Reservoir Zonation and Permeability Estimation

download Reservoir Zonation and Permeability Estimation

of 12

Transcript of Reservoir Zonation and Permeability Estimation

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    1/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    RESERVOIR ZONATION AND PERMEABILITY ESTIMATION:

    A BAYESIAN APPROACH

    Adolfo DWindt. PDVSA E&P

    Copyright 2007, held jointly by the Society of Petrophysicists and Well LogAnalysts (SPWLA) and the submitting authors.

    This paper was prepared for presentation at the SPWLA 48th Annual LoggingSymposium held in Austin, Texas, United States, June 3-6, 2007.

    1

    ABSTRACT

    We propose a new hybrid approach to computeunbiased permeability estimates in uncored wells usingthe theory of Hydraulic Flow Units (HFU) based on the

    Carman-Kozeny equation. First, a linear regressionscheme is applied to obtain the optimal number of HFUpresent in core data. Next, the results obtained are usedas input for the nonlinear optimization scheme based onthe probability plot from which statistical parameters ofeach population are obtained. Subsequently, Bayes ruleis applied for clustering core data into its respectiveHFU. Finally, an algorithm based on Bayesianinference is applied to predict permeability in uncoredwells.

    The methodology is applied to a Venezuelan sandstonereservoir and to a Middle East sandstone reservoir.Application of the methodology allows permeability prediction in cored wells with correlation coefficientsabove 0.95 for the field cases under analysis.Permeability profiles in uncored wells compare wellwith pressure transient test results.

    Among primary applications are better productivityindex assessments, enhanced petrophysical evaluations,and improved reservoir simulation models. Coupling ofNonlinear optimization with Bayesian inference provesa robust way for performing data clustering providingunbiased estimations.

    INTRODUCTION

    Any reservoir description program should address the problem of describing the pore space geometry bysubdividing the reservoir into units and assign to themvalues for those rock parameters being described(Haldoresen, 1986). Core analysis provides afundamental source of reservoir information because itis the only physical specimen recovered from thereservoir suited for comprehensive rock description at a pore level (microscopic level). Unfortunately, coremeasurements are both expensive and scarce. On the

    other hand, wireline logs, representing a larger volumeof investigation (macroscopic level), is one of the mostabundant and economical sources of reservoirinformation being the primary tool for analysis andreservoir description. Permeability is one of the mostimportant petrophysical parameter and it is difficult toestimate in the absence of core measurements.Therefore, relating pore throat attributes (obtained onlyform core measurements) to wireline log measurementsis always a challenge.

    Amaefule et al. (1996) proposed the hydraulic flow unitconcept to be used as a principle for subdividingreservoir in different rock types reflecting differentpore-throat attributes. In this regard, the FZI (flow zoneindicator) represents the primary parameter foridentifying those rock types constituting the foundationof this reservoir characterization tool.

    Many techniques have successfully been applied inorder to both identify the number of clusters present incore data and to properly assign data into its respectivecluster. Among these techniques can be mentioned thefollowing: cluster analysis, probability plots

    (Abbaszadeh et al., 1996), neural networks (Aminian etal., 2003), multivariable regression (Guo et al., 2005),fuzzy logic (Cuddy et al., 2000), and multi-lineargraphical clustering (Al-Ajmi et al., 2000). Abbaszadehet al. (1996) suggested the use of non-linearoptimization. However, it has not been applied beforeto determine the number of HFU present in core dataand their statistical properties. Kapur et al., (2000)combined wireline logs and petrologic description viaBayes Theorem in order to produce probability logs forfacies identification. A similar approach is applied inthis paper. The objective of this paper is to determineHFU by applying non-linear optimization coupled with

    the Bayes rule to perform data clustering.Subsequently HFU is inferred in uncored wells via abayesian inversion scheme.

    HYDRAULIC FLOW UNIT CONCEPT

    A flow unit is defined as a volume of rock where porethroat properties of the porous media that governhydraulic character of the rock are consistently predictable and significantly different from those ofother rocks (Abbaszadeh et al., 1996). A reservoir

    U

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    2/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    ought to be divided into flow units to properly describeits performance when it is subject to different production schemes. Two approaches have beendeveloped in the industry for performing thissubdivision: a geological point of view and engineering

    point of view. Here it will be used the engineeringapproach based on dynamic definitions. Mohammed(2006) suggested the use of the concept of engineeringfacies or dynamic rock type in order to avoid anyconfusion with the geological facies definition used inthe geological approach.

    Based on fundamental theory, assuming a bundle ofstraight capillary tubes, and introducing the concept ofmean hydraulic radius (Bird et al., 1960), permeabilitycan be estimated by:

    2e

    mhk r

    2

    (1)

    where is effective porosity, tortuosity, and re mh isthe mean hydraulic radius.

