The Alvheim Field in the Norwegian North Sea was discovered in 1998. Two wells were drilled in 2018 in the Gekko structure to confirm oil column height and to evaluate reservoir quality in the Heimdal Formation. A comprehensive wireline logging program, including NMR and formation testing, was optimized to reduce formation evaluation uncertainty. Evaluating fluid properties, oil column height, and reservoir quality were primary objectives. Well A was first drilled on the south of the structure, followed by Well B on the north of the structure. Reservoir quality encountered in both wells was very good, and a project to develop these resources is currently in the selection phase. Formation evaluation uncertainty encompassing pore geometry distribution, permeability, reservoir quality, and hydrocarbon identification are mitigated by studying the nuclear magnetic resonance (NMR) log response. NMR fluid typing has been widely used in the oil industry since the 1990s. NMR fluid typing today is a combination of the contrast of spin relaxation time T 1 , the spin-spin relaxation time T 2 ( T 1 T 2 ), and the diffusivity ( T 2 D ) of formation fluids (Chen et al., 2016). NMR fluid typing can be obtained from a continuous log and/or stationary log measurements. This paper showcases excellent, textbook-quality NMR data, as well as the integration of NMR data in the petrophysical workflow. High-confidence fluid properties and fluid contacts are determined. This paper also highlights a comparison of NMR data acquired in stationary vs. continuous depth-based log modes in both wells. The continuous log data quality is equivalent to stationary data, implying continuous log data quality is sufficient for reliable NMR fluid properties evaluation without depending on time-consuming stationary NMR measurements. Reducing logging operations rig time is very advantageous in the North Sea, where drilling rig operations cost is high, and enhanced rig time management is constantly required.

In this paper, we examine fluids interpretation techniques in a prolific oil field in offshore West Africa. A sourceless logging program, consisting of logging-while-drilling (LWD) nuclear magnetic resonance (NMR), resistivity, and formation tester, was chosen to log the reservoir section in 6.5-in. holes. The purpose of this study is to answer questions related to asset appraisal and development with these limited measurements. Core data available are porosity, permeability, water salinity, Archie m and n , and Dean-Stark S w . A comparison of the core and NMR log indicates that NMR total porosity is not affected by hydrocarbon in the pore space. We use a statistical method called factor analysis to deconvolve independent fluid modes from the T 2 distribution and pick the T 2 cutoff. The NMR irreducible water saturation ( S wirr ) computed with this cutoff agrees with Dean-Stark S w . Continuous S w is calculated with Archie’s equation with lab-measured parameters and validated against Dean-Stark S w above the transition zone. The Timur-Coates model is used to estimate matrix permeability. The first application of this interpretation workflow is to confirm the free-water level (FWL) derived from pressure gradients. We found the S w profile largely controlled by heterogeneity in rock textures. The presence of both good and poor-quality rocks makes log-based FWL picking difficult. We use S wirr from NMR to indicate rock quality and simplify our final interpretation. The FWL found by sourceless log interpretation is consistent with the initial FWL found by pressure gradients. The second application is perforation design. Zones with good porosity and low mobile water volume are selected for perforation, and a safe distance is maintained from FWL. As a result, all producer wells exhibit zero water cut.

The computation of permeability is vital for reservoir characterization because it is a key parameter in the reservoir models used for estimating and optimizing hydrocarbon production. Permeability is routinely predicted as a correlation from near-wellbore formation properties measured through wireline logs. Several such correlations, namely Schlumberger-Doll Research (SDR) permeability and Timur-Coates permeability models using nuclear magnetic resonance (NMR) measurements, K-lambda using mineralogy, and other variants, have often been used, with moderate success. In addition to permeability, the determination of the uncertainties, both epistemic (model) and aleatoric (data), are important for interpreting variations in the predictions of the reservoir models. In this paper, we demonstrate a novel dual deep neural network framework encompassing a Bayesian neural network (BNN) and an artificial neural network (ANN) for determining accurate permeability values along with associated uncertainties. Deep-learning techniques have been shown to be effective for regression problems but quantifying the uncertainty of their predictions and separating them into the epistemic and aleatoric fractions is still considered challenging. This is especially vital for petrophysical answer products because these algorithms need the ability to flag data from new geological formations that the model was not trained on as “out of distribution” and assign them higher uncertainty. Additionally, the model outputs need sensitivity to heteroscedastic aleatoric noise in the feature space arising due to tool and geological origins. Reducing these uncertainties is key to designing intelligent logging tools and applications, such as automated log interpretation. In this paper, we train a BNN with NMR and mineralogy data to determine permeability with associated epistemic uncertainty, obtained by determining the posterior weight distributions of the network by using variational inference. This provides us the ability to differentiate in- and out-of-distribution predictions, thereby identifying the suitability of the trained models for application in new geological formations. The errors in the prediction of the BNN are fed into a second ANN trained to correlate the predicted uncertainty to the error of the first BNN. Both networks are trained simultaneously and therefore optimized together to estimate permeability and associated uncertainty. The machine-learning permeability model is trained on a “ground-truth” core database and demonstrates considerable improvement over traditional SDR and Timur-Coates permeability models on wells from the Ivar Aasen Field. We also demonstrate the value of information (VOI) of different logging measurements by replacing the logs with their median values from nearby wells and studying the increase in the mean square errors.

