Sie sind auf Seite 1von 0

MULTIPHASE FLOW THROUGH CENTRIFUGAL PUMPS

Dr. Rajesh Sachdeva, Dr. D. R. Doty


and
Dr. Z. Schmidt
MULTIPHASE FLOW THROUGH CENTRIFUGAL PUMPS
bY
DR. RAJESH SACHDEVA, University of Tulsa (presently with Simulation Sciences Inc.)
DR. D.R. DOTY AND DR. 2. SCHMIDT (both of University of Tulsa)
ABSTRACT
Recently, there has been considerable interest in the
petroleum industry on multiphase flow through pumps.
Pumping gassy fluids has applications in both pipeline
and electric submersible pumps. The present study
discusses a dynamic and a correlational model for three
electric submersible pumps (ESPs).The dynamic model
is applicable to pipeline centrifugal pumps (since the
flow physics is exactly the same) provided a correlation
for bubble size is known.
INTRODUCTION
Centrifugal forces dominate gravitational forces in
multiphase impeller flow. As such, it has been shown by
investigators that the diffuser performance can be
ignored and impeller behavior determines centrifugal
pump behavior. The basic model presented in this paper
is develoiped for electric submersible pumps (ESPs) but
is equally applicable for gas-liquid flow through any
centrifugal (radial or axial-type) pumps. This paper
heavily draws on parts of papers presented elsewhere by
the same authors*3*4*5 . The interest in developing an
offshore multiphase pump has
been mainly due to major economic benefits associated
with laying a single multiphase pipeline (as opposed to
one gas and one liquid pipeline). Development and
modelling multiphase pipeline pumps is becoming more
important as the oil companies are venturing further into
the sea for oil. Data from an Amoco-Centrilift study were
used to validate the model developed. There have been
quite a few studies in the nuclear industry3 on modelling
impellers. It has been shown that none of the models of
the nuclear industry can be used for ESPs in gassy
wells, This is because the nuclear industry models are
pump specific, applicable to substantially lower void
fractions, and do not consider the effects of inlet
pressure. The development of a general model for
multiphase flow through centrifugal pumps would be a
complicated task primarily because of complex pump
geometries. A two-dimensional multiphase pump model
would require the knowledge of phasic holdup and
velocity profiles in a pump and this data is not at all easy
to measure. The results show very encouraging
performance prediction for both the radial and axial
pumps. Limited success was also obtained in correlating
the pump pressure increase (Model 2). Performance of
axial and radial pumps is also compared.
PREVIOUS WORK
Most of the work on multiphase flow through ESPs has
been done in the nuclear industry. A comprehensive
review is given by Sachdeva3. The nuclear industry
models cannot be used because of reasons cited earlier
and will not be discussed.
2 MULTIPHASE FLOW THROUGH CENTRIFUGAL PUMPS
The work in the nuclear industry can be summarised into
three main phases:
4
b)
c)
Black Box Methods: These were largely
unseuccessful. Investigators realised thatat least a
qualitative idea of the flow physics inside the pump
was required.
Experimented Studies: Various U.S. and Japanese
investigators3 qualitatively studied the movement of
the gaseous phase within the impeller. It was found
that bubbly flow caused lower head degradation than
slug/churn-turbulentflow. Investigators also found
that the gas slug growth around the impeller eye led
to unstable flow towards the left-hand side of the
pump (rate-head) curve.
Analytical Models: basic one-dimensional, multiphase
pump models were developed by investigators. The
most motable works wete these of Zakem6 and
Furnya The basic model was the same in both
cases but the solution methods were slightly
different. Both ignored effects of pump inlet pressure
and gas compressibility. Neither had good success
in correlating with experimental data.
Studies in the petroleum industry include:
(a) Lea and Bearden: This study consisted primarily of
gathering data for the K-70, I-42B and C-72 pumps
and a demonstration that under certain conditions,
gas through an ESP may be advantageous because
of a gas lift effect. The present study uses the data
gathered by Lea and Bearden.
(4
Tarpley4: Here, the author presented an extremely
simplistic model for pump head degradation in an
ESP. The energy required to compress the gas was
taken to be the only factor contributing to head
degradation. As the results of the study show, this
energy has a negligible effect on the pump head
degradation. The method did not appear to work for
inlet gas void fractions over 2%. Note that below 2%,
there is hardly any pump head degradation and thus
this model is of extremely limited use.
Limited amount of Russian literature3 is also available on
the subject. Although, the Russians were among the first
to study the problem, their literature seems to be the
least advanced. None of their models involved the study
of the actual flow physics in a pump.
