SPE 37147

Three-PhaseSource

Bradley A. Roscoe,

Holdup Determination in Horizontal

SPE, Schlumberger-Doll Research

Co!Mght 16S6, %wiefy of Petroleum Engineers, Inc.

This paper was prepared for presentation al the 1996 SPE lntemafional Conference onHorizonta Well Technology hefd in Calgwy, Canade, l&20 Nmmks 19%.

Ths pepr was selected for pnssenfaticn by an SPE Program Commiftse following review ofinformation Qnfained in an abstract submitted by the author(s). Cmtents of the paper, aspfesenled. have not been reviewed by the Sockty of Petroleum Engineers and are subject tocorrection by the autho!fs). The material, as presented, does not necessarily raflect anyposition of the .%Mefy of Petroleum Engineers, its ofticers, or members Papers presented atSPE meetings are subject to publication review by Editorial Committees of the Socieiy ofPetroleum Engmoers. Permission to wpy is restriiled to an abstract of not more than 300words. Illustrations may not be copied. The abstract should contain conspicuousachnoMedgment of where and by whom the paper was presented. Write Librarian, SPE, PO.Box 633636, Richardson, TX 7S0S3.3636, U. S.A., fax 01-214.952-9435.

AbstractIn horizontal wells, it can be very difficult to interpretconventional production logging tools due to the fluidsegregation in the borehole. This is a even more of a problemwhen there are more than two phases present in the borehole,i.e., oil, water, and gas, A pulsed-neutron tool measures manyparameters which are differentially sensitive to all threepossible borehole phases. Therefore, it is possible to combinethe information available from a pulsed-neutron tool todetermine the 3-phase holdup in horizontal wells.

One of the major difficulties in evaluating the response ofa tool to 3-phase holdup is obtaining good data under realisticdownhole conditions, i.e., realistic gas densities. Laboratorymeasurements cannot readily be made under these conditions;therefore, modeling techniques must be used to evaluate andcharacterize tool response. To validate a computer model,laboratory data are needed for benchmarking; therefore, forthis study, over 400 laboratory formation measurements wereperformed using air to simulate gas. These formationconditions were also modeled using Monte Carlo techniques.The agreement between measured and modeled data proved tobe good enough that modeling can be used to confidentlypredict the tool response with air or realistic gas.

Once the ability to predict tool response under realisticdownhole conditions exists, it is possible to combineinformation from a pulsed-neutron tool to quantitativelydetermine the holdup of all three phases. This is accomplishedby combining the inelastic near/far ratio with the near and farcarbon/oxygen (C/0) ratio. This approach to the holdupmeasurement has been demonstrated using a combination oflaboratory data, Monte Carlo modeling, and field data. Theresults of this study have demonstrated that the RMS accuracyof this measurement is about 60/0on each of the three phases.

&iltiBm

Wells Using a Pulsed-Neutron

IntroductionAs horizontal wells have become more prevaient, the ability toreliably evaluate the production performance of these wellshas become increasingly important. Existing productionlogging techniques, such as spinners, that have beensuccessfidly used in vertical wells cannot always be applied tohorizontal wells with full confidence because of segregatedflow in the borehole. For this reason, new techniques-must bedeveloped to evaluate oil and water flow rates in horizontalwells.

To determine the flow rates of the oil and water phases in ahorizontal well, one must either 1) measure the individual oiland water flow rates directly, or 2) measure the individual oiland water velocities in addition to their hoidups, (It should benoted, that for most production logging applications inhorizontal wells, measuring only the holdup or only thevelocitv of the rsroduction fluids is usuallv insufficient todeterm~ne the so&ce of production problems.~ This paper willaddress part of the second approach, the measurement ofindividual oil, water, and gas holdups. Once determined, theseholdups can be combined with velocity information, obtainedfrom several possible approaches’”z to obtain oil and waterflow rates.

BackgroundPulsed-neutron tools have previously been used toqualitatively determine the 3-phase holdup in horizontal welis.This approach uses the borehole sigma and the inelasticnear/far ratio for this determination. The method is consideredqualitative since tool calibration information is not available~or the ratio measurement or the sigma of the gas.

Recent work reported by Peeters et. al.’ has attempted toquantify the pulsed-neutron measurement for holdup inhorizontal wells. Their approach utilizes three measurementsfrom a single pulsed-neutron tool centered in the borehoie:C/O windows ratio, borehoie sigma, and capture near/far ratio.The measurements are combined through a iinear responsematrix to produce the desired holdup measurements. Thecoefficients for the matrix are determined by regression ofmodeled or measured tool responses to known conditions.

A more quantitative approach has been employed with theRST* Reservoir Saturation TO01.4’5 This tool was primarily

* Mark of Schlumberger

895

2 B.A. ROSCM SPE 37147

designed to measure the oil saturation of the formation withoutdepending on formation water salinity. This wasaccomplished by using a carbon/oxygen measurement. Alarge part of the problem of converting a C/O measurementinto oil saturation is the effect of the borehole on themeasurement. To properly determine the formation oilsaturation, the borehole composition must be knownreasonably well. In general, a 5% error on the boreholecomposition can cause a 15 S.U. error on the formation oilsaturation. For this reason, the RST tool was designed withtwo detectors to compensate for the borehole effect. However,from previous statements, it is obvious that the RST tool canalso measure the 2-phase borehole holdup in a well.