    Equation (1) provides a relationship between permeability and mean hydraulic radius showing itsstrong relationship with pore geometry. By combining

    porosity, rmh, and surface area per unit grain volume(Sgv) with equation (1), the Carman-Kozeny model for ageneralized geometry is obtained (Amaefule et al.,1993)

    3e

    22 2

    gv e

    1k

    F S 1

    (2)

    where k is given in m2 and Sgv is in m-1. The

    effective porosity is obtained either from well logs orcore measurements.

    From (1) follows that

    2mhe

    kr

    (3)

    The mean hydraulic radius has a strong correlation withdifferent petrophysical parameters such as (Amaefule,et al., 1988): stress corrected porosity and permeability,capillary pressure derived pore throat radius, formationfactor, cation exchange capacity, saturation exponent,and relative permeability among others. Additionally,the mean hydraulic radius can be correlated with thecharacteristic length used in the definition of Reynoldsnumber for porous media (Jones, 1987). Thus, the

    selection ofek/ as a predictor of pore space

    attributes is both useful and physically sound. Based onthese observations and from equation (3), equation (2)is rearranged to give the following:

    1

    0 0314 1e

    e e s gv

    k

    . F S

    (4)

    The units of k are md. Now, a reservoir quality index(RQI) is defined by

    e

    kRQI 0.0314

    (5)

    Also, the Flow Zone Indicator (FZI) is defined as

    s gv

    1FZI

    F S

    (6)

    RQI and FZI are given in m. According to equations(5) and (6), (4) is rewritten as:

    zRQI FZI (7)

    and

    1e

    z

    e

    (8)

    Thus, in a log-log plot, core data corresponding to a particular hydraulic flow unit will plot as a unit slope

    straight line with intersect at z =1 equals to FZI.Having obtained FZI we are in capability ofdetermining intrinsic petrophysical properties of a givenhydraulic unit and, by doing so, a reservoir can bedivided into a discrete number of hydraulic units. Oncea HFU or engineering facie is identified, permeability iscalculated by (Amaefule et al., 1993):

    32

    21014

    1

    e

    e

    k FZI

    (9)

    The Carman-Kozeny equation provides good estimatesfor well-sorted samples from which the average particlesize diameter is known (Wu, 2004). However,knowledge of grain diameter and specific surface areais critical in the Carman-Kozeny model. The latterrepresents a major limitation of such model. Also,applicability of the Carman-Kozeny model isquestionable in the presence of diagenesis (Abbaszadehet al., 2000).

    2

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    3/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    One of the advantages of the engineering faciesdefinition over the geological definition of facies is thatthe FZI has a strong correlation with irreducible watersaturation, specific surface, grain size, and mineralcontent (Svirsky et al., 2004). Moreover, J-Function

    derived water saturation values can be obtained usingthe FZI concept (Desouky, 2003) and be successfullyused for allowing accurate 3D reservoir modeling(Obeida et al., 2005).

    Despite its limitations, the Carman-Kozeny model has been widely applied both in sandstone and incarbonates with consistent results. Nevertheless, whenreservoir rock deviates from the capillary bundle modelthe Carman-Kozeny model fails in predicting

    permeability. Therefore it is important to introduce amodification. A variable exponent in the porosity groupis introduced as well as a correction for the degree of

    cementation (Civan, 2002-2003). This modificationleads to obtain straight lines with slopes different than

    one in the RQI-z plot.

    NON-LINEAR OPTIMIZATION

    Because of random errors and minor fluctuations ofgeological factors controlling petrophysical attributes,data will cluster around the straight lines showing somescatter (Abbaszadeh et al., 1996), consequently the FZIwill be distributed around an expected value.

    Since the FZI can be represented as the product of

    several factors, according to the Central LimitTheorem, its probability distribution will be log-normal(Jensen, 1997). This observation has profoundimplications because the properties -mean and standarddeviation- of the log-normal distribution or, more

    broadly speaking, the Gaussian distribution, are wellknown, being given this probability distribution by:

    21 x

    21f x, , e2

    (10)

    In equation (10) and are the expected value and the

    standard deviation.

    In a probability plot the logarithm of the FZI will plotas a straight line. Unfortunately, when several

    populations (flow units) are present in the data, it iscommon to observe superposition of severaldistributions.

    For the case of a multi modal distribution the probability plot is not linear, rather a smooth curve isobtained, thus, attempting to identify straight line insuch a plot is though and inaccurate. Moreover, because

    of superposition effects, the number of clusters presentwill be masked introducing bias to interpretationsextracted from the plot. Hence, data clustering using the

    probability plot requires a rigorous approach.

