You are on page 1of 12

SPE-184314-MS

PVT Analysis: Validity Checks and EOS Tuning Techniques for Retrograde
Gas Condensate Reservoirs
S. U. Ibeh and S. E. Chubueze, Shell Professorial Chair, Department of Petroleum Engineering, FUT

Copyright 2016, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Nigeria Annual International Conference and Exhibition held in Lagos, Nigeria, 2– 4 August 2016.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
PVT analysis is pertinent to field-wide development and optimization. This is because we need to know
the over-all behavior of the fluid from reservoir through production and process facilities, down to the
refinery. Its emergence as a separate discipline has been supported by the development of modern
computer software that utilize EOS models for simulating experiments and representing fluid phase
behaviours. PVT simulations are carried out to determine operating conditions that will maximize the
surface liquid content and extend the production plateau duration at the lowest possible cost. In such
simulations, Laboratory derived data are used to tune the EOS models and the results incorporated for
reservoir simulation and studies. Obtaining a good match between EOS and laboratory data depends on
data quality and for retrograde gas condensates, this can be very challenging due to their complicated
phase behavior. Inadequate match results in computational errors and unrepresentative results when used
in reservoir simulations and thus, jeopardize reservoir management decisions which are based on it.
The major contributions of this paper are-first,a review of critical issues and success factors in PVT
analysis and its associated challenges. Second, methods for quality checking PVT data. Third, techniques
for improving Lab – EOS match and their influence in PVT results.
INTRODUCTION
The oil and gas business is technology dependent, risk ridden and capital intensive. Therefore, reducing
uncertainty is key to staying in business. A lot of tests are usually carried out to evaluate a particular
parameter, for instance, well testing and coring as expensive as they are, are usually employed just to
measure and quality check permeability.
No operator would claim to have known his fluid until a detailed PVT analysis has been carried out.
Even when correlations are used to evaluate basic PVT properties on the well site, decisions based on such
evaluations are generally minor and short term.
The most recommended PVT tests for retrograde gas condensates are the CCE and CVD. While CVD
replicates events at reservoir conditions, CCE simulates fluid behaviours at separator conditions.
Retrograde gas condensate reservoirs have temperature between the critical temperature and the
cricondentherm. Depending on their proximity to the dew point curve, they could be regarded as
saturated/rich or under saturated/lean. At pressures above the dew point, the reservoir fluid content and
2 SPE-184314-MS

production is single phase gas. During depletion at constant temperature, pressure goes below the dew
point initiating the condensation of liquids/condensates that existed in equilibrium with the gas phase.
These liquids drop out in the reservoir and being the wetting phase are trapped by capillary induced forces.
The immobility of the stranded condensates creates productivity concerns for operators. As pressure
continues to decrease with production, the liquid droppings revapourize into the gaseous phase, however,
the pressure at which revapourization occurs is way below the abandonment pressure and so cannot be
attained as it is generally constrained by economics.
PVT measurement and analysis is critical to understanding retrograde phenomena due to the high
compositional effect associated with their development. Laboratory derived data need to be matched to an
EOS model which can be used to make up/compensate for properties that were not measured. Uncertainty
in lab measurement and the collection of unrepresentative fluid samples pose a major challenge to EOS
tuning/matching and accurate fluid description. In this paper, challenges of unrepresentatively and
methods for validating PVT /CVD data were presented and were possible, the PVTi compositional
simulator is employed.
Black oil and Compositional Simulation
Black oil models represents multicomponent hydrocarbons and all fluid phases as two pseudo compo-
nents- vapour and liquid. All calculations are based on four PVT properties namely Oil FVF, Gas FVF,
Solution GOR, Volatized OGR. Compositional effects are not explicitly modelled however, small
compositional effect can be represented by variation in solution GOR[1] [2]. A major advantage of black
oil model is reduction in computing time.
On the other hand,compositional models supposes that every component is different. Higher compu-
tational time is required. This is because, for each pressure step, we first have to calculate the moles in
vapour and liquid phases, and for each component - the mole fractions in each phase and their
corresponding properties. However, whenever a multi - phase development is anticipated at any point,
then compositional modelling will be necessary. Facilities for minimizing its attendant increase in
computational time will be discussed later.
Fluid Sampling and Representativity – The Challenge
Fluid can be sampled down/bottomhole or at surface. In either case, the collected samples should be
maintained at the sampling condition. If production is single phase and bottomhole pressure and pressure
at any point in the facilities has not gone below the dewpoint – that is constant composition of the fluid
stream, then either surface or downhole sampling can be done. On the other hand, if a multi - phase has
developed, the separator gas and oil samples are collected and recombined mathematically to obtain the
original feed stream compositions [3].
Representative fluid sampling is key to obtaining quality data. For saturated rich gas condensates, a
small pressure drop could result in two phase and the liquid lost to the reservoir can compromise data
quality. Also unstable temperature and pressure condition can affect equilibrium compositions of vapour
and liquid phases. Gravitational forces cause components of hydrocarbon system to segregate-this implies
that heavier ones are located at lower depths/structure while lighter ones will be positioned upstructure.
Fluids sampled/collected as single phase is most representative regardless of where it has been
collected-surface or downhole[3]. This is because we want to get a fluid that represents the volume being
drained by the well and not necessarily a compositional representation of the sampled point which is just
a local volume of the reservoir since composition vary greatly in both areal and vertical directions.
PVT Experiments
Constant Composition Expansion
In Constant Composition Expansion experiment, the test fluid is put in a visual PVT cell and maintained
at reservoir temperature. The pressure, starting a little above the initial reservoir pressure, is reduced in
steps [of about 10-15psi][4]. Reduction in pressure is typified by the removal of mercury from the cell.
SPE-184314-MS 3