In addition to the obvious 2-phase holdup measurement,other data (in several forms), already available from the RSTtool, are sensitive to the presence of gas in the borehole.Quantifying this information and combining it with thealready available C/O measurement provides sufficientinformation to quantitatively determine the 3-phase holdup inthe borehole.

To ensure the best quantitative answer, it is advantageousfor the parameters measured by the tool to have approximatelythe same depth of investigation. In addition, these parametersneed to be easily quantifiable. For this reason, the followingtool measurements were chosen for this measurement: nearcarbordoxygen ratio, far carbonloxygen ratio, and the inelasticnear/far ratio. Since all these measurements are inelasticmeasurements, their depth of investigations are about equaland do not vary with formation and borehole sigma. Otherparameters sensitive to borehole holdup, such as boreholesigma and capture ratios, can be used, but their varying depthof investigation can complicate their quantitative use for thismeasurement. In addition, their use adds additionalrequirements in the form of assumptions about the boreholesalinity.

One of the biggest problems in characterizing the toolresponse for a 3-phase holdup measurement is the acquisitionof realistic data due to the difficulty and hazards of performinglaboratory measurements with gas under realistic downholeconditions. Air is usually used to simulate the effects of gasfor these measurements; however, real downhole gas hasappreciable density and elemental constituents that cannot beignored. Therefore, tool characterization depends heavily ontool response modeling. For this reason, this study uses amultistep process to develop the quantitative analysis required.These steps include:

1. Performing benchmarked laboratory measurementssimulating 3-phase holdup conditions in horizontal wellsusing air to simulate gas.

2. Modeling the tool response using Monte Carlo techniquesunder the conditions measured in the laboratory tobenchmark the model,

3. Develop interpretation procedures that quantitativelypredict tool response using air to simulate gas.

4. Model the tool response to 3-phase holdup using realisticdownhole gas characteristics.

5. Apply interpretation procedure to modeled data.6. Apply interpretation procedure to acquired field data.

Finally, the issue of running the tool centralized or

eccentralized in the borehole must be addressed. In ahorizontal well, it can be difficult to centralize the tool in theborehole. An argument can be made that it is better toeccenter the tool and really be sure of its location in theborehole. However, eccentering the tool can introduce someproblems such as a nonlinearity of the borehole tool response.Therefore, this study has also tried to quantify the effects oftool position in the borehole.

Laboratory MeasurementsMeasurements to characterize the RST-A response underconditions simulating a horizontal well with 3-phase holdupwere performed at the Schlumberger Environmental EffectsCalibration Facility (EECF) in Houston, Texas. Since theseformations are usually used to simulate vertical conditions,borehole liners were fabricated to divide the borehole fluidinto three equal volumes. This simulated the effect ofsegregated fluids in the borehole. Different liners weredesigned for the centered and eccentered measurements toallow for the fluid displacement by the tool and therequirement of a rathole for the tool (Fig. 1).

Dividing the borehole into three equal volumes allows 10possible borehole fluid combinations that simulate holdup in ahorizontal well. These are shown in Table 1. Since it is notpossible to have a realistic downhole gas present in theborehole for these laboratory measurements, air was used tosimulate gas.

A series of measurements was performed using variousformations at the EECF. The formations were selected toprovide variations in lithology, porosity, borehole diameter,and formation oil saturation (see Table 2). Most of theseformations were run with the tool both eccentered andcentered in the borehole. All of the measurements used a 7-inch, 23-lb/ft casing. Fresh water was used in the formationand borehole for all measurements. Some additionalmeasurements were performed with 100 kppm brine in theborehole. Over 400 laboratory measurements were performedfor this study.

Monte Carlo Model of RST-A ResponseSince laboratory measurements with realistic gas conditionsare not possible in the laboratory formations, characterizationof the tool response depends heavily on tool responsemodeling. For this reason, a large part of the effort in thisstudy has been to develop and test Monte Carlo models topredict tool response. To this end, a RST-A model for theinelastic measurement has been developed using the MonteCarlo modeling code MCBEND,’ which was developed byAEA Technology in Winfrith, UK. This model requires about18 hours to mn on a SUN Spare 20 for each set of formationand borehole conditions. The outputs from the model aretallied such that the following tool responses are available:near and far carbon and oxygen yields, near and farcarbon/oxygen ratio, near and far carbon/oxygen windowratio, and near/far inelastic countrate ratio.

To benchmark the RST-A model (as applied to 3-phaseholdup problems), the laboratory measurements describedabove were modeled with MCBEND. The results from themodeling were then compared with the results obtained with

898

SPE 37947 Three-Phase Holdup Determination in Horizontal Wells Using a Pulsed-Neutron Source 3

the measurements as shown in Fig. 2. These figures show thatthe modeling can be used to predict tool response. Each figurehas a line fit to the data that can be used as a calibration toconvert from modeled results to measured results. Bothcentered and eccentered data are shown on the plot.

Borehole Gas HoldupThe inelastic near/far countrate ratio can be used to detect andquantify gas in the borehole. This is demonstrated in Fig. 3for centered and eccentered tool positions with measured datain which air was used to simulate the presence of gas. In thisfigure, the lines on the plot show the data trends for variousfractions of air in the borehole.

