Tuesday, May 28, 2019

Modeling Logs for Horizontal Well Planning and Evaluation

Horizontal wells can increase production rates and ultimate recovery, and can reduce the number of platforms or wells required to develop a reservoir. They can also help avoid water or gas breakthrough, bypass environmentally sensitive areas and reduce stimulation costs.

As exploration and development budgets tighten, companies are becoming more efficient by drilling fewer, well-placed holes. Reentry and multilateral wells are growing in number, along with short-radius wells. There are greater expectations and smaller margins for error in driling today's horizontal wells. 

Drilling horizontal wells presents formidable challenges. Planning trajectories, choosing fluids, steering, formation evaluation and completion- each stage is a huge task. Several stages-planning,steering and formation evaluation- benefit from combining the efforts of geologist, log analyst and directional drillers. 

A powerful partner in all these stages is forward modeling, or log simulation. Other industries are using simulation to help train pilots , model aircraft and automobile reliability and response, design buildings, test weapons, record music, predict weather- the list is endless. In the oil field, modeling helps make efficient use of logs in horizontal wells in two ways- first by predicting logging-while-drilling (LWD) tool response to guide directional drilling, and second by constraining formation evaluation when the conventional assumptions of a vertical well no longer hold. 

Directional drilling practice and technology have evolved to the point where , given a good plan, the target can be hit with high accuracy. The drill bit can be placed within a target the volume of an engineer's office at a depth and lateral offset of a few miles. Trajectories are becoming more complex as directional drillers push the technology to its limits in "designer" wells. To improve the odds of these wells hitting the target, they are carefully planned in two steps: definition of the target from maps and logs, then design of a wellbore trajectory to hit it. 

No plan, unfortunately, is foolproof. Uncertainties in the position of the target , combined with unpreditictability of structural and stratigraphic variations, even in developed fields, can cause directional drillers to lose their way. The chance of going astray declines significantly, however, with the use of real-time formation evaluation logs and comparison of the logs with modeled cases to gauge the position of the tool within the sequence of beds. The INFORM Integrated Forward Modeling program provides an interface for building a formation model and simulating log response, allowing drillers to anticipate what's ahead. We look first at modeling for horizontal well planning, then explore how the INFORM system facilitates postdrilling visualization of LWD and wireline logs in horizontal wells.  

Model First, Then Drill

Often the objective of drilling a horizontal well is to penetrate the reservoir but stay close to a caprock shale or gas-oil contact- to drill parallel to a boundary or a contrast in material properties- for thousands of feet. Such a viewing angle is unusal for electromagnetic tools, the tools most commonly used for steering. Other measurements, such as gamma ray and density, are also affected by the horizontal geometry, giving an asymmetric response as they lie against the floor of the borehole. 

Because most resistivity tools probe several feet into the formation, they are affected by resistivity inhomogeneities in the vicinity of the well and even ahead of the drill bit. This early warning feature is beneficial to directional drillers, who harness it to steer wells into target layers or away from problem zones before they are encountered by the bit. This "proximity effect" can be accurately modeled during predrilling planning to provide a road map for drilling.

In a planning example from the North Sea, Jim White of Schlumberger Wireline & Testing in Aberdeen, Scotland, used log modeling to demonstrate the feasibility of landing the well in a thin sand and avoiding high-resistivity , calcite-cemented , tight streaks. Forward modeling computed the response of the CDR Compensated Dual Resistivity tool with its two depths of investigation- shallow from the phase shift measurement and deep from the attenuation log. When the wellbore came to within 3 ft of the calcite zone, the modeled attenuation and phase shift curves crossed, because the deeper-reading attenuation measurement senses the high-resistivity calcite. 

The CDR logs acquired when the well was drilled corroborated the modeled predictions. Based on the simulations, the signature of the lower boundary- the deep reading crossing over the shallow was recognized while drilling, and the well was steered away. Had the well entered the cemented zone, drillers estimated they would have spent several days trying to get back on target. 

Geologist from Chevron Niugini are using INFORM forward modeling to plan and geosteer horizontal wells in the Iagifu Hedinia field, within the Southern Highlands Province of Papua New Guinea. Located in Papuan Fold and Thrust belt, this field is part of a double anticline complex in the Hedinia thrust sheet. The major oil reservoir is the Lower Cretaceous Toro sandstone. Within the Toro, the hydrocarbon accumulation consists of an oil band up to 218 meters [ 715 ft] thick overlain by a gas cap. Gas cap expansion and gravity drainage are the major drive mechanisms for the field, with support from the Toro aquifer making a minor contribution.

Development well planning and drilling are complicated by the complex fold geometry. Unfortunately, the rugged karst topography created in the Darai Limestone at surface prohibits the acquisition of usable seismic data. For predicting the subsurface reservoir geometry, geologist rely on surface geological mapping, side-scan radar imagery, dipmeter data and correlation logs from adjacent wells.

In order to maximize productivity and ultimate recovery from the horizontal wells, wells are programmed to be horizontal in the Toro oil reservoir at a level of 15 m [50 ft] above the oil-water-contact. This enables the wells to produce oil at lower solution gas/oil ratios (GOR) and should delay breakthrough from the advancing gas front. 

