We present an approach for estimating in-situ relative permeability and capillary pressure through the joint inversion of array resistivity logging and formation test data. Considering a scenario of drilling a vertical well into an oil-bearing formation with water-based mud, the mud-filtrate invasion process can be regarded as a controlled experiment under reservoir conditions. Array resistivity logging can sense the formation resistivity perturbed by the two-phase flow invasion. Formation testing with fluid sampling can also provide information on the radially varying saturation and the associated changes in mobility, as well as information on the effect of capillary pressure. A facies-based workflow is developed to invert for the relative permeability and capillary pressure from the abovementioned two data sets. The inversion strategy is adjustable based on a sensitivity analysis as well as on the data available and the operational sequence of collecting the data. A hybrid inversion framework combining deterministic and stochastic optimization approaches is developed for the inversion of the data.
The tensor-type apparent conductivity acquired by triaxial induction tools enables the determination of both resistivity and resistivity anisotropy of a reservoir. The most popular anisotropic model for triaxial data interpretation is a planarly layered formation with transverse isotropic anisotropy in each bed. This model, the so-called 1D model, is used widely in commercial products in the industry to determine anisotropic resistivity. Such a model, however, does not consider invasion effect. When there is invasion developed in a permeable reservoir, the resistivity derived from inversion based on this 1D model can be substantially affected. The resulting effect can be so strong that the interpretation can become unreliable and even wrong in the worst situations. In this paper, we present an inversion-based answer product that includes the invasion in the interpretation model so that the invasion effect on interpretation is taken into account.
The formation model that we use in the inversion is a multilayer model in which the invasion, e.g., the simple pistol invasion or the annulus invasion, can be modeled in each bed. Due to the existence of invasion, the problem is no longer 1D. A new highly efficient forward solver has been developed to accurately and rapidly model the response of triaxial induction tools to both bed boundary and invasion in multilayer formations. The inversion is based on the Gauss- Newton approach with the L2-norm regularization technique that not only ensures the stability of the inversion, but also enhances the resolution of reconstructed model in thinly layered formations.
The richness of information contained in triaxial induction data makes the inversion work well in various situations. Numerical experiments demonstrate that the inversion is able to reliably correct for the invasion effect and find the right anisotropic resistivity in the virgin zone of invaded beds. In field cases, the anisotropic resistivity determined by the new inversion is more consistent with the petrophysical expectation in the presence of invasion. Moreover, the new inversion provides more-reasonable anisotropy estimations, which would otherwise go abnormally high with the 1D inversion and cause erroneous interpretation in the reservoir.
A multi-step, inversion-based workflow has been developed for analyzing logging-while-drilling density and neutron measurements in high-angle (HA) and horizontal (HZ) wells. The workflow produces accurate layer properties (i.e., bulk density, photoelectric factor (PEF) and neutron porosity) by taking account of bed thickness, borehole effects, and tool response to boundary crossings and adjacent bed effects. The workflow has been validated using both synthetic and field data. A layered earth model is determined as the final result of this workflow, and it can be used as input for subsequent petrophysical interpretations.
The inversion relies on fast forward models for each of the nuclear measurements. These forward models are based on flux derived sensitivity function maps obtained from Monte Carlo modeling. An initial parametric model, including borehole geometry, mud properties, geometric structure, and layer properties, defines a layered earth model along the wellbore trajectory. The initial geometrical model consists of the bed boundary locations and their dip which are determined automatically from the measured compensated density image. Initial layer density, PEF, and apparent limestone porosity are automatically estimated from the respective measured logs. After the initial model setup, a flexible three-step iterative inversion determines 1) mud density and mud PEF, 2) sensor standoff from the borehole wall, 3) layer density, layer PEF, boundary location, and dip. The geometrical model determined from the density inversion is then fixed for the subsequent neutron inversion, which determines apparent limestone porosity for each layer. Gauss-Newton optimization with line search, adaptive regularization scheme, and parameter constraints is used to minimize the weighted
The workflow is validated using synthetic data sets that include both thick and thin-bedded formations and eccentric tool position in the borehole. The true geometrical structure, layer formation density, PEF, and neutron porosity can be recovered within the accuracy of the forward modeling even in the beds where the layer thickness is thinner that the measurements ability to fully respond to the layer property. The workflow has also been applied to field data sets. The inversion results show that it is possible to determine a common geometrical model for the density and neutron measurements, even though they have significantly different responses to the layering. Otherwise, it would not possible to manually derive accurate layer properties due to the asymmetric nature of the neutron measurements and the common practice of attempting to compare the non-azimuthal neutron measurement with the azimuthal density measurements.