The effect of porosity on the inelastic near/far ratio is smallby comparison with its response to air in the borehole. Thedata points along each air fraction line lie high or low aroundthe line depending on the “total hydrogen index” of the fluidsin the borehole. For example, the 660/0 air line has a lowernear/far ratio for oil in the borehole than it does for water inthe borehole.

As far as sensitivity to 100% air in the borehole, theinelastic near/far ratio changes by about the same amountwhether the tool is centered or eccentered. However, for airholdups ranging from O to qs~., the centered tool hassubstantially more sensitivity to the air because of itsproximity to the air region. For similar reasons, theeccentered tool has more sensitivity than the centered toolwhen the air holdup is between 66 and 100%.

Modeling was used to predict the RST-A inelastic near/farratio under the same measurement conditions as those in Fig.3. This was used to benchmark the modeling, and acorrelation was developed so that modeled results could betransformed into equivalent measured results (see Fig. 2).This means that the calibrated modeled response gives thesame results as the measurements.

After the benchmarking of data of Fig. 3 via Fig. 2, theformations were remodeled substituting realistic downhole gasinto the model. The gas was modeled as 0.3 g/ml methane(CH4). The modeled inelastic near/far ratio under theseconditions is shown in Fig. 4 for the tool centered andeccentered, For this “realistic” gas, the dynamic range of theinelastic near/far ratio is reduced by about 30 to AOO/o relativeto the air data for both tool positions. Like the air data, the gasholdup lines drawn on the figure show the same general trendin that the centered tool has its maximum sensitivity betweenO and qq~. holdup while the eccentered tool has its maximumsensitivity between 67 and 100°/0gas.

Two-Phase (Oil/VVater) Holdup From C/OAs in the gas phase measurement discussed above, thepositioning of the tool in a segregated borehole impacts thelinearity and sensitivity of the holdup measurement. Thissection will address this issue with respect to the 2-phase,oil/water, situation with a near/far C/O interpretationapproach.

For this section, the interpretation approach utilized will bethe standard RST approach’ in which carbon and oxygenelemental yields can be expressed as sums of the contributionsdue to the matrix, pore space, and borehole as:

Cn = N, +N2cDS0 +N3Y0 .....!........................................(l)

0n=N4+N,0(l -So)+ N,(l-Yo) ...........................(2)

where the N’s are the near detector sensitivity factors ofcarbon and oxygen to the matrix, pore space, and borehole.(Similar equations can be written for the far detector where thefar detector sensitivity factors are designated by F’s.) Todetermine the sensitivity factors, four laboratorymeasurements (or calculations) are performed where only theformation or borehole fluids are changed (i.e., formationporosity, lithology, borehole size, and casing size are fixed).These measurements are usually referred to as the endpointmeasurements since they correspond to the endpoints of thetool response. From these data, the near and far sensitivityfactors are determined using the four eIemental yields from themeasurements with the above equations. This results in fourequations and three unknowns for both the carbon and oxygenexpressions, allowing the determination of the three sensitivityfactors for each elemental yield. To characterize the completetool response, this process is repeated varying porosity,lithology, etc.

Once the tool response is known from the abovecalibration, for every depth, the sensitivity factors arecalculated based on known borehole size, lithology, porosity,etc. and the near and far C/O ratios are measured. Combiningthese data results in the following two equations:

c N, +N2cPS0 +N3Y0

‘n ‘~= N, +N,4(1-SO)+N,(I- YO) ““”””””””-’’’”””(3)

R,=:=F1+ FIOSO + F3Y0

F4+F,cD(1-SO)+ F,(l-YO)................

r,....(4)

which have two unknowns. The solution to these equationsyields the borehole holdup and the volume of formation oil.

The above analysis assumes that the borehole fluid is ahomogeneous mixture of oil and water as is the case withvertical wells. However, in horizontal wells, the segregatedborehole fluid can introduce nonlinearities into the analysis.The purpose of the following analysis is to characterize theeffect of this nonlinearity on the determination of holdup in ahorizontal well. The data used for this analysis will includeeight points per formation porosity, i.e., oil and water in theformation with four different oil/water combinations in theborehole depicting several holdups with segregated flow (seeTable 3). The data points shown will be from modeling (themeasured data show the same result.)

For this analysis, the coefficients for the interpretationmodel are calculated from the four endpoint measurements ofthe formation being considered as described above. Thesecoefficients are then applied to the other four data points of thedata set to look at the linearity effects with segregatedborehole fluid. (This approach solves for both the oilsaturation and holdup simultaneously). The results of thisanalysis can be seen in Fig. 5 for a centered and eccenteredtool .

The top plot of Fig. 5 shows the reconstructed oil holdupfor data modeled in a 7-inch casing (8.5- and 10-inch

897

4 B.A. ROSIXX3 SPE 37147

borehole). The reconstructed holdup is extremely nonlinearif the tool is eccentered. In this case, it appears that the tooldoes not see the top third of the borehole due to thenonlinearity. If the tool is centered, then the tool response isfairly linear.

Since the effect being observed is a nonlinearity caused bythe tool’s limited depth of investigation, the effect should bereduced in a smaller casing size. The bottom plot of Fig. 5shows modeled results for a 5.5-inch casing in a 8.5-inchborehole. As can be seen, the effect is reduced, but still asignificant issue.

