Methods/Algorithms for Bubble/Dew point calculation

PETR 5328
By
Team 13:
Oluwatosin Ogundipe
Samuel Apetuje

Page | 1

Contents
INTRODUCTION........................................................................................................... 3
METHODS.................................................................................................................... 5
Mixture Saturation Points from Cubic Equations of State (3)......................................................5
Bubble Point Pressure Calculations................................................................................5
Dew Point Pressure Calculation..................................................................................... 6
CONVENTIONAL SUCCESSIVE SUBSTITUTION METHOD..............................................7
MVNR METHOD........................................................................................................ 8
SATURATED LIQUID AND SATURATED VAPOR..........................................................8
SELECTION OF ITERATION VARIABLES..................................................................10
DEFINING EQUATIONS.......................................................................................... 11
THE CORRECTION STEP........................................................................................ 12
STARTING ESTIMATES FOR SATURATION PRESSURE CALCULATIONS......................12
CONCLUSIONS AND RECOMMENDATIONS...................................................................17
REFERENCES............................................................................................................. 18

Page | 2

INTRODUCTION
Determination of reservoir fluid properties, such as bubble and dew point cannot be
overemphasized for reservoir engineering calculations such as material balance, inflow
performance, reservoir simulation and reserves estimation. A small error in the calculation of
pressure-volume-temperature (PVT) properties, can lead to very significant errors in a reservoirs
model design, operation and control, such subsequently leads to the development of
inappropriate reservoir simulation model and ultimately inaccurate reservoir performance
prediction. In an ideal situation, we would love to able to always obtain actual measurements of
PVT properties such as bubble and dew points, from laboratory experiments/analysis. However,
such procedures are very costly and time consuming. Alternatively, these properties can be
determined with the aid of empirical correlations or iteratively using an equation of state (EOS).
While empirical correlations are very easy to use, they do not include phase behavior of mixture,
whereas an EOS consists of detailed thermodynamics of the fluid but presents the challenge of
having to provide a good initial estimate.
Bubble point is defined as the pressure and temperature conditions at which the first bubble of
gas comes out of solution in oil (1). Dew point is defined as the pressure and temperature
conditions at which the first liquid droplet forms (2).
Due to the significance of these properties, we have been tasked with determining different
methods/algorithms with which to calculate the bubble/dew point.
The condition for thermodynamic equilibrium between phases is described by the following
mathematical equations:
1.

The overall material balance:

L + V = 1;

(1)

Page | 3

Where L is the moles of liquid and V is the moles of vapor per mole of mixture.
The component material balance:
zi = Lxi + Vyi
(2)
where zi is the mole fraction composition of the initial fluid and xi and yi are the mole

2.

fraction compositions of the liquid and vapor phases respectively.

Mole fraction constraints:

3.

nc

z i=1

(3)

i=1

nc

x i=1

(4)

i=1

nc

y i=1

(5)

i=1

4.

Finally the criteria for thermodynamic equilibrium

f^ li =f^ vi
(6)
l
where f i

v
is the fugacity of component i in the liquid phase, and f i is the

fugacity of component i in the vapor phase.

Combining (1)-(5), and solving for xi and yi yields,
x i=

zi
1+V ( K i1)

(7)
and
y i=

K i zi
1+ V ( K i1)

(8)

Page | 4

K i=

where

yi
xi

From the mole fraction constraints,

y

( i x i)=0

nc

(9)

i=1

This leads to
K
( i1)
zi
=0
1+ V (K i1)
nc

(10)

F ( V )=
i=1

(10) is called Rachford-Rice Equation.

When V0, no vapor is present this means bubble point. (10) becomes,
nc

z i K i=1
i=1

(11)

When V1, no liquid is present. This means dew point. (10) becomes,
nc

Ki =1
i=1

(12)

The methods investigated in this paper include; using cubic equations of state, the conventional
successive substitution method, and the minimum variable newton-raphson (MVNR) method.