During drilling to the Toro objective, the landing phase is critical to the success of the horizontal well program. With an unstable Alene shale section overlying the Toro, it is important to minimize the amount of horizontal section drilled before encountering the top Toro. Conversely, encountering the Toro during the build section of the well course, before reaching horizontal, can result in loss of productive interval since this hole section may be too close to the current gas-oil contact and would not be perforated. The Alene is drilled with mud weights in the range of 12 to 14 ppg, while the current reservoir pressure in the Toro are in the 4.5 to 5.5 ppg equivalent range. To prevent loss circulation problems and possible loss of the hole, it is necessary to identifiy the top of the Toro casing point before penetrating more than 1.5 to 3 m [ 5 to 10 ft] of the sandstone. 

An accurate predictive model of the Toro anticlinal geometry resulting from recognition of overlying stratigraphic markers while drilling -as well as the ability to determine the structural attitude of these layers - increases the probability for a successful landing phase. With INFORM processing, a model of stratigraphic interval above the target can be built using well logs and dipmeter data from nearby wells along with geological structure models developed for the planned horizontal well. LWD responses for the potential range of structural dips within a particular area of the anticlinal fold can be simulated. 

Sunday, May 12, 2019


When a wavefront hits a boundary at vertical incidence, the amount of compressional energy reflected and transmitted is dependent only on the contrast of acoustic impedance- density times compressional velocity- of the rocks at that boundary. But when the incident angle is not 0 degree, the amount of compressional energy reflected of tranmitted depends on the angle of incidence, or source offset, and contrast in densities and shear and compressional velocities. In such cases, the reflection AVo can be measured and analyzed to yield information about lithology and pore fluid through their effects on density and compressional and shear velocities. 

Carrying out a walkaway VSP with the receivers straddling such a boundary allows direct measurement of the variation in amplitude with offset that arises from lithology and fluid properties above and below the reflector. The results can be analyzed for fluid and lithology identification in a wide zone around the well. Formation properties inferred from VSPs can be integrated with those interpreted from well logs and measured directly from cores. In this way the VSP can also provide independent calibration of the same amplitude variation seen across a surface seismic reflection point gather- a gather is thee collection of traces that reflect at the same point, but at different angles, or offsets.

Calibrating the surface seismic AVO data with the VSP AVO response brings added value by:

  • establishing viability of using AVO to map a reservoir.
  • reducing the risk involved with the added cost of AVO studies
  • improving the reliability of AVO interpretations
  • quantitatively assessing the effects of processing on the AVO response.

To establish whether AVO is applicable as an interpretation tool for a particular reservoir, the expected AVO response is usually modeled. This requires knowledge of the model parameters, including shear velocity. Dipole shear sonic logging tools are used to measure shear velocities even where this velocity is slower than the borehole fluid velocity.

However, use of density and velocity log data to model anticipated AVO anomalies has not always succeeded in fully explaining the AVO response observed on surface seismic gathers. The reasons for this are many and include reflectivity mismatches between surface seismic and log data, wave propagation effects through fine layers, tuning effects (constructive and destructive interference at seismic wavelengths), geometric effects, processing-related issues and intrinsic anisotropy.

Borehole seismic data can quantify these effects. VSPs provide an independent measure of the seismic AVO response and the ability to include necessary effects in the forward modeling to satifactorily explain the origins of the surface seismic AVO response. Anisotropy is one such effect - one that can both mimic and mask AVO responses, giving false hope for or concealing the presence of hydrocarbons. 

Informaiton about anisotropic velocities for forward modeling often comes from measurements made on cores. But being scale-dependent, anisotropy may be different at the seismic wavelength scale. Therefore, it is better to measure  the elastic anisotropy at the seismic scale. 

In 1994, at Schlumberger Cambridge Research in Cambridge, England, Doug Miller proposed a method to do this using the arrival times from a walkaway survey to provide a measure of compressional velocity anisotropy in a shale, and from this to characterize the elastic properties of that shale, governing compressional and vertically polarized shear waves.

Shale consists of finely- layered clay platelets and exhibits an anisotropy called transverse isotropy (TI). The acoustic properties vary depending on whether waves propagate with particle motions parallel or perpendicular to the platelet layers- often thought of as horizontally or vertically because the clays usually lie flat.

Miller proposed that the vertical slowness - the inverse of velocity - of a shale may be measured across an array of geophones for each shot point offset along a walkaway profile. And the horizontal slowness can be measured at a single receiver location for adjacent shots in the same profile, providing the subsurface layers are essentially flat. A crossplot of these measurements for each shot position defines the compressional anisotropic response of the shale. A curve fitted to these data points provides a solution to the equations that deliver shear anisotropy through a complete description of the elastic properties of the shale.