The workflow provides an accurate method for quantitative petrophysical interpretations. In addition, the results lead to a better understanding of density and neutron measurements in HA and HZ wells.
Crossbedding can cause significant effect on triaxial induction measurements. Previously we investigated the effect of crossbedding in detail for low angle and vertical wells with a new layered forward model incorporating crossbedding. In this paper, we extend the crossbedding forward model to high angle and horizontal wells. This is achieved by using the continued fraction to tackle the oscillating and oftentimes slowly convergent integrand. Moreover, we demonstrate with numerical experiments that crossbedding can have the same strong effect on triaxial induction measurements as in vertical and low angle wells. The numerical results show that the classical layered model assuming transverse isotropy is not adequate in the presence of strong crossbedding in high angle and horizontal wells. A new layered crossbedding forward model turns out to be a more appropriate choice in such scenarios.
Presentation Date: Wednesday, October 19, 2016
Start Time: 2:45:00 PM
Presentation Type: ORAL
Nuclear measurements in high-angle wells, such as density, neutron, and sigma logs, are often affected by shoulder bed effects and should not be processed directly by the methods which are typically used in vertical wells. Those measurements also have their own unique features and different depth of investigations because of the distinguishing physical principles. It is difficult to have a common formation model that reconstructs the tool responses for all the measurements. In order to remove the geometrical effects and build a consistent formation model among density, neutron, and sigma measurements, we developed a model-based inversion for multiple nuclear measurements in high-angle wells. The inversion algorithm has been validated using synthetic data and applied to field data set. Compared to the input measurements, the inversion results show consistent models, and provide more accurate layer petrophysical properties.
Presentation Date: Thursday, October 20, 2016
Start Time: 8:55:00 AM
Presentation Type: ORAL
Evaluating movable hydrocarbon in shale reservoirs using conventional logs is challenging. In shale reservoirs, the producibility is related to maturity of organic matter and ultimately the type of hydrocarbon present in the rock. Measurements such as total organic carbon (TOC) provide useful input; however, differentiating the organic carbon into kerogen and liquid hydrocarbon requires additional information. Similarly, although the insensitivity of nuclear magnetic resonance (NMR) to kerogen is an advantage, the overlapping T2 response of various fluids brings uncertainty to T2-based NMR interpretation.
Recent laboratory-based experiments have shown significant contrast in the T1/T2 ratio between water in small pores and bitumen or hydrocarbon hosted by organic pores. The two may appear as distinct signal amplitudes on a T1-T2 map. Typical downhole NMR logs have a low signal-to-noise ratio (SNR), which causes the inverted map to have broadened distributions obscuring the individual fluid amplitudes. In spite of broadened distributions, the mean signal corresponds to the relative fluid concentrations and can be converted into saturation, provided the T1 of oil and water with respect to T2 are known. The challenge is that the T1/T2 ratio of various fluid components may be nonunique and is usually unknown.
A new workflow integrates the TOC provided by spectroscopy, bulk density, and T1-T2 NMR measurements to quantify the volumes of kerogen, bound and producible hydrocarbon, and water. The workflow relates TOC to the volume fraction of kerogen, bitumen, and light hydrocarbon. The difference between NMR and matrix-adjusted density porosity provides an estimate of kerogen. A typical T1-T2 map in organic shale is comprised of the signals due to bitumen, light oil, and water. Bitumen and light oil usually have distinct T2 distributions and can be separated if the overlapping water signal is eliminated. The workflow exploits the T1/T2 ratio contrast between oil and water to eliminate the water signal. Although input of the T1/T2 ratio of water is required, the T1/T2 ratio of oil is optimized through an iterative process such that the derived volume fractions of bitumen and light oil satisfy the TOC-to-volume relationship.
The NMR sensitivity to short T1 and T2 components is imperative for the successful application of this method. The data from a new NMR tool fulfilled the requirement and provided the essential input. The integration provided a complete pore fluid volume analysis in several shale reservoirs. The predicted volumes compare well with core-derived measurements.