Page | 5

METHODS
Mixture Saturation Points from Cubic Equations of State (3)
Bubble Point Pressure Calculations

1. Provide an estimate of the bubble point pressure

2. Estimate the K-factors using Wilsons Correlation
T
Pc ,i exp 5.37(1+ i)(1 c ,i )
T
K i=
P

(13)
Where,Tc,i is the critical temperature of component i
Pc,i is the critical pressure of component i
i is the acentric factor of component i
j +1
j
3. Estimate vapor phase composition from y i =z i K i , i = 1,2,,nc, where j is an
iteration counter
v
4. Calculate the vapor and liquid phase fugacity coefficients [( i , i = 1,2,,nc) and (

li , i = 1,2,,nc)] using current estimates for bubble point pressure and vapor phase
composition. The liquid composition equals the feed composition.
5. Calculate new K-factors from,
L
V
ln K i=ln i ln i

(14)

nc

6. Evaluate

F= z i K i1

7. Evaluate

ln Li
P
V
ln i
z i K i (
)
P
n
dF
=
dP i=1

(15)

i=1

(16)

8. Calculate the (j+1)th estimate of the bubble point pressure from

Page | 6

P j +1=P j

F
j
dF
dP

(17)

9. If not converged return to 3

Dew Point Pressure Calculation

1. Provide an estimate of the dew point pressure

2. Estimate the K-factors using (13)
3. Estimate liquid phase composition from

x ij+1 =

zi

j
K i , i = 1,2,,nc, where j is an iteration

counter
v

4. Calculate the vapor and liquid phase fugacity coefficients [( i , i = 1,2,,nc) and (

li , i = 1,2,,nc)] using current estimates for dew point pressure and vapor phase
composition. The liquid composition equals the feed composition.
5. Calculate new K-factors from (14)
n
z
F=
Ki 1
6. Evaluate
i=1
i
c

7. Evaluate

ln Vi
P
z i ln iL
(
)
Ki
P

(18)

(19)

nc

dF
=
dP i=1
8. Calculate the (j+1)th estimate of the dew point pressure from (17)
9. If not converged return to 3
Bubble and dew points may also be calculated for a specified pressure, that is, temperature is the
unknown parameter.
These calculations are complicated by the fact that it is not generally known in advance if the
saturation points of the mixture actually exist at a specified temperature and pressure. For
example, a bubble point calculation for a temperature higher than the critical temperature would
Page | 7

not converge because the bubble point line ends in the critical point. Also, the mixture may have
two dew points as in the case of gas condensates. This may cause convergence problems in the
calculations. At best, either the upper or lower dew point will be located with this method.
This method converges easily from low to medium pressure than at high pressures.

CONVENTIONAL SUCCESSIVE SUBSTITUTION METHOD

The phase equilibria equations that incorporate the most commonly used equations of state
(Soave-RedlichKwong and Peng-Robison) in the petroleum engineering industry, are nonlinear
and must be solved iteratively. The method of successive substitutions is one of the most widely
known and used methods in saturation point calculations involving EOS. Here we give an
illustration of the algorithm.
1. Estimate initial value of Ki from (13)
2. Compute F(V=0) and F(V=1) (10). If F(0) > 0 and F(1) < 0, two phases exist and
equation 10 is solved using Newton Raphson for V. If F(0) < 0 or F(1) > 0, then most
likely a single phase is present, in which case a value of V= 0.5 is assumed and the
calculation done.
3. Calculate xi and yi respectively from (7) and (8).
l
v
4. Compute f i and f i from derived expressions for fugacity, and evaluate
f li
Fi = v
fi

and

K i=

Fi y i
xi
nc

5. Check for convergence using the criterion

( F i1 )2
i=1

7
, ( =10 )

(20)

If (20) is satisfied stop, if not proceed to step 6

6. Compute F(0) and F(1)
If F(0) < 0 set V= 0, i.e single phase liquid
If F(1) > 0 set V = 1, i.e single phase vapor
Page | 8

If F(0) > 0 and F(1) <0 solve (10)

7. Update xi and yi from (7) and (8) respectively, then go to step 4
In the single phase region, where V= 0 or V =1, the Fis may converge to a value different
from unity. Nghiem et al(4) suggested iterating until (20) is satisfied even in single phase
cases.
This method however, almost always diverges for bubble and dew point calculations near
the critical region; a region of high importance when studying multiple-contact, miscible
oil recovery methods involving CO2 or rich-gas injection because the path of the over-all
fluid mixture passes through this region. Also, this method almost always diverges for
some mixtures near their saturation pressure at conditions way from the critical region.