Table 4 summarizes the results of this 2-phase holdupanalysis using C/O ratios, This table includes RMS accuracyestimates and 18-second precision for this measurement.From these data, it is obvious that a centered tool always givesbetter accuracy and precision than an eccentered tool.However, if the tool must be run eccentered, accuracy can beimproved by making corrections based on calibrations of toolresponse as shown in the plots of Fig. 5. The effect of thisapproach is reduced sensitivity to low oil holdups.

Another interesting result of this analysis is that formeasurements using the RST-A tool, where the borehole fluidis unknown, a centralized tool will give better precision onformation oil saturation than an eccentered tool (see Table 5).This is because centralizing the tool makes the near and farcarbonloxygen response more orthogonal with respect toborehole compensation. For some conditions, this can meanreducing station times by a factor of 4. For measurements inwhich the borehole fluid is known, station time requirementsshow no improvement.

3-Phase Holdup DeterminationTo determine the holdup of all three phases using the near andfar C/O ratios and the inelastic near/far ratio, a two-stepinterpretation is currently used (these could be combined intoa single step). The interpretation is based on the assumptionthat the three borehole holdups sum to unity and thatformation gas is zero, i.e., YO+YW+Y =1 and SO+SW=l.

kThe first step is to obtain the gas oldup from the inelasticnear/far (N/F) ratio. This is obtained from calibration datasimilar in form to those shown in Figs. 3 or 4 and can beexpressed as:

Yg = f(y~) ........................................................................ (5)

Once the gas holdup is determined, the water and oil holdupscan be determined using a modified version of the normalRST C/O interpretation, The modifications made are to allowfor the inclusion of gas holdup into the formulation as shownbelow:

(6)~ = N1 +N20S0 +N3(Y0 +Y,~yo)

“ N, +N,O(l-SO)+N,(l-YO -Y,) ““”’””””””””””””””””

F, +F,OSO +F3(Y0 + Y,(p~m))R,= F4+F,@(l_so)+Ff(l_yo _y,) . . .. . . . . . ..- (7)

where ps and PO are the downhole gas and oil densities. In

these equations, the numerator and denominator have beenmodified from the original RST interpretation (Eqs. 3 and 4)to allow for the contribution of gas in the borehole. In thedenominator, this results in a reduction of the amount ofoxygen present in the borehole. In the numerator, it isassumed that gas will increase the carbon response by a factorapproximated by the relative densities of gas and oil.

As with the normal C/O processing, the sensitivity factorsare obtained from laboratory calibrations which use theborehole size, casing size and weight, lithology, and porosityas inputs. Once the C/O ratios are measured and the gasholdup is calculated horn Eq. 5, Eqs. 6 and 7 reduce to two

equations and two unknowns that can then be solved for theholdup and oil saturation.

In this formulation of the problem, it is assumed that thetool is equally sensitive to all sections of the borehole. As hasbeen shown, this is a reasonable assumption if the tool iscentered; however, if the tool is eccentered, this formulationwill introduce some bias. Under many conditions, this biascan be reduced by providing a correction based on data similarto those in Fig. 5.

3-Phase Holdup Results - Centered ToolThe interpretation procedure outlined above has been appliedto the measured and modeled 3-phase data obtained in thisstudy to evaluate the expected performance of thismeasurement. This includes the measured data (with air) andthe modeled data (with air and 0.3 g/ml gas).

An example of the measurement parameters for a 16-p.u.oil-saturated limestone formation with an 8.5-inch boreholeand 7-inch, 23-lb/ft casing is shown in Fig. 6 in a three-dimensional plot. In this plot, the comers of the plotsrepresent the conditions when only a single phase is present inthe borehole, i.e., YW=1 is all water, YO=1 is all oil, and Ys=lis all gas (0.3 g/ml in this example). From one comer toanother, the fraction of each phase changes linearly along thatline. The point in the center of the plot has water, oil, and gaspresent in equal fractions. The ‘z’ axis of the plot shows themagnitude of the various parameters being displayed.

The data from Fig. 6 are obtained from modeling since arealistic gas was used (0.3 g/ml). As can be observed, themodeled data are dominated by the physics of themeasurement, while the statistics of the Monte Carlocalculations are small by comparison. This makes identifyingthe tool response trends fairly straightforward. Also note thatthe inelastic near/far ratio response is quite orthogonal fromthe other responses, giving a clean gas signal. The near andfar C/O ratios show a similar response; however, they areorthogonal enough to differentiate the borehole and formationsignals.

When the holdup data of this study are analyzed throughthe above interpretation model, one can compare the predictedholdups to the known holdups to give an estimate of theaccuracy of the approach. An example of these errors inreconstruction is shown in Fig. 7 for the input data of Fig. 6.These data show that the reconstruction is fairly good. Theoverall accuracy of this approach can be estimated by takingRMS errors of the data. This was calculated using data forseveral formations with different borehole size, porosity, and

898

SPE 37147 Three-Phase Holdup Determination in Horizontal Wells Using a Pulsed-Neutron Source 5

saturation. This resulted in RMS errors of 6.3, 6.1, and 4.8°/0for the gas, oil, and water holdups, respectively.