    Non-linear optimization provides a robust way todecompose a superposition of a multi-modal Gaussiandistribution into its component parents. The goal is tominimize the cost function given by the least squarescriteria

    m

    1i

    2

    icalcii

    measi FZIlnFFZIlnF (11)

    where icalci FZIlnF and are the calculated

    and measured cumulative probability of obtaining avalue less than or equal to ln(FZI

    imeasi FZIlnF

    i). The term

    imeasi FZIlnF is obtained from measured data.

    The cumulative probability distribution for amultimodal distribution is given by (Sinclair, 1976):

    N

    1j

    ijjicalci zFfFZIlnF (12)

    where fi is the fraction of the data belonging to aparticular population, N is the number of HFU, and F(z)is given by (Jensen, 1997):

    N

    1i

    i2

    erff12

    1zF

    z(13)

    In equation (13) erf() is the error function and z is thestandard normal variable defined by

    lnFZI

    lnFZI

    lnFZI-z

    (14)

    The standard variable z may be estimated from rationalapproximations (Jensen, 1997), ln(fzi), andln(fzi) are themean and standard deviation of ln(fzi)

    One of the major advantages of this approach is that themean and standard deviation of each population arecalculated values, not approximations. On the other, itis critical to provide the algorithm with initial guessvalues close to the solution because non-physicalsolutions may arise. Also, convergence problems mighttake place. Another problem that has to be dealt with ishow to figure out the optimum number of HFU presentin the core data.

    3

    U

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    4/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    Because the optimum number of clusters is not knownin advance, an iterative procedure is required todetermine the number of hydraulic flow units. In orderto overcome these two adverse situations, it is

    performed a preliminary graphical clustering analysis as

    follows:

    1. Assume the number of HFU in dataset

    2. Estimate FZI values from the RQI-z plot.3. Perform non-linear optimization in order to obtain

    the minimum of the misfit function given by (11)for the assumed number of HFU

    4. Increase the number of HFU and go back to step 2;perform this step until the misfit function reaches aplateau.

    Once the optimum number of HFU has beendetermined, the mean, standard deviation and the

    fraction of the sample corresponding to each HFU canbe determined from the minimization of equation (11).An alternative linear optimization algorithm (Al-Ajmi,

    et al., 2000) using the RQI-z plot can be applied withequivalent results.

    BAYES THEOREM

    Once the statistics ( and) of each distribution havebeen obtained, the next step is to properly assign coremeasurements to their respective HFU. The basis ofthis approach is the fact that a particular FZI maycorrespond to any HFU, but it will take place with

    different values of probabilities for each HFU. Thismay be visualized according to the probability treeshown in Figure 1.

    FZIi

    FZIi

    FZIi

    .

    .

    .

    .

    ff HFU1

    HFU

    1

    ffHFU2HFU2

    ffHFH

    N

    HFHN

    i 1P lnFZI lnfzi HFU

    i 2P lnFZI lnfzi HFU

    i nP lnFZI lnfzi HFU

    FZIi

    FZIi

    FZIi

    .

    .

    .

    .

    ff HFU1

    HFU

    1

    ffHFU2HFU2

    ffHFH

    N

    HFHN

    i 1P lnFZI lnfzi HFU

    i 2P lnFZI lnfzi HFU

    i nP lnFZI lnfzi HFU

    Figure 1: Probability Tree showing alternative branches foreach HFU leading to the same FZI

    Each branch in the probability tree leads to the sameFZI but with different probability. Then, from theBayes Theorem (DWindt, 2005):

    4

    N

    1i

    ii

    iii

    HFU)fziln()FZIln(Pf

    HFU)fziln()FZIln(Pf)fziln()FZIln(HFUP

    (15)

    The symbol means in the neighborhood of (Kapur,2000). N is the number of HFU. The termP(HFUiln(FZIj)ln(fzi)) is the probability ofobtaining a particular HFU given that a particular

    ln(FZI) is in the neighborhood of ln(fzi). The term fiis an a priori estimate of the probability occurrence of agiven HFU. The term P(ln(FZIj) ln(fzi)HFUi) is the

    probability that a particular ln(FZI) is within certaininterval given that it belongs to a particular HFU, it iscalculated using equation (10).

    The application of equation (15) permits thedetermination of the boundaries of each HFU in asimple way. Probability logs can be generated for eachHFU so the boundaries are easily identified. The

    procedure is simple and intuitive; the decision rule isbased on a probability value.

    INVERSE PROBLEM: BAYESIAN INVERSION

    Once core measurements have been clustered into theirrespective parents, the inverse problem must beaddressed. That is, predicting hydraulic flow units onwells without core measurements based only wirelinelogs.

    Figure 2 shows a probability tree with the differentalternatives or paths available leading to a particular setof wireline logs. It is assumed that different HFU mayresult in the same set of log values, but this takes placeat a different probability values.

    XXii

    XXii

    XXii

    ..