These research efforts have been put to practical use in the BP-operated Forties field in the UK sector of the North Sea. The ultimate aim is to enable AVO attributes to be mapped with confidence from 3D surface seismic data. To achieve this,  a detailed evaluation of shear velocity anisotropy in the formations overlying the Forties sand has been undertaken to build a velocity model. The data used included acoustic measurements from preserved shale and sand cores, a full suite of logs- including standard density and DSI Dipole Shear Sonic Imager Logs- in addition to walkaway, rig source and vertical-incidence VSP data.

Initially, two models were generated, one assuming the shale overlying the reservoir sand was isotropic and another in which TI anisotropy was introduced. Differences in amplitude response between the two models were immediately observed, particulary at far offsets for the interface between the shale and the reservoir sand at 1.07 normal incidence time. 

The predicted response assuming an anisotropic shale was validated by the amplitude measured in the calibration walkaway. This implies that the effect of the anisotropic velocity in the shale must be taken into account before attributing the AVO response in the surface seismic data to effects of fluid in the reservoir. 

It is clear from this study that the combination of AVO measurement from VSP and log-based, anisotropic forward modeling provides a powerful methodology for calibrating AVO responses observed on surface seismic data near wells in low dip structures. Where AVO analysis is used as the basis for hydrocarbon indication in fields with existing wells, the method helps identify the origin of observed AVO effects, determining whether large-scale AVO analysis and reprocessing effort are worthwhile in terms of achieving the desired objectives. 

The greater understanding of observed AVO effects should minimize the risk of missing genuine hydrocarbon-related AVO anomalies or of misinterpreting anomalies caused by other factors, such as anisotropy.


Tuesday, April 9, 2019

Borehole Seismic Data

Seismic surveys in the borehole deliver a high-resolution quantitative measure of the seismic response of the surrounding reservoir. Although these measurements may be used alone to image local features, they may also be tied with well data-logs and cores- and then related to more extensive surface seismic data. Advances in borehole geophysics are helping realize the full potential of existing data to create a sharper image of the reservoir. 

It's a matter of resolution. Surface seismic surveys deliver  one of the few quantitative measurements of reservoir properties away from wells, making the technique central to structural mapping of the entire reservoir volume. However, surface seismic waves cannot resolve features smaller than 30 to 40 ft [9 to 12 m] . On the other hand, logs and cores resolve features on the scale of a few feet down to about 6 inches [15 cm]. Reconciling these two measurement scales to get the optimal picture of the reservoir volume is a problem that has long challenged the industry.

Borehole geophysics has a foot in both the logging and surface camps. From the vantage of the wellbore, seismic data often have higher resolution than their surface seismic counterparts. Depths of each borehole receiver are also known, providing a better tie to the formation properties provided by petrophysical, core and other in-situ measurements and relating them to the 3D seismic volume. 

The idea of locating a receiver downhole and a seismic source at surface is not new. For more than half a century, the check shot has helped to correlate time-based surface seismic surveys with depth-based logs. Check shots check the seismic travel time from a surface shot to receivers at selected depth intervals. Subtraction of times, combined with the depth differences, yields vertical interval velocities and thus relates well depths to surface seismic times. 

In vertical seismic profiles (VSPs), the spacing between downhole geophone levels is considerably closer than for check-shot surveys. VSPs use high-quality full waveforms that include reflection information rather than just the time of first arrivals - or first breaks- to create an image of reflections near the wellbore. Building on this technique, 2D reflection images have been obtained by offset and walkaway surveys with sources and receivers in a variety of configurations that address most reservoir problems.

Yet, despite these and other developments, borehole geophysics has for many years failed to gain the status in reservoir characterization that some industry specialists think it deserves. Now, thanks to improved quality and increased confidence in the match between borehole and surface seismic data, borehole geophysics seems to be moving into an increasingly valued position.

Before examining how borehole seismic data are being used to successfully integrate other data, this article will illustrate how the scope of VSP is broadening through the development of horizontal, 3D and through-tubing techniques.

Broadening the Scope of VSP Applications

In the deviated and horizontal wells of the North Sea,the most common type of borehole seismic survey is the vertical-incidence VSP. These are often called walk-above surveys because, as the geophone is moved along the deviated section of borehole, the source is kept vertically above it, "walking above" the well.  In VSP terms, a horizontal well is an extreme version of a deviated well. Like other VSPs, deviated well surveys may be used for locating the well in the 3D surface seismic volume and assessing the quality of surface seismic surveys. Also, the technique may be employed for measuring lateral velocity variations and for imaging faults and structures below the wellbore. 

The following example of a walk-above VSP was carried out in late 1994, in a North Sea well with a 1.2 kilometer horizontal section. There were two main objectives. The first was to measure a suspected lateral velocity anomaly that may have been creating artifacts in the surface seismic data. The second was to obtain a high-resolution seismic image below the deviated portion of the well. An additional objective was to obtain seismic image in the horizontal part of the well.

Data were collected in ther vertical and deviated portions of the cased well using the conventional wireline-conveyed ASI Array Seismic Imager tool. In the horizontal section, a two-element CSI Combinable Seismic Imager geophone array was run on drillpipe in combination with a cement bond log. By decoupling the sensor module from the body of the CSI tool, the geophones are isolated from noise and distortions created by the drillpipe. 