Since the temperature of the system is constant, reducing pressure results in increase in volume according
to Boyles’ law. As presssure is decreased, below saturation pressure [dew point], liquids drop out from
the vapour phase. However, the gases are not removed from the cell but allowed to coexist with the liquid.
This ensures that a constant composition of the cell is maintained. For each pressure step, the total
hydrocarbon volume is recorded and reported with reference to volume at dew point (relative volume).
(1)

Data such as saturation pressures[dewpoint), gas compressibility factor, Pressure-volume relationship


of hydrocarbon systems can be obtained from CCE tests.

Constant Volume Depletion


CVD is usually carried out on gas condensate and volatile oil reservoir to simulate recovery and
compositional changes as a result of depletion. The test fluid is placed in a visual PVT cell at saturation
pressure and reservoir temperature. For each pressure reduction step[volume increase by mercury
removal,condensate forms and the initial volume is re-established by simultaneously pumping back the
removed mercury to the initial level and ejection of the gas phase. This is done to ensure that the volume
of the cell is constant, thus the next step begins at the same volume as the previous. The gas evolved is
analysed to determine its composition and properties.

Validity and Consistency Checks

Flash Calculations
Flash calculations are used to determine the compositions, amount and properties of each phase present
at a given temperature, pressure and over-all properties of pure components. For a fluid system, the feed
stream is the sum of the number of moles of vapour, V and Liquid, L phases.
(2)

And for each component


(3)

(4)

(5)

Again for equilibrium, the fugacities of each component in vapour and liquid phases must be equal [1],
hence,
(6)

This result in,


(7)

The mole fraction of each component in vapour and liquid phases at any given pressure step will be
defined as,
(8)

(9)
4 SPE-184314-MS

Compositional Material Balance


This is carried out to ascertain the validity of laboratory derived data. Uncertainty resulting from
laboratory measurements and fluid sampling procedures make a questionable fluid model. Where CVD
and separator measurements are available, surface recovery calculations can be carried out without
reverting to an EOS[5]. CVD test is recommended for fluid that show high compositional effect such as
the near critical fluids-volatile oil and retrograde gas condensate. Errors due to negative values in the
number of moles of liquid mole fraction and lack of monotonicity are rectified using the COMB facilities
in PVTi. Assuming one mole of gas at dew point, the cell volume is calculated from real gas equation as[5]:
(10)

From the liquid saturation volume given in CVD test, we can calculate:
(11)

And
(12)