Many methods of calculating water saturation require knowing the chloride concentration in formation water. Chlorides have a strong effect on water properties, and they impact saturation estimates that are based on resistivity, dielectric dispersion, or thermal neutron absorption. Here we introduce a new direct quantitative measurement of formation chlorine from nuclear spectroscopy, enabling a continuous log of water salinity within a limited radial depth. Neutron capture spectroscopy is sensitive to chlorine and is a natural fit for measuring its concentration, except that the spectrum contains chlorine from both the formation and borehole. The borehole chlorine background can be large and is highly variable. Historical efforts to derive water salinity from spectroscopy have relied on ratios of chlorine and hydrogen, which are affected by the borehole and hydrocarbons. The direct use of chlorine provides a more reliable basis for salinity interpretation after isolating its formation signal. We partition the borehole and formation components of chlorine via two unique spectral standards. The contrast between the two standards arises from differences in gamma ray scattering based on their point of origin. The shape of the borehole chlorine standard must be adjusted along depth to account for environmentally dependent scattering, which we achieve with a continuously varying function of borehole and formation properties. The algorithm is derived from 129 laboratory measurements and 2,995 numerical simulations spanning a diverse range of conditions. The remaining signal is converted into a log of formation chlorine concentration. In combination with total porosity, chlorine concentration sets a minimum value for water salinity. Adding an organic carbon measurement enables the simultaneous estimation of water volume and salinity. Chlorine concentration can also be combined with a selected water salinity to compute a water volume for comparison with other methods. Finally, chlorine concentration enables calculation of a maximum expected sigma, which can identify the presence of excess thermal absorbers in the matrix. The systematic uncertainty on the chlorine concentration ranges from 0.03 to 0.07 wt%, depending on borehole size. The resulting salinity accuracy is inversely proportional to porosity. A potential limitation of the measurement is its depth of investigation, reaching 8 to 10 in. for 90% of the signal. The chlorine concentration is sensitive to filtrate or connate water, depending on formation permeability and invading fluids. We first present the technique to measure formation chlorine, supported by modeling, laboratory data, and core-log comparisons. We then propose petrophysical workflows to interpret the chlorine concentration.

Running pulsed-neutron logs in Malaysia has previously been plagued by results with high uncertainties, especially in brown fields with complex multistacked clastic reservoirs. Together with a wide range of porosities and permeabilities, the acquired logs quite often tended to yield inconclusive results. In addition, the relatively fresh aquifer water (where salinity varies from 5,000 to 40,000 ppm) makes reservoir fluid typing and distinguishing between oil and water even more challenging. As a result, the inconsistencies and uncertainties of the results tend to leave more questions than answers. Confidence in using pulsed-neutron logging, especially to validate fluid contacts for updating static and dynamic reservoir models, deteriorated within the various study teams. Due to this fact, the petrophysics team took the initiative to conduct a three-tool log off in one of their wells with the objective of making a detailed comparison of three pulsed-neutron tools in Malaysia’s market today. The main criteria selected for comparisons were the consistency of the data, repeatability, and statistical variations. With recent advancements in pulsed-neutron (multidetector) tool technology, newer tools are being equipped with more efficient scintillation crystals, improving the repeatability of the measurements as well as the number of gamma ray (GR) count rates associated with the neutron interactions. In addition, the newer tools now have up to five detectors per tool, with the farthest detector supposedly being able to “see” deeper into the formation, albeit at a lower resolution. With these new features in mind, the log off was conducted in a single well with a relatively simple completion string (single tubing, single casing), logged during shut-in conditions only, and the logs were acquired directly one after the other (back to back) to avoid bias to any particular tool. Both sigma and spectroscopy measurements were acquired to compare the capabilities of each tool. Due to the relatively freshwater salinity, the carbon-oxygen (C/O) ratio from the spectroscopy measurements is used to identify the remaining oil located in the reservoirs, while the sigma measurements determine the gas-oil or gas-water contact, if present. This paper will illustrate the steps taken by Petronas Carigali Sdn Bhd (PCSB) to compare the raw data and interpreted results from the three pulsed-neutron tools. Consequently, a comparison from all the tools was made to the current understanding of the reservoir assessed. The points from these comparisons will then show which tools are favored over the rest.