As with any survey, the desired seismic image is produced using the reflected, or upgoing, wavefield. So the first processing task was to separate downgoing waveforms from upgoing. For walk-above surveys in horizontal wells, this is far from straightforward, since unlike vertical and deviated wells, there is no apparent time difference across the array between the downgoing and the reflected upgoing waves. It is therefore impossible to use conventional techniques to distinguish between reflections and downgoing waves. To improve the image a number of special techniques were used, including:
  • multichannel filtering to attenuate noise and sharpen the desired signal
  • downgoing wavefield subtraction using a long filter length to estimate the downgoing wavefield
  • median filtering techniques to estimate and subtract the energy scattered by faults
  • enhancement of the desired upgoing signal
  • equalization of the reflected wavefield amplitudes from the horizontal and the build up sections.

The final image showed three important features: the two faults marked A and B, which appear where suspected in the reflected image, and the dip of the strata below the well. Formation MicroScanner data acquired during openhole logging were compared with the VSP, confirming the fault locations-seen as chevrons in the VSP - and the apparent dips.

In this case study, VSP processing was performed before Formation MicroScanner data were ready to interpret, and the VSP helped the interpretation by outlining the major features. The two data sets were then interpreted and refined together, providing a more complete description of near-well geology than was otherwise available. The results met the main objectives of the survey and delivered an image below the horizontal section. 

An alternative strategy for acquiring and processing horizontal VSP data exploits the different responses of geophones and hydrophones to differentiate downgoing energy from upgoing energy in horizontal wells. Geophones are clamped to the formation, and sense its motion. In contrast, hydrophones are suspended in the borehole fluid and are sensitive to fluid pressure changes as seismic wave passes in any direction. When the two sensor types show the same signal polarity for a downgoing wave, they show different polarities for the upgoing wave.  By taking the difference between signals received at the two types of sensors - for a signal consisting of a direct pulse followed by a reflected pulse- the direct wave is canceled and the reflection enhanced.

Complications arise from differences in the coupling and impulse responses between geophones and hydrophones. However, this approach has recently been applied in the field, enabling the extraction of related wavefields in a horizontal well and the imaging of reflectors below the receivers.

 3D VSPs

VSP imaging surveys, such as walkaways, have been used for a number of years to image structural complexity away from the borehole. These walkaway profiles are essentially two-dimensional, confined to the vertical plane containing the surface source and the borehole. 

Because of the proximity of the receivers to the target, like all VPSs, these 2D images usually have the advantage of being of higher resolution than their surface seismic counterparts. But, by definition, 2D walkaways don't describe the full volume of the reservoir. Fortunately, the acquisition principle may be extended to cover three dimensions by repeated profiling in parallel lines - in effect, by collecting a series of 2D walkaway surveys similar to marine 3D seismic data acquisition. 

The progression from 2D to 3D in VSP surveys is similar to the progression in the surface seismic technique , and offers equivalent benefits. Thus, 3D VSPs allow high resolution imaging to augment surface 3D surveys and make it possible to obtain  images beneath surface obstacles, such as platforms, and near-surface obstructions, such as shallow gas zones. In addition, because the acquisition conditions and processing steps of VSP surveys are accurately reproducible, 3D VSP opens up the possibility of time-lapse, or 4D, seismic surveying. 

However, progressing from 2D to 3D substantially increases the need for planning and logistics control. Similarly, the processing requirements are almost an order of magnitude greater. 

The first 3D VSP survey was run in 1987 in the Adriatic Sea Brenda field, operated by AGIP. Since then, there have been two 3D VSP surveys in the Norwegian Ekofisk field for Phillips Norway- where a large gas plume over the center of the structure prevents imaging using conventional 3D surface seismic techniques. Other Norwegian surveys probe the Eldfisk and Oseberg fields. 

In the UK North Sea, a 41-line, 3D walkaway VSP survey has been carried out in Shell Expro's Brent field. In this case, the aim was to acquire a survey with improved resolution compared with the 3D surface seismic survey. The image was then be used to produce an accurate structural map to aid the planning of horizontal development wells in the Brent slump- a crestal zone of complex faulting and collapse which contains a significant portion of the field's remaining oil reserves. 

The survey was executed from a well with a trajectory that allowed positioning the geophones to give three-dimensional illumination of the slump zone. The receivers consisted of five shuttles with fixed triaxial sensors, clamped 2000 ft [606 m] above the target during the entire survey. Once in the well but prior to shooting, the coupling between each of the shuttles and the formation was evaluated using internal shakers to ensure distortion-free data.

The seismic source consisted of a cluster of three 150 in 3 sleeve guns. To supply sufficient gas for 41 lines of 200 shots per line, four 5100 cubic meter nitrogen-filled tube skids were used. Simultaneously with the downhole data acquisition, each shot location on the surface was recorded using two differential GPS navigation system. 

To make the survey cost-effective, it was vital to minimize time spent acquiring data -every extra minute per sail line meant an additional 41 minutes of rig time. For example, to reduce the time the vessel took to maneuver between lines, a strategy was devised to wrap each line efficiently into the next. In the end, the data were acquired within the planned survey time of two and a half days, including a conventional VSP.