While the holdup calculations are performed, it is possibleto estimate the precision of the holdup measurement bypropagating the measurement errors through the analysis.This was done assuming 18 seconds of data accumulation forthe measurement, which is equivalent to logging at 500 ft/hrusing a 5-level depth average. For these conditions, the I-aprecision were estimated to be 1, 15, and 15% for gas, oil,and water holdup, respectively. These precision will varydepending on the formation porosity, borehole size, and casingsize.

For an unambiguous interpretation, it is recommended thatholdup measurements in horizontal wells be performed inconjunction with velocity measurements. Since the velocitymeasurements are stationary measurements, a holdupmeasurement could be performed at the same time as thevelocity measurement, giving more than adequate precision tothe measurement.

Centralization Error Effects on 3-Phase HoldupAs stated previously, one of the advantages of eccentralizingthe tool is that its position is well known, while attempting tocentralize the tool can introduce tool positioning uncertainty.To evaluate the effect of errors in tool centralization, a seriesof Monte Carlo calculations were performed in which toolposition was varied from fully centralized to fullyeccentralized. These calculations were performed forformations with an 8.5-inch borehole in both 16 and 33 p.u.sandstone formations. The calculations were performed withtwo different casing sizes: a 7-inch, 23-lb/ft and a 5.5-inch,17-lb/ft casing.

The data were analyzed to show what would happen to thecalculated borehole holdups if the tool was assumed to becentralized, even when it was not. This was accomplished bycalculating the near and far detector sensitivity coefficientsbased on data with the tool centralized. In addition, the toolgas response (Eq. 5) was based solely on the centralized tooldata. These coefficients were then used with the modeled tooldata for all tool positions to predict the holdups one wouldcalculate with this centralization error.

Figure 8 shows the RMS error on the borehole holdups asa function of centralization error for the two casing sizesmodeled. The RMS error calculation includes data at bothporosities and with both water and oil in the formation. In thefigure, the RMS error at zero centralization error is abackground error level due to the nonlinearity of the responseand statistics of the modeling. For both casing sizes, itappears that there is no significant increase in the RMS errorup to about 0.5 inches. Based on this analysis, as long as thetool is within 0.5 inches of being centered, there is negligibleeffect on holdup accuracy.

Field ExampleCommissioned in 1993 as the first well in the offshore sectionof this reservoir (18 p.u. sandstone), the well has been oncontinuous production with interruptions only for workovers.The horizontal section was drilled with a 8.5-inch bit andcompleted with a 5,5-inch, 17-lb/ft casing and cement. The

well inclination across the reservoir section varies from 75 to88.5 degrees. It penetrates the original OWC, enablingmovement of the OWC to be monitored. The objectives of theproduction logging program were to 1) determine the flowingproduction profile, 2) determine the source of the waterproduction, and 3) determine any movement of the OWC.

Both velocity measurements and holdup measurementswere performed with the RST tool in this well. Unfortunatelyfor this study, the well did not have any gas in the borehole (asverified by the inelastic data). However, it is an excellentexample of 2-phase oil/water production in a horizontal welland is still a good example of what can be accomplished in ahorizontal well.

The RST tool was run eccentered in the borehole for thiswell. As mentioned earlier, if not corrected, this results insome bias in the final holdup estimates. For this example, thebiases were removed by a calibration similar to that shown inFig. 5 (derived from modeling). The RST holdupinterpretation was based solely on modeling results.

The logging results are shown in Fig. 9. The top graph inthe figure shows the comparison of PVL* Phase Velocity Logand WFL* Water Flow Log water velocities. The agreementbetween the two techniques is excellent. The next graph downshows the oil velocity measured with the PVL. The followinggraph shows a comparison of water holdup measurementsfrom the RST-A tool and the LIFT” Local ImpedanceFlowmeter Tool. (The LIFT tool measures the water holdupin a wellbore by scanning across the wellbore with sixseparate impedance probes which individually measure thelocal water holdup.g In both conventional and horizontalwells, the data from these six probes can be processed to givea global water holdup measurement.) The agreement betweenthe LIFT and RST-A holdups is excellent except where knownproblems occurred with the LIFT measurement. (Note thatthe uncalibrated LIFT points were due to operational problemsand the over-ranged point was past the maximum LIFTsensitivity). The bottom graph in the figure shows the oil andwater flow rates calculated from the velocities and holdup.These data indicate that most of the water production is frombelow 900 ft while the oil production is fairly uniform over theinterval.

This log shows an example of when the holdup answer, byitself, in a horizontal well, does not completely diagnose theproduction problem and why additional velocity information isusually required for an accurate interpretation. Looking atonly the holdup data from this well, a rather uniformlydecreasing water holdup going up the well is observed. This,in itself, does not readily identify that most of the waterproduction is coming from the lower part of the well.

Summary and ConclusionsThe ability of the RST-A tool to measure 3-phase holdup inhorizontal wells has been quantified in terms of accuracy andprecision of the measurement. This has been accomplished bymerging experimental (using air to simulate gas) with modeleddata (using 0.3 g/ml CH4 for gas) into a database with over400 borehole/formation conditions. This approach hasallowed for the benchmarking of the Monte Carlo model (forthe inelastic measurement) and predictions of the tool

899

6 B.A. ROSCOS SPE 37747

response with gas under realistic downhole conditions.The inelastic neadfar ratio was shown to be an excellent

stand-alone indicator of gas holdup. This measurement ismore accurate with the tool centralized in the borehole thanwhen it is eccentralized due to nonlinearities caused by toolpositioning.