The most widely used Russian electrical and electromagnetic logging tools are the high-frequency electromagnetic sounding tools (VEMKZ/VIKIZ), the unfocused induction tool (IK), the unfocused lateral logging tool with gradient and potential arrays (BKZ), and the focused lateral logging tools with two three- electrode arrays (BK). A direct comparison of responses from these tools with responses from three benchmark Western tools—the Schlumberger HRLA high-resolution laterolog array tool, Schlumberger AIT array induction tool, and Schlumberger Rt Scanner triaxial array induction tool—was achieved using simulated tool responses from 1D and 2D isotropic and anisotropic synthetic models of a conventional clastic reservoir typical of Western Siberia. A brief history of the Russian logging tools includes corresponding bibliography on tool theory, hardware, and processing options. We present the tool configuration for the most common versions of the Russian tools, describe some numerical algorithms used in this paper to calculate the tool responses in 1D and 2D models, discuss and illustrate some workflows that provide quality control (QC) of Russian logs on the basis of radial 1D modeling, address environmental corrections, and introduce standard interpretation of Russian logs. We also briefly mention the advanced processing tools, such as 2D inversion of BKZ data and joint 2D inversion of BKZ and VEMKZ logs. The comparison of tool responses in isotropic and anisotropic benchmark models includes comments from the point of view of Russian and Western log analysts. The comparison illustrates that in simple, mostly uninvaded models, focused responses of the HRLA and AIT tools have some advantages compared with the unfocused BKZ and VEMKZ measurements. However, simulated tool responses in realistic 2D isotropic and, especially, anisotropic models showed that the advanced 2D inversion- based processing is required for both Russian and Western tools to determine parameters of the virgin formation and invaded zone accurately.

Sustained E&P activity levels and slim margins on highly valued Permian Basin acreage drive operators to leverage information as much as possible and in ways not seen in the recent past. Data accuracy, especially in this fast-paced, competitive environment, is strongly desired. Core analyses provide subsurface static calibration, but the thick stratigraphic section comprised largely of sublog scale facies, challenges a cost-effective approach to collect sufficient calibration data. Saturation determination is a key petrophysical deliverable that has multiple uses, including landing zone assessment. Calibration of saturation models may originate in several ways: proprietary or joint venture core, industry consortia databases, data trades with other operators, government databases, or publications. Internal and external reviews of subsurface model inputs have repeatedly shown that Permian Basin saturations, in particular, have a wide distribution and large uncertainty. Accurately measuring core fluid saturations in tight rock continues to pose significant challenges originating from the currently accepted laboratory methods, assumptions used to interpret those data and more broadly, due to increased relative uncertainty associated with tight, low-porosity formations. For example, crushing core samples, which enhances fluid extraction in tight rocks, causes systematic fluid losses in the case of core samples of liquid-rich mudstone formations, which are not typically quantified. Instead, as-received air-filled porosity is commonly assumed to represent hydrocarbons that were forced from core during acquisition/retrieval due to gas expansion. Additionally, fluid extraction from commercially available retorting systems have widely variable fluid collection efficiencies (<100%) resulting in significant inconsistencies between the weight of collected fluids and sample weight loss during retorting experiments. The Dean-Stark technique removes not only water and oil, but an unknown volume of solvent-extractable organic matter, and it only allows for direct quantification of the extracted water volume. Finally, fluid and solid losses during handling in the laboratory are unassessed in current commercial laboratory procedures. The reconciliation of fluid volumes with fluid and sample-weight data delivered by either of the two techniques, i.e., retort or Dean-Stark, requires numerous assumptions about pore fluid properties, which are typically not verified through direct measurements. We demonstrate that such assumptions can lead to extreme uncertainty in estimates of water saturation. To address such critical uncertainties, a new retort-based core analysis workflow using improved core characterization and fluid-extraction techniques was developed. In one advancement, this workflow employs NMR measurements systematically performed on all as-received and crushed samples to quantify fluid losses during crushing. This approach also uses a specially developed fluid collection apparatus with close to 100% fluid collection efficiency. In addition to these advances in measurements, the workflow is optimized to avoid fluid losses during sample handling and includes repeated grain density and geochemical measurements at different stages for quality control (QC). As a result, the new workflow reduces the uncertainties in acquired data and better addresses the assumptions, i.e., parameter corrections for fluid losses, in interpreting measured data into core total porosity and core fluid saturations. The workflow is demonstrated for a set of Delaware Basin Wolfcamp A formation samples and the results suggest that previous crushed-rock core analysis protocols underestimate water saturation by at least 30% or ~15 saturation units (s.u.) for this liquid-rich mudstone formation.