At any pressure step, the total number of moles remaining in the cell will be calculated from:
(13)

On the component basis, the total number of mole of each component is given by,
(14)

(15)

(16)

(17)

(18)

(19)

A plot of log KijPj against the Hoffman characterization factor, F provides the best test of validity for
CVD tests. This graph trend should be parallel without deviations or bumps. The Hoffman characteriza-
tion factor, F is given by:
(20)

Recombination of Separator Data


Where sampling was not carried out earlier at subsurface, the well stream composition is calculated from
mathematical recombination of separator gas and oil. Since separators are usually operated above standard
conditions hence the oil is not stabilized and contains some dissolved gas, Laboratory measurement should
show the separator and stock tank compositions. The best validity test of separator data is the Standing’s
K-values.
(21)
SPE-184314-MS 5

But below the dew point, the fluid is flashed to standard conditions and Standing’s K-values employed
in the determination of vapour and liquid recoveries[6], figure A1.

Figure A1—K- values versus pressure plot

Tuning the EOS Model


EOS models are used in phase behaviours and property prediction of reservoir fluids. Real reservoir fluids
consists of thousands of mixtures interacting in a complex and often with limited identification. Char-
acterized components consists of-the pure components, with well-defined properties (Pc, Tc and AF); the
SCN groups whose molecular weight and specific gravity may be estimated or measured; and the heavy
residue consisting of the plus fraction[7][8][9]. Thus, compositional analyses are unreliable and SCN groups
are not well defined This insufficient component description is responsible for the weak performance of
EOS models as phase behaviour and PVT property prediction tool. Therefore, experiments performed on
real fluids are used to adjust EOS parameters to reasonably reproduce the laboratory measured data.
Because there is no clear-cut and guaranteed method of doing this, this process is time consuming and
largely by trial and error. Even among the strategic techniques[10][13] which promises a procedural
approach are often inconsistent as one method contradicts the other for a particular fluid type over which
they have been found valid, this is due to the uniqueness inherent in reservoir fluids.
Regression
EOS tuning by regression involves the non-linear minimization of an objective function given by
(22)

Which is the summation of the weighted squared deviations, di is the data point to be adjusted,␸(d)pred
and ␸jexp are the predicted and experimental values respectively, wj is the weighting factor indicating the
degree of confidence in the data to be fitted. Data points with wide disparity between the predicted and
experimented values are assigned higher weighting factors in order to force the predicted close to the
experimental. The lower (dmin) and upper (dmax) limits of dj are used to establish the scale of variation
hence,
6 SPE-184314-MS

(23)

Therefore,0 ⱕ dj ⱕ 1.
The derivative of the objective function with respect to dj is a measure of the sensitivity of the F(d) to
dj. Parameters with higher absolute values of the derivative indicates higher sensitivity. Properties of the
plus fraction such as critical pressure, critical temperature, acentric factors and volume shift parameters
are readily available for regression.
The goal of regression is to find the value of dj such that the objective function, F(d) is a minimum.
Differences between measured and calculated values are due to the weakness of EOS and input data, so
regression should be carried out to evaluate the input data and not indefinitely adjust the EOS coefficients
[7]

Model Application
The fluid used for this study was characterized up to C30⫹ with estimates of molecular weight and specific
gravity of the pseudo components, Table 1. Laboratory report of CCE, CVD and Separator tests carried
out on the fluid simple was obtained. Two CCE tests were reported, one at 1300F and 219.10F (reservoir
temperature). The PVTi compositional simulator was used for the study.