The 3D processing involves an extension of methods already developed for 2D walkaways-data preparation and navigation check, triaxial projection, wavefield separation, deconvolution and migration. 

In this case, the processing consisted of separate preparation and processing of all 41 lines up to the deconvolution stage. Then all 41 reflected energy profiles were accessed by the 3D VSP migration algorithms to place the reflections correctly in space.

The successful processing of these surveys required an experienced geophysicist with strong interpretative skills to make the correct decisions at each stage of the processing -for example, to ensure that all possible questions related to the influence of data quality had been resolved. These skills ensured that the image was interpreted in terms of reservoir structure without processing artifacts.

The migration process requires the computation of raypaths from each source and every receiver to every reflection point in the subsurface. The rays are traced through a velocity model of the subsurface that can vary in complexity between flat layers ( a 1D layercake) to complex structures in 2D or 3D.

For simple structures , a layercake velocity model, which reduces computation time, is sufficient.  However, using this model in more complex subsurface may lead to erroneous positioning of reflections and the incorrect focusing of real events. More complex velocity model increase the number of ray-trace computations required, but are better able to position reflected events and focus the wave energy.

The Brent structure varies in the dip direction but changes very little along strike. Consequently, the velocity model is more complex than a plain 2D model but not as complex as a full 3D model; the structure varies in one horizontal direction and is extruded into the other horizontal dimension to form a so-called "2.5D" model. In this, the volume may be thought of as filled with an infinite number of 2D sections. This allowed computational efficiency due to symmetry and ensured a close match with the actual Brent structure.

Shell concluded that the Brent 3D VSP improved vertical resolution and significantly improved horizontal resolution- resolving features on the order of 100 to 150 ft [ 30 to 45 m] as opposed to the original 3D surface seismic resolution of 200 to 300 ft [60 to 90 m] . The interpretation of the slump features has confirmed conclusions reached independently , demonstrating the technique's potential and reducing the risk of a proposed new 3D surface survey.

 Through-Tubing VSPs

The third application broadening the scope of borehole geophysics is the VSP through tubing. Thanks to hardware developments, cost-effective VSPs can be run in mature fields that promise significant economic benefits

Traditionally, borehole seismic surveys are acquired in exploration wells when they are drilled. However, in older fields, borehole seismic information is often needed to aid the reservoir engineer in areas where no new wells are planned, or to plan a new well. Now a slim seismic receiver may be deployed by a simple masted logging truck to acquire borehole seismic data through production tubing and inside casing during workover or while the well is still on production. This reduces acquisition costs and makes surveys in multiple wells possible during the same mobilization. 

In this way, a full range of borehole surveys may be carried out and the data may be used to tie log and production information to new 3D surface seismic surveys being run in older producing fields.

The slim seismic tool has a 1 11/16- inch outside diameter and may carry one single-axis geophone group or three orthogonally mounted accelerometers.The mechanicallytt actuated anchor has a maximum opening of 7 in. [17 cm] . The tool is adapted for operation with a monocable wireline and through-wellhead pressure fittings. This allows for operations in producing wells with surface pressure. As with any system, a range of seismic information may be obtained in vertical or deviated wells, from check shots to walkaway VSP images.

For example, an offest VSP survey was acquired through tubing and through casing in an abandoned wwell in an inland shallow water field in south Lousiana, USA, using a marine vibroseis unit as a source to acquire high-resolution data. The offset VSP survey was designed to confirm the location of a low-angle fault-indicated by logs-which could not be seen on the surface seismic images. The fault's orientation was needed to reduce the risk of an infill development well and was easily spotted using the offset VSP image.

Using Borehole Geophysics to Integrate Data

A the heart of developments to improve data integration is the recognition of the complementary nature of some measurements. Perhaps the best example of this is the relationship between sonic logs and seismic data. In these two measurements, the physical interaction with the reservoir is the same, but at a different scale of resolution. The sonic tool measures formation compressional slowness, which is dependent on many factors, including the formation porosity and lithology. 

Compressional slowness combined with density provides the one-dimensional acoustic impedance of the formation, the same property that underlies seismic reflections. 

But seismic waves are sensitive only to relative changes in acoustic impedance, unlike sonic slowness measurements, which sample absolute values. Therefore, acoustic impedances from logs provide sufficient information to model most, but not all features of the seismic response. The total travel time measured by sonic logs is a required contribution to the bulk response of the low-frequency surface seismic surveys. Then, synthetic seismograms may be constructed and the response of the formation simulated by altering parameters such as porosity, fluid type and lithology. The synthetics can be used to interpret real data.

Although the scope of VSPs is expanding, the wealth of information relating to lithology, fluid contacts and the seismic responses that they produce is not always used to its fullest extent. This is particularly true when it comes to evaluating and improving the information content of surface seismic data. Now, existing technologies are being used in new ways to provide additional direct quantitative mesurements of the seismic response of the reservoir adjacent to wells.