The 2-phase holdup (oil/water) measurement wasevaluated using the near and far carbon/oxygen ratios for theeffects of tool positioning. This analysis was the normal dualdetector C/O analysis but used to evaluate the nonlinearities inthe measurement due to segregated flow. These resultsindicated that the tool response could be nonlinear if the toolwas run eccentered with segregated flow, but that it could alsobe corrected over a large part of the dynamic range withproper calibration. It was also shown that running the toolcentralized in segregated flow does not exhibit this nonlinearbehavior.

A filly integrated 3-phase holdup analysis was evaluatedusing the inelastic near/far countrate ratio and the near and farcarbou/oxygen ratios. This analysis was performed using aRST analysis modified to take into account the gascontribution. The analysis used both measured and modeleddata in combination for a centralized tool. For this analysis,the estimated accuracies are approximately 6 p.u. for each ofthe three phases, The estimated precision (18 seconds ofdata) for this measurement are approximately 1, 15, and 15°4for the gas, water, and oil holdups, respectively.

A field example indicated the accuracy and precision of the3-phase holdup measurement are in line with the predictionsfrom the modeling. Comparison of the holdups obtained withthe RST tool compared quite favorably with holdupmeasurements performed by another totally independenttechnique, establishing the validity of both measurements. Inaddition, this field test example demonstrated that modelingresults could be used to interpret log data accurately.

Finally, this field example reaffirmed the notion thatvelocity or holdup measurements by themselves will notalways give a correct interpretation of production problems;but that holdup and velocity measurements together can.

AcknowledgmentsThe author would like to thank British Petroleum forpermission to publish their data and Steve Bamforth for hisassistance in obtaining this permission. The author would alsolike to thank Norman Winkelmann of Schlumberger HoustonProduct Center for his assistance in acquiring the 3-phaseholdup measurements under laboratory conditions.

Nomenclature@= Porosity

BOPD = Barrels Oil Per DayBPD = Barrels Per Day

BWPD = Barrels Water Per DayC“ = Near Carbon YieldCr = Far Carbon Yield

C/O = Ratio of Carbon to OxygenF,= Far Detector Sensitivity FactorsN,= Near Detector Sensitivity Factors

N/F = Inelastic Near to Far Ratio

on=of=

owc ‘

R“ =R~=

RMs =so=Sw=

Y.=Yg =Y.=Y.=

Near Oxygen YieldFar Oxygen YieldOil Water ContactNear C/O ValueFar C/O ValueRoot-Mean-SquareFormation Oil SaturationFormation Water SaturationAir HoldupGas HoldupOil HoldupWater Holdup

S1 Metric Conversion FactorsBPD X 1.589873 E-O] = m’/d

fiX 3.048 E-01 =mR/h x 8.466667 E-05 = lllk

ft./rein x 5.08 E-03 = 111/S

in x 2.54 E+OO = cmlb/ft X 1.488164 E+OO = kg/m

References

1.

2.

3.

4.

5.

6.

7.

8.

McKeon, D.C., Scott, H.D., and Patton, G.L.: “Interpretation ofOxygen Activation Logs for Detecting Water Flow inProduction and Injection Wells,” Trans. SPWLA 32nd AnnuafLogging Symposium, Midland, Texas, (June 16-19, 1991), paperBB.Roscoe, B. A., Lenn, C., Jones, T. G. J., and Whittaker, C.:“Measurement of the Oil and Water Flow Rates in a HorizontalWell using Chemical Markers and a Pulsed-Neutron Tool,”Paper SPE 36563 presented at the 1996 SPE Annual TechnicalConference and Exhibition, Denver, Colorado, (October 6-9).Peeters, M., Oliver, D., and Wright, G.: “Pulsed Neutron ToolsApplied to Three-Phase Production-Logging in HorizontalWells,” Trans. SPWLA 35th Annual Logging Symposium, Tulsa,Oklahoma, (June 19-22, 1994) Paper L.Scott, H.D., Stoner, C., Roscoe, B. A., Plasek, R. E., andAdolph: , R.A. : “A New Compensated Through-TubingCarbon/Oxygen Tool For Use In Flowing Wells,” Trans.SPWLA 32nd Annual Logging Symposium, Midland, Texas,(June 16-19, 1991), Paper MM.Stoner, C., Scott, H.D., Plasek, R. E., Lucas, A. J., and Adolph,R.A.: “Field Tests of a Slim Carbon/Oxygen Tool for ReservoirSaturation Monitoring,” Paper SPE 25375 presented at the SPEAsia Pacific Oil & Gas Conference & Exhibition, Singapore,(Feb. 8-10, 1993).MCBEND: A Monte Cario Program For General RadiationTransport Soiuiions - Users Guide, AEA Technology, Wirrfrith,U.K., 1995.Roscoe, B. A., Stoner, C., Adolph, R.E., Boutemy, Y.,Cheeseborough, J., Hall, J., McKeon, D., Pittmsn, D., Seeman,B., and Thomas, S.: “A New Through-Tubing Oil-SaturationMeasurement System,” Paper SPE 21413 presented at theMiddle East Oil Show & Conference, Bahrain, (Nov. 16-19,1991).Halford, F.R., MacKay, S., Bamett, S., and Petler, J.S.: “AProduction Logging Measurement of Distributed Local PhaseHoldup,” Paper SPE 35556 presented at the EuropeanProduction Operations Conference and Exhibition, Stavanger,Norway, (April 16-17, 1996).