    ....

    ..

    ..

    ..

    P(HFU

    1

    P(HFU

    1))

    P(HFU

    2P(H

    FU2))

    P(HFUn

    P(HFUn))

    i 1P X HU

    i 2P X HU

    i nP X HU

    XXii

    XXii

    XXii

    ..

    ....

    ..

    ..

    ..

    P(HFU

    1

    P(HFU

    1))

    P(HFU

    2P(H

    FU2))

    P(HFUn

    P(HFUn))

    i 1P X HU

    i 2P X HU

    i nP X HU

    Figure 2: Probability Tree showing alternative branches foreach HFU leading to the same set of wireline log readings

    In the case of multiple wells logs, equation (15) ismodified so that the probability of occurrence of HFUgiven a wireline log data set may be calculated, therequired expression is the following (DWindt, 2005):

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    5/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    i j i

    i j n

    i j ii 1

    f P X HFUP HFU X

    f P X HFU

    (16)

    where fi the a priori estimate of a particular HFUobtained from core data, Xj represents a group of logvalues which are near a set of given log values . Theterm P(HUiXj) is the probability of obtaining aHFU given that the wireline log readings are in theneighborhood of and P(XjHUi) is the probability ofobtaining a given set of log readings given a particularHFU. When using log data we are dealing with discretedata, thus probability given by equation (16) can beestimated by:

    j j

    i j

    j

    n X HFUP HFU X

    n X

    (17)

    The term n(X jHFUj) is the number of data points belonging to HFUj in a given interval or bin and n(X j) is the number of all data points falling in the

    bin. For bins without data, probabilities are interpolatedusing an inverse distance method (Isaaks, 1989).

    FIELD CASES

    I Sandstone Reservoir:

    This formation is a Eocene clastic reservoir located atthe center of the basin of Lake Maracaibo (Venezuela);

    at approximately 11,000 ft (TVD). Sand deposits are primarily of channel-type corresponding to a fluvial-deltaic deposition system. Total reservoir thickness isup to 900 ft. A total of 21 wells have been drilled withcore data acquired only on three of them. Permeabilityranges from less than 0.1 md up to 2000 md. Net pay

    porosity is between 12% and 25%, non-pay rocks areconsidered to have less than 10% porosity.

    Data Corrections. Porosity and permeability core datahas to be both Klinkenberg and stress-corrected tosimulate reservoir-confining conditions. Stresscorrections are made according to Jones (1986).

    Figure 3 shows a permeability-porosity cross plot thesandstone reservoir at net overburden (NOB)conditions. Variation of 2 orders of magnitude for agiven porosity indicates that other factors rather than

    porosity itself- are governing formation transmissibility.

    An exponential or potential model for the k-relationship will not properly reproduce formation

    permeability (Jennings et al., 2001).

    0.0001

    0.001

    0.01

    0.1

    1

    10

    100

    1000

    10000

    0 0.05 0.1 0.15 0.2 0.25Porosity (fraction) @ NOB

    Permeability(mD)@

    NOB

    Figure 3: Permeability-Porosity relationship for sandstonereservoir

    Hydraulic Flow Unit Identification. Preliminary

    analysis of the probability RQI-z plot (Figure 4)indicates that there is not clear boundary among thedifferent flow units. Furthermore, it is challenging todetermine the number of flow units present in the coremeasurements

    0.001

    0.01

    0.1

    1

    10

    0.01 0.1 1PHIz @ NOB

    RQI@N

    OB(microns)

    Figure 4: RQI-z relationship for sandstone reservoir; the plotdoes not show clear boundaries between HFU

    From the minimization of the cost function given byequation (11) it is determined that 7 clusters or HFU arerequired to properly model core data (Figure 5).For 7 clusters the cost function flattens indicating that afurther increase of the number of clusters does notsignificantly reduce the objective function.

    5

    U

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    6/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    0.001

    0.01

    0.1

    1

    0 1 2 3 4 5 6 7 8

    Number of Clusters

    CostFunction

    Figure 5: Cost function (least-square criteria) as a function ofnumber of clusters

    Once the optimum number of cluster is determined, theminimization of equation (11) allows the calculation ofthe statistics of each parent or HFU. Figure 6 shows thematch given by non-linear optimization. A closeapproximation on the probability plot is obtained byusing 7 HFU.

    6

    It is important to state that any attempt of fittingstraight lines directly in the plot lead to a wrong numberof HFU. In this particular case maybe 4 or 5 straightlines might be fitted leading to wrong statistics.