MVNR METHOD
SATURATED LIQUID AND SATURATED VAPOR

At a given temperature the pressure at which an infinitesimal amount of vapor first appears is
known as the bubble point pressure. Thus, for a saturated liquid the overall material balance is
simply
L=1

(21)

and the component material balances are

xi = zi ;

(22)

Similarly, at a given temperature, the pressure at which an infinitesimal amount of liquid first
appears is known as the dew point pressure, and for a saturated vapor the overall material
balance is simply
V=1

(23)
Page | 9

and the component material balances are

yi = zi ;

(24)

The saturation (bubble and dew point) pressures are related to the fugacities of each component
in the liquid and vapor phases by

saturation pressure ps =

f li
()/( vi )
n

, for saturated liquid

(25)

, for saturated vapor

(26)

i=1

and

ps =

f vi
()/( li)
n

The

i=1

MVNR method was first presented by Fussel and Yanosik(5) as a robust algorithm for solving the
non-linear system of equations describing thermodynamic equilibrium using the RK EOS. The
l

single stage separation equations 1, 2, 3, 5, 6 and the expressions for f i

and f i

were

considered, which make up a 4n + 2 equations which are to be solved simultaneously in order to

determine L, V and n liquid phase compositions, xi, n vapor phase compositions, yi, n liquid
l
v
phase fugacities, f i , and n vapor phase fugacities, f i . These equations are non-linear and

have to be solved iteratively. One way of handling the iterative sequence is to correct all 4n +2
unknowns at each iteration, which would be very time consuming. The minimum variable
Newton-Raphson (MVNR) provides an efficient way of organizing the iterative sequence, as it
reduces the size of the correction step by eliminating as many unknowns as possible from the

Page | 10

correction step. This is achieved by breaking the unknown variables into 2 groups. The first
group contains the iteration or independent variables which are the unknowns to be corrected,
and the second group contains the dependent variables. One equation must be used to define each
of the dependent variables, and all these equations are collectively referred to as the defining
equations.
The remaining equations which are equal to the number of iteration variables are known as error
equations. This yields a four-step iteration process for MVNR which are
1)
2)
3)
4)

Select the iteration variables and assume values for them

Calculate the dependent variables using the defining equations
Calculate the error using the error equations, and
Correct the iteration variables using the Newton-Raphson method.

The solution convergences when the error in Step 3 is less than an acceptable error tolerance, .
If not, the process is repeated.
SELECTION OF ITERATION VARIABLES
l
For bubble-point liquid, (6) and (25) along with f i

v
and f i make up a set of 3n + 1

equations that are to be solved simultaneously to determine the bubble point pressure, ps, n vapor
l

phase compositions, yi, n liquid phase fugacities, f i

and n vapor phase fugacities, f i . The

iteration variables required by MVNR is n + 1; pressure p, and n vapor phase compositions. This
is referred to as p-y iterations.
For dew point, similarly there are 3n +1 equations to be solved simultaneously to determine the
l
dew point pressure , ps, n liquid phase compositions, xi, n liquid phase fugacities, f i

and n

Page | 11

vapor phase fugacities, f i , with n + 1 iterations variables (just like for bubble point liquids)
which are pressure p, and n liquid phase compositions. This is referred to as p-x iterations.
Initial estimates for p and yi and p and xi for p-x and p-y iterations respectively are required.
Fussel and Yanosik(5) used the component K values calculated by the empirical correlation (6)

exp 5.37 ( 1+ i ) 1
K i=

1
T ri

)]

pri

(27)
to flash the overall fluid mixture at the system temperature and an estimated saturation pressure,
yielding liquid and vapor compositions. Depending on the iteration used (p-x or p-y), the
appropriate phase compositions are used as initial estimates of the iteration variables, xi and yi.
The other phase composition is equal to that of the fluid mixture (22).
DEFINING EQUATIONS

Next is to solve the defining equations for the dependent variables. The iterative steps for p-x and
p-y are,
1) Calculate by solving the Redlich-Kwong EOS, the compressibility factors for the liquid
and vapor phases ZL and Zv
l