900

SPE 37147 Three-Phase Holdup Determination in Horizontal Wells Using a Pulsed-Neutron Source 7

Tablal: Borehole fluldconfigurations used in3-phase holdupmeasurement to simulate a horizontal well.

Yw Y. YA Bottom Middle ToP Table4: Accuracy and precision (18 second) of the RST.Aoll% % % Section Section SectIon holdup meaaurementj Yo, for the tool eccentered and centered In

1 100 0 0 Water Water Water the borehole with sevaml casings and aegregatad borehole fluids.

2 66 33 0 Water WaterThis analysis solves for +SO and Y. at the same time.

3 33 66 0 Water

4 66 0 33 Water

: ~ ~

Water

5 33 0 66 Water

6 33 33 33 Water

7 0 100 0 oil oil oil

8 0 66 33 Oil oil Air

9 0 33 66 Oil Air Air

10 0 0 100 Air Air Air

Table 2: Formations used for >phaae holdup chamcterkatlon. Table 5: Accuracy and preclalon (18 eecond) of the RST-A

Tool Borehole Formationvolume of formation oil measurement, $S., for the tool eccentered

Poeition Diameter Llth. (:) Fluidand centered In the borehola with several casings and aegmgatedborehole flulda. This analyala solves for $430 and Y. at the came

Eccent. & Cent. 8.5 Lime 17 Water & Oil time.

Eccent. & Cent. 8.5 Lime 43

“ ~

Water & Oil

Eccent. & Cent. 10 Lime o

Eccent. & Cent. 10 Lime 47 Water &Od

Eccent. 10 Sand 33 Water & Oil

Table 3: Fluid combhratlons used with the data for 2-phaaeholdup calculatlona from C/O. The holdup combinations usedsimulated segregated borehole holdup In a horizontal well. Thosecombinations used for calculation of interpretation coefficientsare labeled as “andnolnte’.

—r—–

I I s. I Y. I Y. I Endpoint I,1 0.00 0.00 1.00 yes

2 0.00 0.33 0.67

3 0.00 0.67 0.33

4 0.00 1.00 0.00 yes

5 1.00 0.00 1.00 yes

6 1.00 0.33 0.67

7 1.00 0.67 0.33

8 1.00 1.00 0.00 yes

eFig. 1: Tool positioning In borehole with different Ilners allowingfor cantered and eccedered 3-phaae holdup measurements.

901

8 B.A. ROSC06 SPE 37147

; 0.8 —mw

o- 0.4 -0

L~ (),2

z

0.0

0.0 0.2 04 0.6 0.8 1.0 1.2

Near C/O (Modeled with MCBEND)

l,o~ I I I I J

F’knslFar C/O Ratio .......-..~..-..-.....-..6..~.00--.*...-

0.0 0.2 0.4 0.6 0.8 1.0Far C/O (Modeled with MCBEND)

o.=

z& --?.0au~:

:s 1.8 –awl

.........7 .. ..... .................... ..............o=.= w

~ 1.4 –

alc

0.8 1.0 1.2 1.4 1,6 1.8 2.0 2.2Inelastic Near/Far Ratio (Modeled)

s:2~’.’..:.::1..::lz‘-k-~-‘ 2s: .*.: :*\a 2.0 .. . ..;.................i ................j...*. .. .....

y Y . O.(Mz . --”1

~ 1.2k------*-”----;-~m---

*I** v=lnn d

‘a ““-- .0

l.glo’ r, I I I I0 10 20 30 40 50

porosity (%)

2.4

; 2.2 4

2 ..,.., ......................................... .. ,

:%1.8

:1.6 — G..; ...... : ~ ..........t?.#....._...... ... ..........

0 +~ ‘j Yi=0,67i - ~.-Z=1,4 — ...... .. . .......... . ................... ......... .

g ! Y = 1.00: ....................... .1,2 — ................ ... .......... . ... . .

+

}Illlil lllillj lilllli(lllilllli‘“%0 o 10 30 40 50

Poroa& (“A)

Fig. 3: RST-A Inelestic nedfer countrate mtio es e funotion ofporosity measurad In IC logging mode. The tool was runeccentered and centered Irr a 7-inch, 23-lb/ft casing in both 8.5-and 10-inchborehoies. Air was used to simuiste gaa for thasemeaauramenta. Tha iinee ara drawn to show tha trenda In thedata for various air hoidups In the borehoie.

Fig. 2: Near C/O ratio, far C/O ratio, and Inelastic nearlfar MOfrom maasuramenta versus the MCBEND modelad reaulta. Theline showe a Ilnear flt to both the eccentered and centerad tooidata.

902

SPE 37147 Three-Phaae Holdup Determination in Horizontal Wells Using a Pulsed-Neutron Source 9

2.2- , , r , ,I I I

Cantered Tool [j;

;2.0 - Ineleatlc NIF Ratio .......................{...................................

E 0.3 glml ges ;.