    0.1

    1

    10

    -3 -2 -1 0 1 2 3

    Z

    FZI

    Match

    Data

    Figure 6: Probability plot of core data showing non-Linearregression match obtained from minimization of misfitfunction

    It has to be noted that the parameters ( and) obtainedfrom minimization of the misfit function correspond tothe lognormal distribution. Therefore, propertransformations are needed to obtain the actual valuesof FZI for each HFU. To complete such a task (Jensen,1997) the following expressions are used:

    2ln FZIi lnFZI+0.5FZI =e

    (18)

    2lnFZI2 2

    FZI ln FZI= e 1 (19)

    Table 1 shows the statistics obtained from theoptimization scheme for each HFU.

    Table 1. Non-linear optimization results. Statisticsfor each HFU

    HFU fi FZIi FZIi1 0.0425 0.1496 0.18632 0.1593 0.6489 0.02773 0.2116 1.3032 0.00404 0.1575 2.0360 0.00985 0.1873 3.4079 0.07406 0.2044 5.8613 0.0952

    7 0.0373 8.5924 0.1409

    For the sake of comparison, alternative choices forclustering data such as: K-means, Wards algorithm,and Kohonens self-organizing maps techniques wereapplied in order to evaluate the consistency of the

    proposed method. Table 2 shows a comparison of theexpected value of FZI for each HFU obtained fromdifferent methods

    Table 2. Comparison of expected value of FZI fromdifferent clustering results

    HFUNon-

    Linear

    Ward

    Algorithm

    K-Means Kohonen

    1 0.150 0.196 0.203 0.226

    2 0.649 0.564 0.627 0.675

    3 1.303 0.969 1.046 1.219

    4 2.036 1.672 1.647 2.038

    5 3.408 2.578 2.513 3.327

    6 5.861 4.582 4.457 5.393

    7 8.592 7.409 7.297 7.890

    From the results presented in table 2 it is observed thatthe four different methods provide values for the FZI ina close range indicating consistency in results. For thecase of the Kohonens self-organizing maps technique,

    the absolute differences with the non-linear results areless than 5% in average (excluding HFU 1). Furthermore, hypothesis test for the mean between Kohonensand non-linear results show that, with the exception ofthe HFU1 and HFU7, differences are not statisticallysignificant.

    Once the optimal number of clusters is determined it isnecessary to establish the boundaries of each HFU inorder to classify the core data. In this paper a Bayesianapproach is applied to perform this task. A probability

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    7/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    of occurrence for each HFU given a particular value ofFZI is determined by means of the Bayes rule. This isaccomplished by applying of equation (15). Thecorresponding HFU for a given FZI will be that onewith the highest probability of occurrence. Results of

    the application of equation (15) are shown in Figure 7.

    0.00

    0.20

    0.40

    0.60

    0.80

    1.00

    0.1 1 10FZI

    ProbabilityofOcurrence

    HFU1

    HFU2

    HFU3

    HFU4

    HFU5

    HFU6

    HFU7

    Figure 7: Probability log for each HFU as a function of theFlow Zone Indicator (FZI). Venezuelan reservoir

    It is important to remark that boundaries for each HFUare easily determined by a probability value. There isno need for clustering algorithms such as Ward or K-means algorithms for clustering core data into itsrespective parent, a simple probability criteria is

    sufficient. The RQI-z plot (Figure 8) shows unit slope

    lines passing through each cluster or HFU following theclassification made based on the Bayes Theorem; here

    the intersection at z=1 provides the expected value FZIfor each HFU.

    0.001

    0.01

    0.1

    1

    10

    0.01 0.1 1PHIz @ NOB

    RQI@N

    OB(microns)

    HU1 HU2 HU3 HU4 HU5 HU6 HU7

    Figure 8: RQI-z plot presenting core data grouped byHydraulic Flow Unit

    Once core measurements are assigned to their

    respective HFU, k- cross-plot (Figure 9) shows avariation of less than a half of logarithmic cycle foreach HFU. Variability within a particular HFU is small.Variation in permeability, for a given porosity, is

    dramatically reduced when reservoir is subdivided intoflow units with distinctive pore-throat attributes.

    0.001

    0.01

    0.1

    1

    10

    100

    1000

    10000

    0.05 0.10 0.15 0.20 0.25Porosidad (fraccion) @ NOB

    K@N

    OB(mD)

    HU1 HU2 HU3 HU4 HU5 HU6 HU7

    Figure 9: k- cross plot for Field Case I

    Hydraulic flow unit inference from log data. Solutionof the inverse problem requires dealing with wirelinelog data but, prior to any kind of analysis, log data must

    be depth-matched to core measurements and correctedfor environmental effects. Also, wireline log responses

    have to be normalized; this is a mandatory and sensitivetask (Hunt et al., 1996).

    Environmentally-corrected wireline logs are correlatedwith FZI via Spearmans rank correlation method(Amaefule et al., 1988). Among the different logsavailable, it was determined that gamma-ray, Neutron

    porosity, and density porosity logs gave the highestcorrelation coefficient with FZI.