Anand, Vivek (Schlumberger) | Ali, Mansoor Rampurawala (Schlumberger) | Abubakar, Aria (Schlumberger) | Grover, Rahul (Schlumberger) | Neto, Orlando (Schlumberger) | Pirie, Iain (Schlumberger) | Iglesias, Jorge Gonzalez (Schlumberger)
The fundamental challenges in formation evaluation of unconventional shale reservoirs are the estimation of producible hydrocarbons and identification of target zones for horizontal wells. Modern gamma-ray spectroscopy tools can estimate the in-situ total organic carbon (TOC) which is an important parameter for determining the hydrocarbon potential of shale reservoirs. However, differentiating the components of TOC into kerogen and bound and free hydrocarbon remains an outstanding problem. Resistivity and dielectric measurements are sensitive only to water-filled porosities, and therefore these measurements cannot be used for partitioning the TOC components. Similarly, nuclear magnetic resonance (NMR) T2 cutoff-based methods for the estimation of clay-bound, capillary-bound, and free fluids are generally not valid because the responses of the different fluids overlap in the T2 domain.
In this paper, we describe a novel methodology for quantitative estimation of fluid volumes in unconventional shale reservoirs from NMR T1-T2 measurements. The methodology is based on the mathematical concepts of data analytics for identifying underlying features and systematic relationships in the data without any a-priori information. The NMR T1-T2 data in an entire logged interval are arranged in a database matrix and subsequently resolved as a product of two positive matrices using an automatic feature extraction technique. The first matrix contains the T1-T2 signatures of the different fluids and the second matrix contains the respective volumetric fractions of each fluid type. The methodology is robust and provides accurate extraction of fluid signatures even in low porosity and low signal-to-noise ratio conditions. The extracted fluid signatures could be easily visualized for gaining further petrophysical insights.
The methodology is used to predict continuous logs of fluid volumes in unconventional shale reservoir. High resolution T1-T2 measurements were obtained in an Eagle Ford shale play using a new-generation wireline NMR logging tool. The higher operating frequency of the tool and a new acquisition scheme implemented in the tool result in enhanced sensitivity to the T1-T2 contrasts of fluids and to the fast-relaxing components typical of shale reservoirs. The signatures of different fluids such as bitumen, hydrocarbons in organic matter and inter-particle pores, clay-bound and inter-particle water could be resolved using the methodology described. The fluid volumes show good comparison with core measurements obtained using Dean Stark and tight rock analysis. Integration of fluid volumes from T1-T2 measurements with other advanced wireline logs such as spectroscopy and dielectric measurements provides a complete formation evaluation in shales. A second field example from a heavy oil reservoir shows that the methodology could accurately resolve fluid signatures including heavy oil, clay-bound and capillary-bound water.
Over the years, triaxial induction tools have found applications in not only determining formation resistivity, anisotropy, and dip, but also in detecting fractures. Fracture detection techniques provide useful information in identifying fractured reservoirs because fractures provide both space for hydrocarbon storage and also the channel for fluid flow. When fractures are developed in a formation, the formation can exhibit triaxially anisotropic conductivity if the fractures are filled with a fluid with a significantly different conductivity than that of the formation. This is particularly true for transversely isotropic formations. In an isotropic sedimentary formation, fractures can also cause the formation to be triaxially anisotropic if the geometric properties, orientation and porosity of fractures are varied from place to place in the fractured zone.
In this paper, we present a new method that uses the information of the conductivity tensor to detect fractures. The new method indicates the presence of contrasting fluid-filled fractures by the difference in three conductivity components of the tensor. This difference is called
We have applied the new method to both synthetic and field examples. Synthetic examples show that the rotation operation is indispensable to remove the ambiguity effect of the principal coordinate system. Furthermore, it is shown that the triaxiality is indeed responsive to the presence of fracture. Previous processing of the field example has showed that fractures exist in many zones. The triaxiality index derived from the new method is found in a fairly good agreement with the fracture index from a previous method.