The next two examples clearly indicate how the integration of all available data may improve understanding of the reservoir. The first example looks at how structural and stratigraphic interpretations may be improved. The second shows how reflection amplitude variation with offset (AVO) from VSPs may be used to calibrate surface seismic AVO.

Morgan's Bluff

In the Morgan's Bluff field of Orange County, Texas, USA, the operator IP Petroleum needed to map the shale edge of its Hackberry reservoir to design a secondary reservoir program.

Substantial existing 2D surface seismic data did not adequately image the reservoir. Therefore, vertical incidence and offset VSPs were shot within a production well. These results were combined with logs and geologic information to map the edge of  the shale. Further, the surface seismic lines were reinterpreted, resulting in an extensive remapping of the Hackberry sand.

The aim was to drill a sidetrack from the shut-in producing Well 8 toward the adjacent Well 10, depending on the exact reservoir boundary, to be determined using the VSP - the Hackberry sand was originally mapped on the strike line that runs through both of these wells.

First , the feasibility of this plan was tested and detailed survey models were constructed using structure maps, log data from the two wells and velocity data from a third well. Borehole seismic data shot in 1986 in the central part of the field were used to construct the general velocity model. In Well 8, sonic logs were available to about 8000 ft , and only nuclear and resistivity logs from there to total depth. A pseudosonic log was constructed from these logs and compared to the velocities from the VSP survey. A synthetic offset VSP was then generated using the same wavefield separation, deconvolution and migration processing to be used with the real data. 

Two scenarios were forward modeled: a gradual shaling out and an abrupt, or faulted, sand termination. From this it was agreed that in either case the shale boundary should be interpretable to within 100 ft using the offset VSP sections, and the go-ahead for the survey was given. Additionally, a second offset VSP to the west of well 8 was designed to confirm the interpretation. A VSP was also to be carried out in Well 8 to build an updated velocity model for migration.

The three downhole surveys were acquired with sources located 4000 ft [1212 m] to the west-southwest, 4300 ft [1300 m] to the southwest and 400 ft [121 m] to the east-southeast. An eight level downhole receiver system was deployed to record 110 levels at 50 ft [15 m] spacing from 8500 to 3000 ft [2575 m to 909 m]. Across each interval, the top and bottom shuttles were overlapped to check for any source amplitude, signature or phase changes during the survey.

Following a standard processing sequence using a flat-layer velocity model and some small velocity changes to match the model to the observed transit times, each of the offset VSPs was migrated. Logs from Well 8 were correlated with the offset VSPs.

Monday, March 25, 2019

Nuclear Magnetic Resonance Imaging

Although well logging has made major advances over the last 70 years, several important reservoir properties are still not measured in a continous log. Among these are producibility, irreducible water saturation and residual oil saturation. Nuclear magnetic resonance (NMR) logging has long promised to measure these, yet it is only recently that technological developments backed up by sound research into the physics behind the measurements show signs of fulfilling that promise. 

 For nearly 70 years, the oil industry has relied on logging tools to reveal the properties of the subsurface. The arsenal of wireline measurements has grown to allow unprecedented understanding of hydrocarbon reservoirs, but problems persist: a continuous log of permeability remains elusive, pay zones are bypassed and oil is left in the ground. A reliable nuclear magnetic resonance (NMR) measurement may change all that. This article reviews the physics and interpretation of NMR techniques, and examines field examples where NMR logging has been successful. 

 Some Basics

Nuclear magnetic resonance refers to a physical principle- response of nuclei to a magnetic field. Many nuclei have a magnetic moment- they behave like spinning bar magnets. These spinning magnetic nuclei can interact with externally applied magnetic fields, producing measurable signals.

 For most elements the detected signals are small. However, hydrogen has a relatively large magnetic moment and is abundant in both water and hydrocarbon in the pore space of rock. By tuning NMR logging tools to the magnetic resonant frequency of hydrogen, the signal is maximized and can be measured. 

The quantities measured are signal amplitude and decay. NMR signal amplitude is proportional to the number of hydrogen nuclei present and is calibrated to give porosity, free from radioactive sources and free from lithology effects. However, the decay of the NMR signal during each mesurement cycle- called the relaxation time- generates the most excitement among the petrophysical community.

Relaxation times depend on pore sizes. For example, small pores shorten relaxation times- the shortest times corresponding to clay-bound and cappilary-bound water. Large pores allow long relaxation times and contain the most readily producible fluids. Therefore the distribution of relaxation times is a measure of the distribution of pore sizes- a new petrophysical parameter. Relaxation times and their distribution may be interpreted to give other petrophysical parameters such as permeability, producible porosity and irreducible water saturation. Other possible applications include capillary pressure curves, hydrocarbon identification and as an aid to facies analysis.

Two relaxation times and their distributions can be measured during an NMR experiment. Laboratory instruments usually measure longitudinal relaxation time, T1 and T2 distribution, while borehole instruments make the faster measurements of tranverse relaxation time, T2 and T2 distribution. In the rest of this article T2 will mean tranverse relaxation time. 