    In order to apply equation (17), a Bayesian algorithmwas implemented with the log measurements and withconventional core laboratory measurements.Application of equation (17) allowed the determination

    of probabilities of occurrence for each HFU for a givenset of wireline logs, being selected that HFU with thehighest probability. Once a HFU was determined,

    permeability was calculated by using equation (9).

    Predicted and measured permeability are shown in alog-log cross plot in Figure 10. Correlation coefficientis equal to 0.97, indicating a high accuracy in the results

    provided by the Bayesian inversion method.

    7

    U

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    8/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    0.01

    0.1

    1

    10

    100

    1000

    10000

    0.01 0.1 1 10 100 1000 10000K core (mD)

    Kcalculated(mD)

    Figure 10: Calculated-Measured permeability cross plot forcore data of three wells; r2 = 0.97

    Figure 11 and Figure 12 show a comparison between

    actual and calculated values for two of the wells withcore data. Calculated permeability profile reproducesthe general tendency shown by core measurements

    10550

    10650

    10750

    10850

    10950

    1 10 100 1000 10000

    K(mD)

    Depth(feet)

    K core

    K Calculated

    Figure 11: Permeability profile for Well A; comparison between Bayesian generated permeability and actual coredata. Venezuelan sandstone reservoir

    11000

    11100

    11200

    11300

    11400

    11500

    0.1 1 10 100 1000 10000

    K(mD)

    Depth(feet)

    K core

    K Calculated

    Figure 12: Permeability profile for Well B; comparisonbetween Bayesian generated probability and actual core data

    These figures illustrate the ability of the inversionscheme to reproduce permeability by using only welllogs assuming that no core data had been taken at all.

    Dynamic validation. Proposed HFU scheme is tested ina well without core data completed in the reservoir.Production data available includes RFT, BUP and PLTmeasurements. The pressure transient test indicated akh of 10400 mD-ft and a skin of 2. From PLT/RFTresults it was calculated a PI of 25.65 STB/D/psi.

    Bayesian inversion was applied using available log datain this well determining a kh of 10950 md-ft (relative permeability data was used to correct for liquidsaturation effects; kro@Swi was estimated at 0.73).A synthetic PI of 24.05 STB/D/psi was calculated byusing pseudo-steady Darcys equation. Calculated PIand kh values differed only in 5% and 6% respectivelyfrom actual data.

    These results indicate that not only permeability is properly reproduced but also that dynamic well

    8

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    9/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    response can be simulated with a high degree ofconfidence.

    Ta l pro s Field c

    9

    II - Middle East Reservoir:

    Core data available from a Middle East sandstone

    reservoir was analyzed using the previously appliedmethodology. Figure 13 is a permeability-porositycross-plot showing variation of to 4 orders ofmagnitude in permeability.

    0.1

    1

    10

    100

    1000

    10000

    0.05 0.10 0.15 0.20 0.25 0.30

    Porosity (fraction) @ NOB

    K@

    NOB(mD)

    v

    Figure 13: Permeability-Porosity relationship for Middle EastReservoir.

    Non-linear optimization on the probability plot allowsthe determination of 6 HFU present in core data(Figure 14). Observe the smoothness of the curve,which cause to be challenging attempting to drawstraight lines directly in the plot.

    0.1

    1

    10

    -3 -2 -1 0 1 2 3Z

    FZI

    Match

    Data

    Figure 14: Probability plot of core data showing non-linearregression match obtained from minimization of cost function

    field case II

    Table 3 presents the statistical properties of each HFUobtained from the non-linear optimization

    ble 3 HFU statistica pertie ase II

    HFU fi FZIi FZIi1 0.3645 0.5848 0.03282 0.1754 1.1432 0.00063 0.1738 1.9595 0.01644 0.1705 3.1936 0.03215 0.0492 4.6912 0.10236 0.0667 7.4872 0.2858

    Application of equation (15) allows the determinationof probability logs (Figure 15) for each HFU permittingthe identification of each HFU boundaries.

    0.00

    0.20

    0.40

    0.60

    0.80

    1.00

    0.1 1 10

    FZI

    Probabilityofocurrence

    HU1

    HU2

    HU3

    HU4HU5

    HU6

    as a function of theow Zone Indicator (FZI). Field case II

    forming clusters with small variability withinthem.

    Figure 15: Probability log for each HFUFl

    RQI-z cross plot (Figure 16) shows 6 different HFUrepresented. Despite data being tightly clustered,

    proposed algorithm is capable of separating one fromother

    0.001

    0.01

    0.1

    1

    10

    0.01 0.1 1PHIz @ NOB

    RQI@

    NOB(microns)

    HU1 HU2 HU3 HU4 HU5 HU6

    C

    Figure 16: RQI-z plot presenting core data grouped byHydraulic Flow Unit Middle East Reservoir

    U

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    10/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    10

    t; less than an order inmagnitude is observed when previously 4 orders ofmagnitude was the order of day.