MODEL FORMULATION
Modelling single phase flow through ESP, can easily give
50% errors, when approached within a one-dimensional
framework3. For comparison, errors for single-phase flow
in tubulars are around l-2%. The pump geometry and a
rotating impeller are the primary complicating factors for
modelling two-phase flow through ESPs. Furthermore, in
multiphase flow, the effects of impeller slip (different from
interphasic slip to be discussed later) are unknown. In
single phase flow, the impeller slip causes the velocity
triangle to deviate from the ideal velocity triangle as
shown in Figure (1). The net effect of this deviation is a
reduction in useful head produced. Also, individual
pump losses (mechanical, hydraulic, etc.) have not yet
been adequately quantified even for single-phase flow.
Given the minimal amount of knowledge for single-phase
flow through ESPs, formulation of a simplified multiphase
model becomes necessary. This avoids assumptions that
cannot be corroborated.
Sachdeva, through numerous runs, has shown that
diffuser performance can be neglected. Similar
conclusions were also reached by Pate1 and Runstadler,
Runstadler and DoIan and Hench and Johnston.
Since the data shows that the impeller dominates the
pump performance, the diffuser performance is ignored.
The basic procedure for a pump model is as under:
(a) develop a simple, one-dimensional liquid-only model.
Since most losses are not considered, this model will
not match the actual curves
(b) formulate a similar one-dimensional, multiphase
model
(c) the difference between (a) and (b) will represent the
additional head degradation due to free gas in the
pump. This difference is subtracted from the liquid-
only curve published by the manufacturer. This
approach is represented in Figure (2).
IDEALISED MULTIPHASE CURVE
This section deals with the development of the curve (B)
shown in Figure (2). Noemenclature used is shown in
Figure (3). The l-42, K-70 and C-72 pumps are shown in
Figure (4).
Flow is assumed to be idealised, i.e., no recirculation, no
impeller slip (different from interphasic slip), etc. Diffuser
3
performance is ignored for reasons discussed earlier. A
two-fluid approach is used for multiphase equations.
Continuitv Equations
Along a streamline z, parallel to the impeller blades, the
phasic continuity equations for gas and liquid are given
by:
M,
= P&4
ML = pr W, (1 -a) A,
Differentiation yields:
I dA, 1 da
--+--
I dWg
1 dps 0
A, dz a dz +Tqdz+p,dz=
I dA, 1 da
--
dW=()
---+--
A, dz (1 -a) dz W, dz
. ..(I)
. ..(2)
. ..(3)
. ..(4)
These equations are valid for both bubbly and churn-
turbulent flow regimes.
Momentum Equations
One-dimensional phasic momentum equations for steady
state flow are given by Wallis* as:
Yl
PLVrf ar
=q+WL-; +$mL
. ..(5)
av,
PA ar
. ..(6)
Here, Zb, and Zb, are the body forces and Zf, and Zf,
are leftover or balancing forces.
In an axial pump, the following geometrical relationship is
true:
d = sir-$?(f) cosy(f)
dz
. ..(7)
Based on Equation (7) the following can also be shown
to be true for an axial pump:
h
= W,sin#?(f) cosy(f)
. ..(8)
and,
4
= A@-$(/) cos&)
. ..(9)
Thus, the relationship between various velocity
components can be easily determined at any point along
the impeller path. Note that the angle y=O for a flat
radial pump.
The body forces due to the impeller can be represented
as:
rbg = p, Q2f
. ..(lO)
and
xb, = pL Q2f
. ..(I 1)
The expressions for body forces are not dependent on
the flow regime in the impeller. The rest of the section
deals with the computation of the balancing forces.
If bubbly flow is assumed to exist in the impeller, the
drag forces that increase the gas-liquid velocity lag can
be represented as3*:
f
drag& = - CD
a
(1 -a)2.78
and
f
draga
=c,
(1 -a).
PIYL
. ..(13)
The drag forces reduce useful head produced and the
above expressions account for the effects of bubble
swarm (particle-particle forces). For churn-turbulentflow,
the term C-,/r,, in Equations (12) and (13) is usually2*3
replaced by a function of (l-a). This reflects reduced
value of C-Jr,, implying vastly increased gas-liquid
velocity lag. Thus, in churn-turbulentflow, the liquid
4
MULTIPHASE FLOW THROUGH CENTRIFUGAL PUMPS
phase accelerates more causing the useful pump energy
to be wasted as liquid velocity head.
The apparent mass forces exist for bubbly flow and tend
to reduce the gas-liquid velocity lag. In churn-turbulent
flow, these forces do not exist by definition. However, we
have retained these forces in the model since the
transition point between bubbly and churn-turbulentflow
is not known in the impeller. The retention of these forces
in the churn-turbulentflow regime causes a slight error.