-1.8 - ....i ...,.. .. .............. . . ..... .;Yg=o.od “;’”””””””””””’’’x’”””s”x”””””

: #x~ =

m2

=- !. ........ ..;...,.., .... .....:.~1.6 == : Y, = 0.33 {-’-j’’’=i----

: ~

1,0 — .......... .......... .......... .............. ...!

0.8I I I I I

-lo 0 10 30 40Poros~t~ (Y.)

50

2.2 ! , I , I , I , I , 1’i

L...-. - .. . .. . .. . .. . . .. . .. . .. . .. . ...+Y.- o.oo.:....~ ...........j... i.....-.

J

~~ :-:If*: Y= 0,67: :=

~1.6 — --’,.’.’.” : -.--’’’’..’.;’-.---.’.~..’.::’ 0 . ............~z

s1,4 — ......................... .. ............ ...... .... ......... . ..........

u * yY9=l. oc!z

E

~1,2 — .......... .......... ... ....... . ... ..............

Eccentered Tool~

\

-1.0 – Inelastic NIF Retio .,: ......................................... , .,., :0.3 g/ml Gas ;

\na I I I 1 I“,”

-lo 0 10 30 40 50Poros~/y (%)

Fig. 4: RST-A Inelaatic nearlfar countrate ratios aa a function ofporosity modeled in IC logging mode. The tooi was modeiedeccentered and centered in a 7-inch, 23-iblft casing in both 8.5-and Ill-inch boraholes. Gaa was modeiad with a dansity of 0.3g/mi. Tha Iinss are drawn to show the trands in tha data forvarioua eir holdups in the borehoia.

120Tb

100R Open Points - Eccantbrsd 100 I ““”””””””””””:’’””””””””””””

z

x

.= 60 --- - .;............-..-i............... ...........,.0

zz 20 —- - -..-+..............~

Modeied Data :

:a

-20 0 20 40 60 60 100 12C

Acutai Oii Hoidup (%)

g100 .. ................ .. . ..........

na

x

E 600

...... .... . . .... .. .. . .

zz 20 —’.’++~

000

:1,

-2020 0 20 40 60 80 100 120

Actual Oli Hoidup (%)Fig. 5: Reconstructed hoidup from modelad dsta in which thainterpretation coefficianta ware calculated from the endpointmeasursmente in 7- and 5.L&inch casing. Thasa dsta include onlyoii and water in tha borahola and formation. Tha interpratstionsolved for both Y. and $SO simultansousiy. The line is drawn toindicata perfect reconstruction.

903

10 B.A. ROSCOf? SPE 37747

NearC/O Ratio ,,+-----

Far CIO Ratio ,/T-.

Inelastk N/F Ratio,+ -----

Fig. 6: Modeled response of the RST-A tool to changes Insegregated borehole fluld compoaltion for a 16-P.u. Ilmeatoneformation with a 8.5-inch borehole and a 7-inch, 23-lb/ft casing.The date were modeled ueing gaa with e density of 0.3 g/ml.

Fig. 7: Error in the reconstruction of borahole holdups using themodeled data of the RST-A tool. The formation used for thisanaiysia waa a 16-p.u. iimestone formation with a 8.5-inchborehole and a 7-inch, 23-lb/ft casing. The data were modeledusing gas with a density of 0.3 g/ml.

904

SPE 37147 Three-Phase Holdup Determination in Horizontal Wells Using a Pulsed-Neutron Source 11

-0.0 0.5 1.0 1.5 2,0 2.5

Centralization Error (inches)

50 I“’I’r’ l’’’ l’’’ l’’’ l’r’ l)’ ‘l’.-8.5-inch Borehole ~

g40 :.... 5.5-inch, 17-lb/ft Casing ~:;

st ~n~;w 30 —=y .............................. ..... .. ....... ...........w

a? ~ Y.

u ~Y320 n ., ,. ..... .

z

(n~lo ---;-” , ( !.......... ................... ......... .......... ....

1b

o I I , 1,,11 I I 11! l,-0.0 0,2 0.4 0.6 0.8 1,0 1.2 1.4 1.6

Centralization Error (inchee)

Fig. 8: Error on the 3-phase holdup estimate for atool that isassumed to becantralized when It Isreally off-center. The RMSerror calculation Includes date for both 16 and 33p.u. sandstoneformations with both water and oil Intheformatlon. In addition,the data include boreholea with 3.phase segregated fluids. Forboth casing sizes, it appeere that there Is no significant IncreaaeIrr the RMS error up to about 0.5 Inches. Baeed on this analysla,as long ss the tool is within 0.5 inches of being centered, thenthere is negligible effect on the holdup accuracy.

600

0I I I I d

100 I I 1 4

.::””!::t:....$..::””:”+”++”‘~{’:[le~ ‘

~’o~l’=4000 ..:.7.:.!............ .......’ .,.-..-.U...:.....:..+..~ oil

OF I I I 1 3 3600 700 600 900 1000 tloo

Relative Depth (ft)Fig. 9: Example of production logging meaaurementa performedin a horizontal well drilled with an 8.5-inch bit and completad witha 5.54nch, 17-lb/ft cemented casing. The meaaurernante includePVL and WFL water velocity maasurementa, PVL oil velocltymeasurements, and RST and Ll~ holdup measurements. TheIinas on the plot connecting the data hava been added to halpindicate the trenda.

905