    Inspection of Figure 17 indicates little variability within

    a given HFU in the k- cross plo

    0.01

    0.1

    1

    10

    100

    1000

    10000

    8200

    8250

    8300

    8350

    8400

    8450

    8500

    8550

    8600

    0.1 1 10 100 1000

    K (mD)

    Depth(ft)

    K core (mD)

    K Calculated (mD)

    0.00 0.05 0.10 0.15 0.2Porosidad @ N

    0 0.25 0.30 0.35OB

    K@N

    OB(mD)

    HU1 HU2 HU3 HU4 HU5 HU6

    C

    fficient of r = 0.96 (Figure 18). Again,e Bayesian inversion showed accuracy in predicting

    permeability.

    Figure 17: k- cross plot Field case II

    For this field-case several well logs were available:gamma-ray, neutron density, FDC, shallow and deepresistivity, microlog, and sonic among others. Rankcorrelation determined that only three well logs wererequired: GR, FDC, and Neutron porosity. Inversionscheme permitted to determine permeability with acorrelation coe 2

    th

    0.01

    0.1

    1

    10

    100

    1000

    10000

    0.01 0.1 1 10 100 1000 10000

    K core (mD)

    Kcalculated(mD)

    ls with core datadrilled in the reservoir. The inversion processaccurately reproduces permeability.

    Figure 18: Calculated-Measured permeability cross plot; r2 =0.96. Middle East reservoir

    Figure 19 presents one of the wel

    Figure 19: Permeability profile for a well drilled in thereservoir; comparison between calculated permeability andactual core data.

    All the procedures previously shown are easilyimplemented in a spreadsheet environment; no specialcommercial software is required for performing thesetasks.

    CONCLUSIONS

    A novel technique is presented and successfully appliedfor clustering core data. Non-linear optimization iscoupled with Bayes Theorem for grouping core datainto its respective parent.

    Bayesian inversion proves to be a powerful tool forpredicting rock properties based only on log data.

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    11/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    11

    Application of the methodology allows thedetermination of the true (not estimates) mean andstandard deviation of each HFU.

    Accurate permeability prediction is achieved by means

    of the application of the Carman-Kozeny model.

    Reproduction of well performance behavior indicatesthat outcomes of the inversion process can be used asinput for numerical simulation purposes.

    Consistency in results when compared to othertechniques validates the application of the proposedmethodology rendering it suitable for data analysis for

    predicting pore-throat attributes.

    Productivity assessments can be performed in an easyand accurate way reducing the need for acquisition of

    expensive testing data.

    REFERENCES

    Abbaszadeh, M., Fuji, H., Fujimoto, F., 1996,Permeability Prediction by Hydraulic Flow Units Theory and Applications. SPE Formation Evaluation.

    December. 263-271

    Abbaszadeh, M., Koide, N., Murahashi, Y., 2000,Integrated Characterization and Flow Modeling of aHeterogeneous Carbonate Reservoir in Daleel Field,Oman. SPE Formation Evaluation and Engineering.Vol 3. 150-159.

    Amaefule, J., Kersey, D., Marshall, D., Poewl, J.,Valencia, L., Keelan, 1988, Reservoir Description: APractical synergistic engineering and geologicalapproach based on analysis of core data. SPE paper18167 presented at the 63rd Annual TechnicalConference and Exhibition of the Society of PetroleumEngineers, Houston, Tx.

    Amaefule, J., Altumbay, M., Tiab, D., Kersey, D.,Keelan., D., 1993, Enhanced Reservoir Description:Using Core and Log Data to Identify Hydraulic FlowUnits and Predict Permeability in Uncored

    Intervals/Wells. SPE paper 26436. 68th AnnualTechnical Conference and Exhibition of the Society ofPetroleum Engineers, Houston, Tx.

    Aminian, K., Ameri, S., Oyerokun, A., Thomas, B.,2003, Prediction of Flow Units and PermeabilityUsing Artificial Neural Networks SPE paper 83586

    presented at the SPE Wester Regional/AAPG PacificSection Joint Meeting. Long Beach, California

    Al-Ajmi, F., Holditch S., 2000, PermeabilityEstimation Using Hydraulic Flow Units in a CentralArabia Reservoir. SPE Paper 63254 presented at the2000 SPE Annual Technical Conference andExhibition. Dallas, Texas

    Bird, R.B., Stewart, W.E and Lightfoot, E.N., 1960,Transport Phenomena.