Accurate interpretation of the petrophysical properties of gas-hydrate-bearing sediments, such as porosity, hydrate saturation and clay content, are of great importance for reservoir characterization and resource evaluation. Typically, these parameters are estimated using either elastic properties or electrical properties instead of both. We propose to take advantage of multiple types of measurements and improve the accuracy of prediction by using an inverse rock physics modeling (IRPM) method, which allows us to combine elastic and electrical attributes. First, we generate constraint cubes of 3D elastic and electrical data in the reservoir parameter domain using suitable rock physics models calibrated by 3D elasticelectrical rock physics templates (RPTs). Then, we extract the isosurfaces from the 3D elastic and electrical data constraint cubes with the marching-cubes algorithm. Finally, we use the iterative least-squares method to find the optimal intersection point of three isosurfaces by minimizing the objective function. To demonstrate the feasibility of this strategy, we apply it to synthetic data and well logs measured at the Ocean Drilling Program (ODP) Hole 1247B drilled on the Hydrate Ridge, South Cascadia Margin. For the synthetic data, the estimated Petrophysical properties are consistent with those produced using noise-free initial synthetic model parameters. In addition, our estimated results for real field localities consistently fit with the core data. The smaller root-mean-square errors between inversion results and referenced Petrophysical properties for both synthetic case (≤ 0.06) and real field data (≤ 0.061) further confirm that the inverse rock physics modeling method is feasible for estimating petrophysical properties by integrating elastic and electrical properties.

The process of well-log correlation requires significant time and expertise from the interpreter, is often subjective and can be a bottleneck to many subsurface characterization workflows. Algorithmic approaches to well-to-well correlation suffer from the inherent heterogeneity of geophysical measurements in the wellbore, both from a geologic and data-quality perspective. We demonstrate a rigorous and repeatable method for well-log correlation by deploying a correlation tool that leverages a machine learning model for pattern matching between well logs and programmed stratigraphic correlation techniques. A supervised-learning approach was used to train a novel deep convolutional neural network (CNN) architecture using over five million data samples, which were derived from thousands of well logs and expert interpreted correlations. To ensure that a robust pattern-matching model was trained, well logs from several US onshore basins with various tectonic regimes and environments of deposition were used to construct training and validation datasets. The result is a universal model for pattern matching of wireline measurements that can incorporate multiple geophysical-log signals as input data and can be deployed at scale without the need for retraining. Overall, the pattern-matching model was able to achieve a level of accuracy of 96.6% and classification area-under-the curve (AUC) of 0.954 on a separate validation dataset. The universal deep CNN is one component of the correlation tool. Algorithmic three-dimensional search logic was constructed around the deep CNN model which determines the optimal correlation and marker propagation pathway. Rules-based criteria have also been applied to the model output ensuring conformance to stratigraphic principles including preserving stratigraphic order and honoring present-day structural trends. We present several examples to highlight the strengths and weaknesses of this machine-learning-based approach to well-log correlation which can be used to efficiently generate high-density datasets for regional exploration, development mapping and reservoir characterization exercises.

Wettability is a crucial factor for the dynamic properties of oil reservoirs. Early recognition of the wettability condition of a recovery may have a significant impact on the development options and of the expected recovery factor. NMR relaxation times of pore fluids are dependent on the wetting through surface relaxation, and are thus known to contain this valuable information. This paper describes an easy-to-implement and reliable procedure to calculate a quantitative wettability index from standard NMR measurements, such as can be made in conjunction with SCAL experiments. The interpretation is fully auditable, and therefore suitable to be part of a standard protocol.