Table 1—Fluid Compositions


Components Mole, %

N2 0.1
CO2 0.62
C1 80.99
C2 6.64
C3 3.41
IC4 0.82
NC4 1.05
IC5 0.51
NC5 0.37
C6 0.85
C7 0.74
C8 0.85
C9 0.81
C10 0.4
C11 0.28
C12 0.23
C13 0.26
C14 0.18
C15 0.19
C16 0.12
C17 0.09
C18 0.09
C19 0.07
C20 0.05
C21 0.05
C22 0.04
C23 0.03
C24 0.03
C25 0.02
C26 0.02
C27 0.02
C28 0.02
C29 0.01
C30⫹ 0.04
MOL WT OF C30⫹ 580
SP GR OF C30⫹ 1.01
SPE-184314-MS 7

As expected, the EOS calculated values did not agree with the experimental observations. Also, a
significant under-prediction of the dewpoint was identified [calculated 6288.037psia and measured
6832psia]. To reduce the disparity between the two. A non-linear regression was carried out. The
weighting factor of the dew point and its corresponding parameters were increased to indicate the degree
of confidence in the data and force the EOS to fit the experimental data. The critical temperature, critical
pressure and acentric factors of the pseudo component was selected, due to the high level of uncertainty
associated with its characterization, as regression variables. The upper and lower limits of the regression
variables was adjusted to 50% each. A sensitivity test was launched to determine the properties which
favours the regression. Results show that the dewpoint was sensitive, though not equally, to the three
parameters but due to the need to use fewer variable, the acentric factor was dropped/removed. The
regression reduced the difference in dewpoint value between the calculated and observed to 0.76% which
was quite acceptable. Significant improvement in the match of other properties was observed, figure 1–6.

Figure 1—Observed and calculated Vapour Z-factor from CCE

Figure 2—Observed and Calculated gas density from CCE


8 SPE-184314-MS

Figure 3—Observed and Calculated Relative Volume from CCE

Figure 4 —Observed and calculated retrograde liquid saturation from CCE


SPE-184314-MS 9

Figure 5—Observed and Calculated gas z-factor from CVD

Figure 6 —Observed and calculated retrograde liquid from CVD

Figure 7—Observed and calculated Cumulative moles of liquid produces from CVD
10 SPE-184314-MS

Figure 8 —Observed and Calculated two-phase Z-factor from CVD

Splitting - Extending plus fraction characterization


The complexity of reservoir fluid composition poses a great challenge to detailed and individual
component identification. Although modern laboratory results characterize components up to C30⫹ - in
which components higher than C29 are lumped as C30⫹ -plus fractions and assigned a SCN. The
uncertainty in the properties of the SCN group and pseudo components influences phase behaviours and
property prediction models[2] [3] [8]. This is because the plus fractions is a mixture of several unidentified
components of various properties. Splitting involves a re-extension of the pseudo components into a
number of groups with SCN and property. Whitson[2] [16] presented a method for plus fraction splitting and
property allocation which has gained wide acceptance. Splitting methods[11][12][14] assume that lighter
hydrocarbons such as the gas condensates exhibit left-skewed distribution while heavier ones show
exponential molar distribution[3] [15].
Pseudoization – Component Lumping
This technique is employed to reduce the number of component that are used in the simulation thereby
reducing the computation time. Classical approach is to characterize hydrocarbon mixtures up to hexane
and then lump/combine higher components into one with a SCN[17]. Components are generally grouped
as light (C1 and N2), intermediates (C2 to C6) and heavy fractions(heptane plus). The degree of fluid
description will be influenced by the process being modelled-fewer details for pressure depletion
processes and higher descriptions for miscibility effects. Components with close molecular weight values
will have similar properties. Lumping schemes involves the determination of the number of groups,
grouping criterion selection and assignment of group/SCN properties[4].

Conclusions/Recommendations
1. It is not possible to match all Laboratory observations with equal accuracy. Therefore, a trade
off/decision must be made at the start of the Tuning process on the most important properties based
on the process being modelled.
2. Errors/Inconsistencies in PVT observations spans through the sampling, handling and experimen-
tal sections, or even due to human imperfections, therefore, the status of all facilities and operating
personnel should be duely monitored to ensure consistency with acceptable standards.
SPE-184314-MS 11

3. All PVT data should be validated prior to analysis. Identified errors should be traced and corrected
to avoid undue and indiscriminate regression on component properties which can lead to non-
monotonic models.
4. The result of all EOS techniques should be checked for monotonicity. Too many regression
variables, improper setting of upper and lower limits for regression variable may contribute to that.

Acknowledgement
The authors wish to thank Shell Chair, FUTO and their affiliates for their contributions to the overall
success of this paper.