    Civan, F., 2002, Fractal Formulation of the Porosityand Permeability Relationship Resulting in A Power-Law Flow Unit Equation A leaky Tube Model. SPEPaper 73785 presented at the SPE InternationalSymposium and Exhibition of Formation DamageControl held in Lafayette, Louisiana

    Civan, F., 2003, Leaky-Tube Permeability Model forIdentification, Characterization, and Calibration of

    Reservoir Flow Units. SPE paper 84603 presented atthe SPE Annual Technical Conference and Exhibitiom.Denver.

    Cuddy, S.J., 2000, Litho-Facies and PermeabilityPrediction From Electrical Logs Using Fuzzy Logic.SPE Reservoir Eval. & EngVol. 3, No. 4, 319-324

    Desouky, S., 2003, A new Method for Normalizationof Capillary Pressure Curves. Oil and GasTechnology. Rev. IFP. Vol 58. No. 5. 551-556

    DWindt, A., 2005, "Hydraulic flow unit identificationby bayesian inference for permeability prediction. MscThesis. Universidad de Zulia.

    Guo, G., Diaz, K., Paz, F., Smalley, J., Waninger, E.A.,2005, Rock Typinc as an Effective Tool forPermeability and Water-Saturation Modeling. A CaseStudy in a Clastic Reservoir in the Oriente Basin. SPEPaper 97033 presented at the 2005 SPE AnnualTechnical Conference and Exhibition, Dallas, Texas

    Haldoresen, H.H., 1986, Simulator ParameterAssignment and the Problem of Scale in ReservoirEngineering. Reservoir Characterization. AcademicPress, inc. 293-340

    Hunt, E., Ahmed A., Pursell, D., 1996, Fundamentalsof log analysis. Part IV: Normalizing logs withhistograms. World Oil. October. 101-102.

    Isaaks, E., Srivastava, R., 1989, An introduction toapplied geostatistics. Oxford University Press.

    Jennings, J., Lucia, J., 2001, Predicting PermeabilityFrom Well Logs in Carbonates With a Link to Geologyfor Interwell Permeability Mapping. SPE paper 71336.

    U

  • 8/3/2019 Reservoir Zonation and Permeability Estimation

    12/12

    SPWLA 48th Annual Logging Symposium, June 3-6, 2007

    12

    SPE Annual Technical Conference and Exhibition.New Orleans, Lousiana.

    Jensen, J., Lake, L., Corbett, P., Goggin, D, 1997,Statistics For Petroleum Engineers and Geoscientists.

    Prentince Hall.

    Jones, S.C., 1987, Using the inertial Coeffiecient toCharacterize Heterogeneity in Reservoir Rocks. SPE

    paper 16949 presented at the 62nd Annual TechnicalConference and Exhibition of the Society of PetroleumEngineers, Houston, Tx.

    Kapur, L., Lake, L., Sepehrnoori, K., 2000, Probabilitylogs for facies classification. In Situ. Vol. 4. No. 1. 57-78

    Mohammed, D., 2006, Personal conversation.

    Obeida, T.A., Al-Mehairi, Y.S., Suryanarayana, K.,2005, Calculation of Fluid Saturations From Log-Derived J-Functions in Giant Complex Middle-EastCarbonate Reservoir. SPE Paper 95169 presented atthe 2005 SPE Annual Technical Conference AndExhibition. Dallas, Texas.

    Sinclair, A. J., 1976, Applications of ProbabilityGraphs in Mineral Exploration. Association of

    Exploration Geologist SpecialVol. 1.

    Svirsky, D., Ryazanov, A., Pankov, M., Corbett, P.,Posyoesv, A., 2004, Hydraulic Flow Units ResolveReservoir Description Challenges in a Siberian OilField. SPE paper 87056 presented at SPE Asia PacificConference on Integrated Modeling for AssetManagement. Kuala Lumpur, Malaysia.

    Wu, T., 2004, Permeability Prediction and DrainageCapillary Pressure Simulation in SandstoneReservoirs. Phd Dissertation. Texas A&M University.

    ACKNOWLEDGMENTS

    The author thanks PDVSA for permission to publishthis work and for permitting the use of field data.

    A note of special gratitude goes to Onaida Pereira forher assistance and Dr. Rodolfo Soto for his advice, andfor providing data for this research. Specialappreciation is granted to Jesus Salazar for hiscorrections.

    ABOUT THE AUTHOR

    Adolfo DWindt has worked for eight years asreservoir engineer for PDVSA ,Maracaibo, Venezuela,dealing with sandstone reservoirs under primary

    depletion and waterflooding projects. His researchinterests include well transient analysis, formationevaluation, productivity assessment, and uncertaintyanalysis. He received a B.Sc degree in 1997 fromUniversidad del Zulia, and M.Sc. degrees from the

    University of Texas at Austin, in 2004 and fromUniversidad del Zulia in 2006, all in PetroleumEngineering.