Prologue The standard model for relating bulk formation resistivity to porosity and water saturation was introduced to the petroleum industry in 1941; it remains the industry standard to this day. The model was discovered empirically by means of graphical analysis. Basically, G.E. Archie discovered that when the logarithm of formation resistivity factor was plotted against the logarithm of porosity the resulting trend could be fitted by a straight line. A similar relationship was discovered connecting the logarithms of resistivity index and water saturation. When these two power laws are combined into a single equation, it can be solved for water saturation (which is not observable from a borehole) in terms of bulk formation resistivity, interstitial brine resistivity, and porosity (all of which can be estimated from observations made in boreholes). This revolutionized log interpretation. There has always been a problem with the model in terms of its "explainability". That is, it cannot be derived in any straightforward way from accepted first principles of physics. It does not contradict any first principle, but neither does it seem to follow ineluctably from them. However, since the model works, most formation evaluators have memorized the relationships that follow from the model and simply "get used to them". That remains the situation to this day. However, there is a path around this obstacle to understanding formation resistivity at a fundamental level, and that way forward is to abandon the resistivity formulation in favor of its reciprocal, conductivity. It is surprising that such a seemingly trivial change could open a new vista into the relationships among formation electrical properties. A conductivity formulation permits the asking of questions about how a formation's conductivity should respond to changes not only in brine conductivity, but also in the fractional amount of brine in a formation, and its geometrical configuration. By answering these questions in an obvious way, and with some analysis of data taken in the laboratory, an intuitively obvious model explaining bulk formation conductivity emerges. The model is not the same as the Archie model. However, when certain parameters are taken to their limiting values, and the model is converted into resistivity space, Archie's power law model is revealed as an approximation to the limiting cases. Thus, from the conductivity formulation, an intuitive understanding of the Archie model emerges. Moreover, the conductivity model can be derived in at least three different ways, each yielding different insights into formation conductivity.

A new sonic-imaging technique uses azimuthal receivers to determine individual reflector locations and attributes, such as the dip and azimuth of formation layer boundaries, fractures, and faults. From the filtered waveform measurements, an automated time pick and event-localization procedure is used to collect possible reflected arrival events. An automated ray-tracing and 3D slowness time coherence (STC) procedure is used to determine the raypath type of the arrival event and the reflector azimuth. The angle of incidence of the reflected arrival is related to the relative dip, and the moveout in 3D across the individual sensors is related to the azimuthal orientation of the reflector. This information is then used to produce a 3D structural map of the reflector, which can be readily used for further geomodeling. This new technique addresses several shortcomings in the current state-of-the-art sonic-imaging services within the industry. Similar to seismic processing, the current sonic-imaging workflow consists of iteratively testing migration parameters to obtain a 2D image representing a plane in line with the desired receiver array. The image is then interpreted for features, which is often subjective in nature and does not directly provide quantitative results for the discrete reflections. The technique presented here, besides providing appropriate parameter values for the migration workflow, further complements the migration image by providing dip and azimuth for each event that can be used in further downstream boundary or discontinuity characterization. A field example from the Middle East is presented in which a carbonate reservoir was examined using this technique and subsequently integrated with wellbore images to provide insight to the structural geological setting, which was lacking seismic data due to surface constraints. Structural dips were picked in the lower zone of the main hole and used to update the orientation of stratigraphic formation tops along the well trajectory. 3D surfaces were then created and projected from the main hole to the sidetrack to check for structural conformity. One of the projected surfaces from the main hole matched the expected depth of the formation top in the sidetrack but two were offset due to the possible presence of a fault. This was confirmed by parallel evaluation of the azimuthal sonic-imaging data acquired in the main hole that showed an abrupt change in the relative dip of reflectors above and below the possible fault plane using the 3D STC and ray tracing. Dip patterns from both wells showed a drag effect around the offset formation tops, further confirming the presence of a fault. A comparison of the acquired borehole images pinpointed the depth and orientation of the fault cutting both wells to explain the depth offset of the projected 3D formation top surfaces.