NMR Applications and Examples

The T2 distribution measured by the Schlumberger CMR Combinable Magnetic Resonance tool, described later, summarizes all the NMR measurements and has several petrophysical applications:
  • T2 distibution mimics pore size distribution in water-saturated rock
  • the area under the distribution curve equals CMR porosity
  • permeability is estimated from logarithmic-mean T2 and CMR porosity
  • empirically derived cutoffs separate the T2 distribution into areas equal to free-liquid porosity and irreducible water porosity.

Application and interpretation of NMR measurement rely on understanding the rock and fluid properties that cause relaxation. With this foundation of the mechanisms of relaxation, the interpretation of T2 distribution becomes straightforward.

T2 Distribution - in porous media, T2 relaxation time is proportional to pore size.  The observed T2 decay is the sum of T2 signalss from hydrogen protons, in many individual pores, relaxing indepedently. The T2 distribution graphically shows the volume of pore fluid associated with each value of T2, and therefore the volume associated with each pore.

Signal processing techniques are used to transform NMR signals into T2 distributions. Processing details are beyond the scope of this article.

In an example taken from a carbonate reservoir, T2 distributions from X340 ft to X405 ft are biased towards the high end of the distribution spectrum indicating large pores. Below X405 ft, the bias is towards the low end of the spectrum, indicating small pores. This not only provides a qualitative feel for which zones are likely to produce, but also helps geologists with facies analysis.

Lithology-independent porosity- Traditional calculations of porosity rely on borehole measurements of density and neutron porosity. Both measurements require environmental corrections and are influenced by lithology and formation fluid. The porosity derived is total porosity, which consists of producible fluids, capillary-bound water and clay-bound water. 

However, CMR porosity is not influenced by lithology and includes only producible fluids and capillary-bound water. This is because hydrogen in rock matrix and in clay-bound water has sufficiently short T2 relaxation times that the signal is lost during the dead time of the tool. 

An example in a clean carbonate formation compares CMR porosity with that derived from the density tool to show lithology independence. The lower half of the interval is predominantly limestone, and density porosity, assuming a limestone matrix , overlays CMR porosity. At X935 ft, the reservoir changes to dolomite and density porosity has to be adjusted to a dolomite matrix to overlay the CMR porosity. If the lithology is not known or if it is complex, CMR porosity gives the best solution. Also, no radioactive sources are used for the measurement, so there are no environmental concerns when logging in bad boreholes. 

Permeability -perhaps the most important feature of NMR logging is the ability to record a real-time permeability log. The potential benefits to oil companies are enormous. Log permeability measurements enable production rates to be predicted, allowing optimization of completion and stimulation programs while decreasing the cost of coring and testing.

Permeability is derived from empirical relationships between NMR porosity and mean values of T2 relaxation times. These relationships were developed from brine permeability measurements and NMR measurements made in laboratory on hundreds of different core samples. The following formula is commonly used:

A cored interval of a well was logged using the CMR tool. The value of C in the CMR permeability model was calculated from core permeability at several depths. After calibration CMR permeability was found to overlay all core permeability points over the whole interval. Over the zone XX41 m to XX49 m the porosity varied little. However, permeability varied considerably from a low of 0.07 md at XX48 m to a high of 10 md at XX43 m. CMR permeability also showed excellent vertical resolution and compared well to that of core values. The value of C used for this well will be applied to subsequent CMR logs in this formation enabling the oil company to reduce coring costs.

Free-fluid index - The value of free-fluid index is determined by applying a cutoff to the T2 relaxation curve. Values above the cutoff indicate large pores potentially capable of producing, and values below indicate small pores containing fluid that is trapped by capillary pressure, incapable of producing.  

 Many experiments have been made on rock samples to verify this assumption. T2 distributions were measured on water-saturated cores before and after they had been centrifuged in air to expel the producible water. The samples were centrifuged under 100 psi to simulate reservoir capillary pressure.  Before centrifuging, the relaxation distribution corresponds to all pore sizes. It seems logical to assume that during centrifuging the large pore spaces empty first. Not surprsingly, the long relaxation times disappeared from the T2 measurement. 

Observations of many sandstone samples showed that a cutoff time of 33 msec of T2 distribuitons would distinguish between free-fluid porosity and capillary-bound water.  For carbonates, relaxation times tend to be three times longer and a cutoff of 100 msec is used. However, both these values will vary if reservoir capillary pressure differs from the 100 psi used on the centrifuged samples. If this is the case, the experiments may be repeated to find cutoff times appropriate to the reservoir. 

In a fine-grained sandstone reservoir example, interpretation of conventional log data showed 70 to 80% water saturation across a shaly sandstone formation. However, on the CMR log most of the T2 distribution falls below the 33-msec cutoff indicating capillary-bound water. Interpretation including CMR data showed that most of the water was irreducible. The well has since been completed producing economic quantities of gas and oil with a low water cut. The water cut may be estimated from the difference between residual water saturation and water saturation from resistivity logs.

 In another example, but this time in a complex carbonate reservoir, the oil company was concerned about water coning during production. CMR log data showed low T2 values below X405 ft indicating small pore sizes. Applying the carbonate cutoff of 100 msec showed that nearly all the water was irreducible, which allowed additional perforation. To date no water coning has occured.