2) Calculate n times(i.e for each of the n-components) f i

and f i

Error Determination
The error equations representing the residual of the thermodynamic phase equilibria equation (6)
is written as
i=f lif vi

(28)

Page | 12

As stated previously the error equations are those remaining after selecting the defining
equations
For the p-y iteration, there are n + 1 iteration variables. (30) represents n error equations and the
remaining error equation is the residual of (25), which is written as
n

n+1= p s
i=1

f li
iv

(29)

Similarly, the additional error for the p-x iteration is

f vi
n+1= p s l
i=1 i
n

(30)

THE CORRECTION STEP

The correction step is necessary to obtain the improved estimates of the iteration variables that
will reduce the residuals calculated by the error equations. The correction step reduces to a
matrix problem with an order directly proportional to n + 1 for the p-x and p-y iterations.

Even though the MVNR method has proved to be a good upgrade on the traditionally known
successive substitution method, experience has shown that the MVNR still requires for an
improve estimate of initial iteration parameters ( i.e pressure and phase compositions) for the
bubble and dew point pressure solutions. It was observed that the K-value correlation used for
the initial estimates usually gave satisfactory results when the fluid system contained significant
amounts of heavy components ( C7+), but was often unsatisfactory when dealing with fluid
systems containing only light components, for example fluids in dry gas-rich zones or CH4, CO2
or N2 rich mixtures.

Page | 13

STARTING ESTIMATES FOR SATURATION PRESSURE CALCULATIONS

In addition to the above stated shortcomings of the MVNR method, the correlation cannot be
used to determine all the K-values required for more complex CO2/hydrocarbon systems, such as
those in which three or more phases form.(7)
For these reasons Baker and Luks(8), employed a more self-consistent method of estimating
starting values of the iteration parameters, in which the EOS is used to obtain both the initial
estimates of saturation pressure and phase composition as well as the final converged solution.
The method is based on approximation of the phase compositions at progressively higher
pressure until a pressure region is reached in which the saturation pressure conditions are
satisfied. This pressure and the corresponding phase compositions are then used to start the
MVNR solution. Employing the fugacity criteria for phase equilibrium, (6) and noting that
l

f i =x i p i

(31)

f iv = y i p vi

(32)

and

For a bubble point calculation the xis are known and the vapor phase composition is
l

y i=

p i x i
p iv

( )
i

vi

xi

(33)

For a dew point calculation the yis are known and the liquid phase composition is given by
x i=

p li y i
p vi

li

( )

vi

yi

(34)

Initially both the pressure and second phase compositions are unknown. The estimation involves
the following steps,

Page | 14

1) Choose a low staring pressure ( e.g. 1 psi)

v
l
2) Compute i and i at the pressure of the new iteration level using the composition

l
from the old iteration level. For the first iteration an approximation of i

v
and i

( for dew point and bubble point respectively) must be used.

3) Evaluate the unknown phase composition using (33) for bubble point or (34) for dew
point. If the sum of the calculated phase compositions is less than 1 then the assumed
pressure is greater than the saturation pressure. If otherwise, the assumed pressure is less
than the saturation pressure.
4) Correct the pressure with the direction indicated in step 3. Baker and Luke(8)
recommended an increment of 100 psi to 300 psi for each iteration until the saturation
pressure is exceeded. At this stage, the half interval type method may be used to narrow
the search range.
5) Correct the unknown phase compositions by normalizing. Repeat step using these
corrected compositions to recalculate the fugacity coefficients at the new estimated
pressure.
As Fussel and Yanosik(4) have shown the MVNR method generally results in more rapid
convergence once the region of the saturation pressure is estimated reasonably well. Thus the
iteration above is stopped once the saturation pressure is within a reasonable range (say 20 psi).
The resulting phase compositions and pressure are then used as initial estimates for the MVNR
l
method. As mentioned in Step 2 above an estimate of i

v
and i must be made for the first

v
iteration. Using i =1 , is a good estimate at low pressures, for a bubble point calculation. For

a dew point calculation, i

can be estimated as
Page | 15

ZL
[ ( p/ psi 1) ]
l
o
i= i exp

(35)

o
where i is the pure component fugacity coefficient evaluated at the pure component vapor

pressure, psi, which can be estimated from the Clausius-Clapeyron equation(9) as

psi = pci exp [

H vi
T ci
1
]
T ci
T

( )(

(36)
The procedure described above for dew points also can be applied to calculate lower dew points,
below which only vapor exists. The only modification required is in Step 4, where
n

x i <1
i=1

indicates the pressure is less than the lower dew point pressure, and

x i >1
i=1

indicates that the estimated pressure is too high.