ABSTRACT The increasing complexity of downhole fluid sampling and the need for optimizing the sampling process call for a model-based approach in planning and interpretation applications. Efficient planning of fluid sampling requires quantitative evaluation of sampling hardware performance over a wide range of deployment conditions. For real-time contamination monitoring, recent work demonstrates that a model-based approach improves contamination estimates, especially in difficult sampling environments and for complex tool geometries (e.g., focused probes). Finally, both real-time and post-job interpretation of formation properties often require a model-based approach, in which parameters of a reservoir model are inverted using downhole fluid analysis (DFA) measurements. The objective of this study is to develop a comprehensive set of forward models of filtrate contamination cleanup, including methods for speeding up model evaluation to enable applications in real-time inversion and uncertainty quantification. Building on previous work in this area, we present numerical forward models for immiscible (e.g., oil sampling in water-based mud) filtrate cleanup. The models cover both conventional and focused sampling tools, and account for complex tool operating modes. Exploiting cloud-based simulation, we demonstrate how accurate proxy models can be constructed from a large number of precomputed numerical simulations. The resulting proxy models enable rapid interpretation workflows. An example of such a workflow is presented for relative permeability inversion from wireline formation tester measurements of water cut and pressure. This recently developed methodology complements laboratory measurements of relative permeability on core samples. The time required for inversion workflow execution is greatly reduced by using fast proxy models instead of full numerical simulations.

ABSTRACT Interpretation of electrical resistivity measurements for assessing hydrocarbon saturation in mixed-wet and hydrocarbon-wet rocks often requires extensive recalibrations of resistivity models. In these conventional resistivity models, the impact of wettability is not reliably incorporated. Recently, we analytically derived a new resistivity model that incorporates parameters to account for wettability and complexity of pore structure. This new model requires experimental verification to enhance its applicability in core- and log-scale domains. In this paper we (1) experimentally quantify the influence of wettability on electrical resistivity measurements, (2) improved and verified our analytically derived resistivity model in rocks with different levels of wettability at different water saturations, and (3) demonstrated the physical meaning of the parameters related to wettability in the resistivity model. We altered the wettability of selected core samples from the same rock type. To prepare the core samples as mixed-wet, a surfactant solution was injected into the samples and then they were aged in decane. These approaches created a range of wettability states in the samples. To quantify the altered wettability of the samples, we used the combined USBM and Amott-Harvey method along with sessile drop contact angle measurements. Next, we used a centrifuge to vary the water saturation in the core samples. We then measured the electrical resistivity of each sample. Finally, we improved the recently introduced resistivity model and compared the estimates of resistivity index from the improved model against the experimentally measured resistivity indices. We successfully verified the reliability of the improved resistivity model for mixed-wet carbonate core samples. The wettability of the core samples was altered to be in the range of 0.1 to 0.4 on the Amott wettability scale. We also demonstrated that all the coefficients required by the improved resistivity model are physically meaningful. One of the unique contributions of this paper is the introduction of a new interpretation diagram, called the wettability triangle. This diagram can potentially be used to quantify wettability from resistivity measurements, if combined with other geophysical measurements. The outcomes of this work are promising for reliable interpretation of resistivity logs in mixed-wet formations for improved assessment of hydrocarbon saturation with minimal calibration efforts.

ABSTRACT Workflows for log analysis in conventional reservoir rocks are difficult to apply in organic-rich mudrocks due to the presence of abundant kerogen (solid, insoluble organic matter). There are two key reasons for this difficulty. First, kerogen is part of the solid matrix but responds like a fluid in traditional porosity log analysis (e.g., density, neutron). Accurately estimating formation volumes from logs, therefore, requires a method to separate the kerogen and fluid signals in the matrix and porosity, respectively. A second difficulty is that the petrophysical properties of kerogen (such as its density and hydrogen index) vary by more than 50% relative, impacted by geologic forces including maturation, and are rarely known. No logging tools can measure kerogen properties directly. This paper describes a novel, wellsite method combining log and cuttings analysis that overcomes these challenges, providing a measured separation of kerogen and pore fluids and a direct estimate of kerogen properties. The cuttings analysis uses diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). Prior to analysis, cuttings are cleaned of formation fluid and drilling mud, isolating the kerogen signal. DRIFTS measures the vibrational frequencies of chemical bonds in matrix components, including kerogen, and newly developed transforms quantify kerogen properties directly from its DRIFTS signal. The method has been optimized for analysis of cuttings and can be performed at the wellsite, for vertical and deviated wells, with any mud type. Examples illustrating the integration of cuttings and downhole log analysis are shown. In one, DRIFTS provides a matrix-adjustment to porosity logs giving a direct measurement of porosity, without the need for subjective assumptions that are needed in standard porosity log interpretations of organic-rich formations. In another, DRIFTS estimates anisotropic elastic and stress profiles along a lateral after calibration in a vertical pilot. Integration of cuttings and downhole log analyses provides a practical method to solve many challenges inherent to log interpretation in organic-rich formations.