Values for cutoffs can also be tailored to particular reservoirs and help with facies analysis, as in the case of the Thamama group of formations in Abu Dhabi Oil Company Mubarraza field offshore Abu Dhabi, UAE. In this field, classical log interpretation showed water saturation of 10 to 60%. However, some zones produced no water, making completion decisions difficult. Permeability also varied widely even though porosity remained almost constant.  Laboratory measurements were performed on cores to determine whether NMR logging would improve log evaluation. 

 Cores showed a good deal of microporosity holding a large volume of capillary-bound water. Free-fluid porosity was found in the traditional way by centrifuging the water-saturated cores. For this reservoir, however, capillary pressure was known to be 25 psi, so the core samples were centrifuged accordingly. This showed that NMR measurements could provide a good estimate of nonproducing micropores using a T2 cutoff of 190 msec. In addition, permeable grainstone facies could be distinguished from lower-permeability packstones and mudstones with a cutoff of 225 msec.  

Additional Applications

Borehole NMR instruments are shallow-reading devices. In most cases, they measure formation properties in the flushed zone. This has some advantages as mud filtrate properties are well-known and can be measured at the wellsite on surface. When fluid loss during drilling is low, as in the case of low-permeability formations, hydrocarbons may also be present in the flushed zone. In these cases NMR tool may measure fluid properties such as viscosity and so distingish oil from water. 

A published example of the effects of hydrocarbon viscosity comes from Shell's North Belridge diatomite and Brown Shale formations, Bakersfield, California, USA. Both CMR logs and laboratory measurements on cores show two distinct peaks on the T2 distribution curves. The shorter peak, at about 10 msec, originates from water in contact with the diatom surface. The longer peak, at about 150 msec, originates from light oil.  The position of the oil peak correlates roughly with oil viscosity. The area under this peak provides an estimation of oil saturation.

 T2 distribution measurements were also made on crude oil samples having viscosities oof 2.7 cp to 4300 cp. Highly viscous oils have less mobile hydrogen protons and tend to relax quickly. The CMR log showed the T2 oil peak and correctly predicted oil viscosity. It also showed that the upper 150 ft of the diatomite formation undergoes a transition to heavier oil.

Capillary pressure curves, used by reservoir engineers to estimate the percentage of connate water, may also be predicted from T2 distributions. Typically these curves- plots of mercury volume versus pressure- are produced by injecting mercury into core samples. Under low pressure the mercury fills the largest pores and, as pressure increases, progressively smaller pores are filled. The derivate of the capillary pressure curve approximates the T2 distribution. Some differences in shape are expected as mercury injection measures pore throat sizes, whereas NMR measurements respond to the size of pore bodies.

 Other applications and techniques are likely to follow with more complex operations that might involve comparing logs run under different borehole conditions. For example, fluid may be injected into the formation that is designed to kill the water, so that residual oil saturation may be measured. 

Function of a Pulsed Magnetic

The CMR tool is the latest generation Schlumberger NMR tool. The measurement takes place entirely within the formation, eliminating the need to dope mud systems witth magnetite to kill the borehole signal- a big drawback with the old earth-field tools. It uses pulsed-NMR technology, which eliminates the effects of nonuniform static magnetic fields and also increases signal strength. This technology, along with the sidewall design, makes the tool only 14 ft long and readily combinable with other borehole logging tools. 

The skid-type sensor package , mounted on the side of the tool, contains two permanent magnets and a transmiter-receiver antenna. A bowspring ecentralizing arm or powered caliper arm- if run in combination with other logging tools- forces the skid against the borehole wall, effectively removing any upper limit to borehole size. 

An important advantage of the sidewall design is that the effect of conductive mud, which shorts out the antenna on mandrel-type tools, is greatly reduced. What little effect remains is fully corrected by an internal calibration signal. Another advantage is that calibration of NMR porosity is simplified and consist of placing a bottle of water against the skid to simulate 100% porosity. T2 properties of mud filtrate samples required for interpretations corrections- may also be measured at the wellsite in a similar fashion. Finally, the design enables high- resolution logging- a 6-inch long measurement aperture is provided by a focoused magnetic field and antenna.

Two permanent magnets generate the focused magnetic field, which is about 1000 times stronger than the Earth's magnetic field. The magnets are arranged so that the field converges to form a zone of constant strength about one inch inside the formation. NMR measurements take place in this region.  

 By design, the area between the skid and the measurement volume does not contribute to the NMR signal. Coupled with skid geometry, this provides sufficient immunity to the effects of mudcake and hole rugosity. The rugose hole effect is similar to that of other skid-type tools such as the Litho-Density tool.

The measurement sequence starts with a wait time of about 1.3 sec to allow for complete polarization of the hydrogen protons in the formation along the length of the skid. Then the antenna typically transmits a train of 600 magnetic pulses into the formation at 320-msec intervals. Each pulse induces an NMR signal-spin echo-from the aligned hydrogen protons. The antenna also acts as a receiver and records each spin echo amplitude. T2 distribution is derived from the decaying spin echo curve, sometimes called the relaxation curve.