Nomenclature
Lab Laboratory
EOS Equation of State
CCE Constant Composition Expansion
CVD Constant Volume Depletion
FVF Formation Volume Factor
GOR Gas Oil Ratio
OGR Oil Gas Ratio
SCN Single Carbon number
dj Regression data/point
dmin minimum limit of regression data
dmax maximum limit of regression data
Zi total mole fraction of component i
L number of moles in liquid phase
V number of moles in vapour phase
xi mole fraction of component I in liquid phase
yi mole fraction of component I in vapour phase
Ki Equilibrium ratio for component i
ØiL Fugacity of component I in the liquid phase
Øiv Fugacity of component I in the vapour phase
Vtot Total hydrocarbon volume
Vrel Relative volume
Vdew Hydrocarbon volume at dewpoint pressure
Vvj Volume of vapour at the jth pressure step

REFERENCES
1. Sugiyanto B.S., Luky H., Dwi H.F., Bagus N., Taufun M. (2012). Multiple EOS fluid characterization for modelling
Gas condensate reservoir with different hydrodynamic systems: A case study of Senoro field. Paper SPE 120822
presented at North Africa Technical Conference and Exhibition held in Cairo, Egypt, 20 –22 February.
2. Pichid V., Abraham D., Luis F., Ayala H. (2014). Identification of Pitfalls in PVT Gas Condensate Modelling using
a modified black oil formulation. J Petrol Explor Prod Technol 4: 457–469 DOI 10.1007/s13202-014-0103-4
3. Curtis H. W., Øivind F., Tao Y. (1999). Gas Condensate PVT-What’s Reaaly Important and Why? Presented at the
IBC Conference⬙Optimisation of Gas Condensate Fields⬙, London, Jan. 28-29.
4. Tarek A. (2006). Reservoir Engineering Handbook. Elsevier: Gulf Professional Publishing. Third Edition.
5. Curtis H. W., Stein B.T. (1983). Evaluating Constant-Volume Depletion Data. Journal of Petroleum Technology,
610 –620.
6. Schlumberger Eclipse Reference Manual-An Introduction to PVT Analysis and compositional simulation
12 SPE-184314-MS

7. Julian Y.Z., and Dan Z. (2000). Plus Fraction Characterization and PVT Regression for reservoir fluids near critical
conditions. Paper SPE 64520presentated at the SPE Asia Pacific Oil and Gas Conference and Exhibition held in
Brisbane, Australia, 16 –18 October.
8. Aguilar R. A. and McCain W. D. (2002). An Efficient Tuning Strategy to Calibrate Cubic EOS for Compositional
Simulation. Paper SPE 77382 presented at the SPE Annual Technical Conference and Exhibition held in San Antonio,
Texas, 29 September–2 October.
9. William D. M. (1994). Heavy Components control reservoir fluid behaviours. Paper SPE 28214. September, JPT
(746 –750)
10. Mahmoud T. A., Ahmed H. E. (2015). EOS Tuning: Comparison between several valid approaches and new
recommendation. Paper SPE-175877-MSpresented at SPE North Africa Technical Conference and Exhibition held in
Cairo, Egypt, 14-16 September.
11. Whitson C. H. (1984). Critical Properties Estimation from an Equation of State. Paper SPE/DOE 12634 presented at
SPE/DOE Fourth Symposium on Enhanced Oil Recovery held in Tulsa, April 16-18.
12. Ahmed T. H., Cady G. V., Story A. L., Vidya V., Sahl B. (1984). An Accurate Method for extending the analysis
of C7⫹. Paper SPE 12916 presented at the 1984 Rocky Mountain Regional Meeting held in Casper, WY. May, 21-23.
13. Ali D. (1998). PVT and Phase behaviours of petroleum reservoir fluids. Elsevier: Amsterdam.
14. Hamoodi A. H, Abed A. F., Firoozabadi A. (1996). Compositional modelling of two phase hydrocarbon reservoir.
Paper SPE 36244 first presented at the 7th Abu Dhabi International Petroleum Exhibition and Conference in Abu
Dhabi, UAE, October 13-16.

You might also like