Stockhausen, Ed (Chevron) | Rasmus, John (Schlumberger) | Xie, Hui (Schlumberger) | Morriss, Chris (Schlumberger) | Ito, Koji (Schlumberger) | Griffiths, Roger (Schlumberger) | Maggs, David (Schlumberger) | Abubakar, Aria (Schlumberger)
Conventional petrophysical workflows that make use of point-by-point inversions of measured log data to determine reservoir properties at every sampled measure point may now be rendered obsolete. These workflows and inversions assume the tool is surrounded by an infinite homogeneous medium and the tool sensors are unperturbed by adjacent layers. This assumption is approximately correct when vertical wells are drilled through thick massive reservoirs. In this case, 1D inversions are available for the deeper-reading induction measurements to provide approximate corrections at layer boundary crossings. However, these inversions are for a single tool, and the position of each boundary is placed to satisfy that tool’s responses without regard to the actual geological boundary position. Additionally, these inversions do not incorporate any nuclear measurements, making the computation of saturation using a 1D squared log of resistivity with a simple measured nuclear log impractical.
Reservoirs today are being developed using high-angle and horizontal (HAHZ) wells using a single platform or pad for multiple wells. This reality, and the fact that thin-bedded formations such as organic shale reservoirs are now being exploited, results in both resistivity and nuclear measurements that are affected by multiple layers at each measured depth, regardless of whether the conveyance is by logging-while-drilling (LWD) or wireline, and in measurements that are not characteristic of any one layer. Interpretation is further complicated by the fact that some measurements are directional and measure at some particular azimuth of the formation, e.g., wireline pad density measurements or best-contact LWD density measurements, and others, such as resistivity and neutron, are omnidirectional and respond to properties at all azimuths.
A new workflow is developed to address the geometrical and tool response issues associated with HAHZ well measurements. First, we take advantage of the high “effective” resolution of the density measurement that is a result of the high relative bed dip to define petrophysical layers as thin as 2-in. true stratigraphic thickness (TST) space. Next, the true-layer log properties are determined from inversion and forward modeling. This allows us to compute the layer petrophysical properties of porosity, saturation, fluid type, and permeability using conventional petrophysical algorithms. Another unique aspect of the workflow is that properties are also determined for the non-crossed layers—those that are proximal to and within the volume of investigation of the measurement, but not actually crossed by the well trajectory, i.e., for parallel bed conditions.
The reservoir hydrocarbon pore volumes and permeability can now be computed on a layer-by-layer basis free of adjacent bed and bed-crossing effects. The integration of these petrophysical layer properties can now be used as input to reservoir property modeling and upscaling exercises. In a particular field case, the new workflow computed layer boundaries and porosity and permeability match core properties in beds as thin as 3 in. whereas the previous vertical well derived properties do not. The use of the newly derived layer properties enables us to more accurately quantify porosity and permeability and explain the unexpected high hydrocarbon flow rates in some layers and the early water breakthrough from water injection in others. With this new information we can now place subsequent development wells in optimal positions to increase ultimate recovery.
The concept of triaxial induction measurements can be traced back to 1960s. The first commercial triaxial induction tool has been in service for more than a decade. The tensor type of measurements acquired by the triaxial induction tool enables extracting not only the resistivity anisotropy but also the dip of a subsurface formation. The first anisotropic resistivity model that was developed is the transverse isotropic (TI) model, in which the principal axis associated with the vertical resistivity (Rv) is perpendicular to the bedding plane. This model is in widespread use in commercial products by oilfield service companies to provide resistivity anisotropy and the dip of a formation.
Image data and outcrops show that crossbedding occurs in multiple depositional environments. The application of an existing fast forward computer model for crossbedded formations has been limited to vertical wells. We have developed a new fast computer model for the same problem that works not only in vertical wells but also in deviated wells. This new forward model enables the user to study the triaxial tool response more efficiently in multilayer formations. In particular, it supports detailed investigation of the tool response where both dipping beds and crossbedding exist in the model.
We have used this model to simulate the triaxial tool response extensively in some typical formations. The simulation results demonstrate that the dipping bed effect is tightly coupled to the effect of anisotropy due to crossbedding in thin-layer formations. The comparison of the new crossbedding model with the previous TI model suggests that the inversion based on the TI model inversion is not adequate in the presence of strong crossbedding, and both resistivity and dip results are adversely affected.
The new crossbedding forward model makes it possible to have an accurate interpretation in dipping crossbedded formations. A new inversion-based interpretation workflow is developed using the new crossbedding model. The new method delivers one bedding dip and multiple bed-to-bed anisotropy dips over the zone of interest. Results show that in the presence of crossbedding, the method is able to distinguish between bedding dip and anisotropy dip, and provides better resistivity interpretation.