ABSTRACT A key objective for formation evaluation in unconventional reservoirs is to estimate reservoir quality by quantifying the volumes of different fluid components. Spectroscopy-based tools can estimate the total organic carbon in a reservoir. Resistivity and dielectric tools are sensitive to the water-filled porosity. On the other hand, nuclear magnetic resonance (NMR) tools have the capability and sensitivity to further partition the hydrocarbon and water into fluid components based on their properties and location in the pore space. T 1 -T 2 maps from NMR logging tools show unique signatures for hydrocarbons, such as gas, bitumen, and producible and bound oil. Similarly, capillary and clay-bound water and water in larger pores have different signatures. These signatures depend on many factors: properties of the fluids (composition, viscosity), properties of the rock (pore geometry) and the geometrical configuration of fluid phases within the pore space. Unless the fluids have very distinct and nonoverlapping signatures in the T 1 -T 2 domain, it is challenging to visually separate the contribution from different fluids and estimate fluid volumes from T 1 -T 2 maps. This problem was addressed by an automated unsupervised learning algorithm called blind-source separation (BSS), wherein the NMR T 1 -T 2 maps of an entire logged interval are factorized into two matrices: the first matrix contains the T 1 -T 2 signatures of the different fluids, and the second contains the corresponding volumes. This method has been shown to work well on multiple field datasets, where there was a sufficient dynamic range in the underlying volume fractions. In this paper, we address two well-known limitations in the BSS algorithm. First, the algorithm assumes a dynamic range in the volume fractions. For this reason, the entire logged interval is considered in the matrix factorization. However, doing so mixes the effects T 1 -T 2 maps due to changes in rock properties with changes in fluid volumes. Second, it assumes that the number of sources (or fluids) is known a priori. This is a well-known ill-conditioned problem. We propose several modifications in the algorithm to address the above limitations. First, we leverage the information that the NMR signature of a fluid is expected to be connected in the T 1 -T 2 domain. Second, we assume that each point in T 1 -T 2 space corresponds, at most, to one fluid. Lastly, we propose a quantitative metric to guide the analyst in selecting the number of components. We demonstrate the application of this method on simulated datasets as well as field datasets from the Eagle Ford formation and Permian Basin.

ABSTRACT Production logging in deviated wells has shown so far limited success in providing a reliable and cost-efficient method for production profiling. The reasons are numerous, but two factors are mainly responsible: conventional array production logging tools constitute very long toolstrings associated with expensive deployments, and data analysis is complex, requiring time-consuming analysis from cased-hole production logging experts. Overall, despite heavy investments made to perform downhole measurements, results are often disappointing and interpretation affected by a great deal of uncertainty. A new instrumentation technology using microelectromechanical systems (MEMS) as well as new interpretation methods is offering new perspectives to this domain. In this paper, we describe a short and modular multiphase flow-sensor platform capable of achieving up to 10 times reduction in toolstring length compared to existing technology. Miniature pressure, temperature, optical, electrical, acoustic, microspinners, and ultrasonic Doppler sensors can be mounted independently from each other and easily interchanged to adapt the tool to address the well-specific challenges and meet the objectives of the surveillance program. By using both the sensor multiplicity made possible by hardware miniaturization, and diversity from multiple measurements, resolution and robustness is greatly improved. In addition, the use of digital power integrated into each individual sensor (smart sensor) provides calibrated data output, which highly facilitates the interpretation. The new technology is an ultracompact array platform, with new sensors designs, whose superior measurements are enhanced by the new interpretation methods it enables.

ABSTRACT Distributed fiber-optic sensing, and specifically the introduction of intelligent distributed acoustic sensing (DAS), has gained the attention of production engineers with the promise of a versatile and cost-effective decision-support tool. These systems can either be permanently installed, or temporarily deployed using diverse types of intervention systems. This article covers the principles of flow allocation using distributed sensing and show how these can be used and combined to identify fluid-entry points, quantify production and identify fluid phases. We will describe the methods used to improve quantitative interpretation from distributed sensing, especially the use of phase-coherent DAS for quantitative measurement of sound speed and its use in analysis of flow velocity and fluid phase. While early DAS systems were previously limited in their flow-detection thresholds we have recently introduced a new sensing system, bringing a 20dB (100×) improvement in signal-to-noise. This offers a significant improvement in measurement and associated interpretation capability. INTRODUCTION Distributed fiber-optic sensors were invented in the 1980s (Hartog, 1983) and introduced into the oilfield in the 1990s. The initial areas of interest, and commercially available technologies, were related to distributed temperature sensors (DTS) and distributed strain sensors (DSS). DTS was applied to leak detection, flow profiling and steamflood-monitoring applications (Smolen and van der Spek, 2003). DSS focused mainly on wellbore integrity, monitoring strain induced on wellbore casings (Li et a., 2004). Some research has also been carried out on the use of DSS systems for distributed pressure sensing, but to date, these have not delivered the required performance and reliability for commercial application.

