barrels of oil equivalent per day (boed), representing more than half of the country's daily production. The rapid and high increase in oil production demonstrates how important these carbonate reservoirs are to Brazil. The Brazilian pre-salt oil fields are in an exploration and development phase, which requires an understanding of the complex geology of these areas. There are many challenges for characterizing carbonate rocks given their high spatial heterogeneity and complex pore systems. The main objective of this study was to propose a workflow for identifying and characterizing carbonate bodies, using a combination of structural attributes and the hybrid spectral decomposition method on these potential targets in the exploration of hydrocarbons.
Presentation Date: Wednesday, September 27, 2017
Start Time: 2:40 PM
Presentation Type: ORAL
SUMMARY Machine Learning and Neural Networks have been used in the oil and gas industry for several years. The main focus of these technologies has been to predict facies distribution from seismic data, or to cluster log data into electro-facies. Some tentative methods for expanding their applicability have been tested, but to date, these have failed to become part of the main interpretation workow stream. The question is how machine learning technology can be used to perform fault interpretation, AVO analysis and geobody detection more quickly and easily. We propose a new type of machine learning approach to accelerate daily interpretation tasks.
Canning, Anat (Paradigm) | Moulière-Reiser, Dominique (Paradigm) | Weiss, Yuval (Paradigm) | Malkin, Alex (Paradigm) | Phillip, Eitan (Paradigm) | Grinberg, Nimrod (Paradigm) | Teitel, Anastasya (Paradigm) | Reznikov, Margaret (Paradigm) | Yehezkel, Vardit (Paradigm)
Summary We present an artificial intelligence approach for enhancing the frequency spectrum of seismic data. We used synthetic seismic data as a training dataset for constructing a neural networks operator that can solve the posed problem. We then applied this operator to real seismic data and obtained broader frequency spectrum. Introduction Artificial intelligence algorithms have received a lots of attention in recent years. In the context of seismic data processing and interpretation, the range of problems which can benefit from an artificial intelligence approach is very broad.
Lima, C. (Independent Researcher) | Lavorante, L. P. (Independent Researcher) | Williams, W. C. (Louisiana State University) | Beisl, C. (UFRJ-COPPE) | Reis, A. F. C. (Petrobras) | Carvalho, L. G. (Petrobras) | Moriss, M. (Paradigm)
ABSTRACT: This study proposes that a systematic comparison using integrated 3D visualization of all pertinent data (midplate seismicity, geological and geophysical variables) could help in identifying areas vulnerable to injection-induced seismicity in the North American plate. From similar studies of the South American plate in Brazil’s Potiguar basin, it is found that intraplate seismicity occurs at uplifted basin borders (areas over thin, hot, weaker lithosphere) where pre-existing faults are prone to be reactivated by small pressure perturbations. Conversely, central basins (areas over thick, cold, strong lithosphere) are not prone to seismicity. With this model we investigate Oklahoma (Ok) and North Dakota (ND), both intense areas of injection. ND activity, in the central basin, shows no induced seismicity. In contrast, Ok activity, in a regional-scale ravine in the uplifted basin border, has seen a 62.5-fold increase in recent seismicity. Modeling of the Ok region shows reactivation of pre-existing faults with injection pressures of 1.75 MPa (254 psi; 0.7ppg) between 2000-2200m depths, values that agree with wellhead injection pressure field data.
1. INTRODUCTION: THE PROBLEM
A huge increase of seismicity in the tectonically stable U.S. is put into evidence, if we examine the USGS Catalog, 2017 comparing the number of earthquakes of magnitude (Mw) greater or equal to 4 that occurred during 2000-2010 and 2010-2016. For this area, see Fig. 1, we jumped from an average of 6.2 events/yr, during 2000-2010, to an average of 28.8 events/yr, during 2010-2016, roughly a 5-fold increase. For Oklahoma, see Fig. 2, a 62.5-fold increase of seismicity has been observed when comparing these same two periods, including two major events (Mw 5.7, 2011; Mw 5.8, 2016). These recent increases are contemporaneous with the increase in shale production as shown in Figs. 1 and 2. In the stable midcontinent, a roughly 5-fold increase is observed in seismicity during 2010-2016. Again, the increase is contemporaneous with US shale production.
The texture of the rock is an important parameter that must be considered when performing a petrophysical characterization of complex carbonate reservoirs due to the relationship it has with the depositional and diagenetic mechanisms and the effect it has on fluid flow capacity of the rocks.
The presented workflow describes a methodology to extract texture information from resistivity borehole image logs in order to generate a high resolution ordered textural facies log which once integrated with conventional logs greatly contribute to the improvement of facies models.
Two approaches are combined to extract texture from resistivity borehole image logs. The first one relies on the quality index from dips of the stratigraphic laminations detected such as contrast, lamination and planarity to delineate laminated intervals, and characterize them in terms of dip facies by means thickness, density of laminations, mean contrast and mean planarity. Considering the fact that dips are irregularly spaced and computed at point level, the dip facies are computed at point level as well. To overcome this challenge of describing continuous rocks interval, either laminated or not, by means of facies observed at irregular depth, a histogram upscaling technique was applied, as a result, a continuous and regularly sampled log was obtained and it was possible to conciliate high-resolution data with conventional logs to qualify intervals in terms of diagenetic or depositional facies. The second approach comprises the texture detection based on covariance information of resistivity image logs and a posterior facies definition using statistical and unsupervised algorithm that will guide the texture identification as a function of contrast. The integration of the high-resolution facies with production data reveals a good correlation between flow potential and the zones obtained by these approaches.