It is worth noting that a variation of this technique is applicable to estimation of phase
compositions for a flash calculation in the two phase region.
The procedures described above use only the equation of state for iteration, the fugacity
coefficients are valid for any two-phase region. Also, the same EOS used to obtain the estimated
pressure and phase compositions, is used to calculate the final solution, thus fewer MVNR
iterations are required. Even though this method requires more CPU time to obtain the initial
estimates of phase compositions than the K-value used by Fussell and Yanosik(4), the fewer
MVNR iterations results in an overall faster and, most importantly, more reliable calculation
method.

Page | 16

T, xi

Initial guess for P

New value of
v
i

Initial guess for

v
i

= (T , P , x)

iv

li =li (T , P , x)

ye
s

y i=

l
i
v
i

( )

no

xi
Page | 17

First
iteration
no

yi =

yi'=yi ?

yI = 1
ye
s

Correct Calculated Bubble point and

CONCLUSIONS AND RECOMMENDATIONS

In determining the saturation points (bubble and dew point pressure) of hydrocarbon fluids, the
use of equations of state offer a more convenient and time saving path as opposed to having to
rely squarely on expensive and laborious laboratory experiments. However, as we have discussed
in this paper, the methods available through the use of EOS have issues unique with whatever
technique one decides to adopt. The Conventional Successive Substitution Method has been
known to exhibit, very slow convergence rate at high pressures and possible divergence near the
critical point, for bubble and dew point calculations. The MVNR method, which relies on a
minimization of the number of unknowns and converges for near critical fluid mixtures, presents
a useful alternative to the CSS method as it also converges faster reducing CPU time. Still, the

Page | 18

MVNR method requires improvement on the estimation of its initial iteration parameters
(pressure and phase compositions), as discussed in the body of this paper.
In deciding what method to employ in bubble in dew point calculations, the following should be
considered;
a. CPU time and expense required
b. Sophistication of computer software
c. The fluid mixture to be analyzed especially in miscible oil recovery processes
There exist computer soft-wares that can switch from say CSS method to MVNR or
another preferred method, if CSS method is exhibiting very slow rate of convergence or
begins to diverge near critical regions.

REFERENCES
1. Schlumberger Oilfield Glossary, http://www.glossary.oilfield.slb.com/en/Terms.aspx?
LookIn=term%20name&filter=bubble%20point, November 5, 2013.
2. Elementary Principles of Chemical Processes, Richard M. Felder, Ronald W. Rousseau,
Pg. 259 260.
3. Phase Behavior of Petroleum Reservoir Fluids, Karen S. Pedersen, Peter L. Christensen,
Pg. 119
4. Nghiem, L.X., Aziz, K. and Li, Y.K.:"A Robust Iterative Method for Flash Calculations
Using the SoaveRedlich-Kwong or the Peng-Robinson Equation of State," SPEJ, June
1983,521-530.
5. Fussell, D.O. and Yanosik, J.L.: "An Iterative Sequence for Phase Equilibrium
Calculations Incorporating the RedlichKwong Equation of State," Soc. Pet. Eng. J. (June
1978) 173- 182.
6. Wilson, G.M.: "A Modified Redlich-Kwong Equation of State, Application to General Physical
Data Calculations," paper presented at the 65th National AIChE Meeting, Cleveland, 1969.

7. Shelton, J.L. and Yarborough, L. Multiple-Phase Behavior in Porous Media during CO2
or Rich-Gas Flooding J. Pet Tech (Sept. 1977) 1171-1178.
Page | 19

8. Baker, Lee E. and Luks, K. D., :" Critical point and Saturation pressure Calculations for
Multipoint systems," SPEJ 1980, 15-24.
9. Himmelblau, D.M.: Basic Principles and Calculations in Chemical Engineering, Prenticehall Inc., Englewood Cliffs, NJ (1962) 242.

Page | 20