ABSTRACT Azeri-Chirag-Guneshli (ACG) is a giant field located in the Caspian Sea, Azerbaijan, operated by BP Exploration (Caspian Sea) limited. The major reservoir zones comprise of sandstone formations with 20 to 25% porosity, 100 to 1,000 md permeability, and an oil column up to 1,000 m. Simultaneous development of several relatively young reservoirs by commingling production and injection in wells, the relatively high cost of intervention in the Caspian region, the configuration of platforms with restricted access to the well heads during drilling operations, and severely limited simultaneous rig and intervention operations, have been some of the key drivers in selecting, adopting and developing appropriate technologies for reservoir and well surveillance in ACG. This paper describes the developments implemented in the field over the last 20 years to address the gaps in conventional technologies and enhance the efficiency of surveillance from: conventional production logging sensors (with surface- and memory-readout conveyance); array production logging, (3) the use of permanent downhole pressure-temperature gauges (PDHG); and permanent installations of distributed fiber-optic (DFO) cables for distributed temperature sensing (DTS) and distributed acoustic sensing (DAS). The paper also demonstrates how integration of data acquired from these technologies with data from other means of surveillance have helped improve our understanding of well and reservoir dynamics and subsequently, make important business decisions at different stages of field life. The paper also discusses new methods developed for automated data streaming, visualization and interpretation with specific reference to (DFO) sensor data for downhole fluid-surveillance applications. The use of such techniques enables seamless integration and real-time interpretation of huge volumes (~1 TB/hr) of DFO data with other petrophysical surveillance data that can be fed into reservoir models allowing for improved reservoir management and proactive planning of well work activities. The diagnostic workflows are used take a holistic approach by integrating reservoir and wells data, both static and dynamic. Use of transient events is part of the method, complementing steady-state data. We show that well behaviors are better understood with permanent installations and continuous data acquisition, while intervention-type surveillance helps calibrate the models, all leading towards a representative flow diagnostic of well condition. A retrospective view of more than two decades of flow-diagnostic examples in producer and injector wells is described in the paper, with key learnings on limitations, applicability and next developments is presented.

Dielectric logging has evolved from a single-frequency mandrel tool in the 1970s to a multifrequency, fully articulated pad tool in the 2000s. Dielectric dispersion, the frequency-dependent dielectric property of sedimentary rocks, provides an additional dimension to petrophysical evaluation over broad frequency up to about 1 GHz. However, the interpretation of dielectric dispersion can be particularly difficult in organic-shale reservoirs, often due to a variety of polarization mechanisms and considerable uncertainties caused by complex mineralogy and organic matter. In this paper, we present an integrated workflow including dielectric core analysis, processing of dielectric-dispersion logs, and petrophysical interpretation through core-log integration. We emphasize the need for accurate matrix-permittivity determination for all current interpretation methods, and explore the possibility to determine matrix permittivity directly from dielectric well logs. Dielectric core analysis is used to validate the interpretation model and calibrate dielectric well logs. For instance, matrix permittivity can be calibrated in the laboratory by optimizing the dielectric constant of each mineral and kerogen. This ensures that kerogen is lump-summed with the matrix for more accurate estimation of hydrocarbon volume. Multifrequency dielectric well-log data are then fitted with an appropriate mixing law or dispersion model to obtain petrophysical parameters, such as water-filled porosity, salinity, textural information, and flushed-zone resistivity. Inspired by the Pickett plot as a visual representation of the Archie equation, we propose a new graphical method that we call Complex-Domain Analysis (CDA) to solve dielectric-mixing-law equations without having to know matrix permittivity. This new method provides a simple way to determine a uniform matrix permittivity or matrix-permittivity endpoints, directly from dielectric log without a need for calculating it from mineralogy, thus very useful for quality control and dielectric interpretation immediately after logging. The integrated dielectric interpretation workflow and CDA method are demonstrated in two case studies in organic-shale reservoirs.