This error is absorbed by the Cdr,, correlation. The
advantage of doing this is that the bubble to churn-
turbulent flow pattern transition boundary within the
impeller need not be known.
The liquid and gas apparent mass terms are given as:
f
amg
= -C/3& -$ Kl
- V,)
and.
f
amc = C(y- ,aa)PYg 2 (Kg - V,)
In absence of any information, a spherical bubble shape
(C=OS) is assumed.
The frictional forces for each phase are calculated as
suggested by Craveri and Wallis and given in
Sachdeva3.
Equations (5) (1 l), (12) and (15) collectively represent
the momentum equation for the liquid phase. The gas
phase is represented by Equations (6), (IO), (13) and
(14). The approximate frictional term is added as
explained in Reference (1). The necessary geometrical
relationships for the axial impeller is given by Equations
(7) and (9).
For liquid-only flow, either Equation (5) or (6) can be
shown to reduce to:
g dP = Q2r
--
PL df
- + $3
. ..(16)
The relationship between r and z requires knowledge of
the geometry of the pump. Since about 100 steps are
required to integrate along the impeller streamline, a
linear relationship between inlet and outlet /?-angle is
assumed.
Note that /I is the angle between the velocity
components U and V. Thus,
where
. ..(17)
. ..(18)
The y-angle is constant for the axial K-70 pump. With the
above assumptions Equation (7) can be expressed as:
. ..(19)
Integration of Equation (19) yields the following
relationship between r and z valid for an any pump:
z=F logtan
[
W-f,) +B,
-logtan
w, -fJ +B,
2 2
1
420)
Note, y=O for a radial pump.
Equation of State
The liquid is assumed to be incompressible. The gas
phase is assumed to behave adiabatically:
p I
- =c
Y
Pg
.I ..(21)
Integration of the above curve will yield the curve (A) in
Figure (2).
5
Differentiation yields:
dP y-l G?J
-& - c 'Y&a -& =
0
. ..(22)
Note that the curve (B) in Figure (2) can be obtained
from the equations of continuity, momentum and state
once the appropriate value of Cdr,, is known.
MODEL SOLUTION
The solution vector [dw,/dz, dw$dz, dp$dz, da/dz,
dP/dz] and is solved along each point of the impeller
and for each flow rate. Since C&, is unknown, trial-and-
error runs yielded C,Jrb value for each data point. This
was then correlated in the following form:
CD pi,
-= K-
rb a: OF
. ..(23)
Based on 326 diesel-CO, data points2*3, the following
values were obtained for the axial K-70 pump:
K = 9.53 x lO-4 El = 3.33
E2 = 2.83 E3 = 5.92
For comparison, the radial C-72 pump of similar size as
the axial K-70 pump gives (173 points):
K = 6.65 x 10 El = 5.21
E2 = 5.22 E3 = 8.94
Similarly, for the l-42 pump (287 points):
K = 5.7 x 10 El = 2.36
E2 = 6.64 E3 = 5.87
The regression analysis coefficients for the correlations
for l-42, K-70 and C-72 pumps were respectively 92%,
94% and 90%. Diesel-CO, data were used for all cases.
In absence of stage-to-stage pressure rise data, a
linearly-averaged stage was correlated for.
RESULTS OF THE DYNAMIC MODEL
Some of the model trends are given in Figures (5) thru
(16). Figures (5) through (8) show the model
performance for the axial K-70 pump and Figures (9)
through (12) for the radial C-72 pump (of similar size as
the K-70 pump), and Figures (13) thru (16) for the l-42
pump. The error analysis for the dynamic model is
shown in Table (1). For comparison purposes the
pressure rise per stage (at bep) for each pump is also
included. The dynamic model predicts the head
degradation well.
The predictions for phasic velocities, void fractions, etc.
are consistant with the observations of the photographic
studies discussed in the earlier study. Both pressures
and gas densities increase along the impeller path as
does the void fraction, This causes the ratio w,.& to
decrease drastically towards the impeller exit as shown
in Figure (17). The model thus shows acceleration of the
liquid phase and the loss of useful head as liquid
velocity head. For the radial or the axial pumps, for
r,=O.l mm or so, there is hardly any head degradation
predicted. Also, at large void fractions and lower inlet
pressures, the gas phase velocity is drastically reduced
and the gas phase almost stalls causing surging and
eventually gas locking. All of the above predictions agree
with the photographic evidence.
The gas-liquid slip is essential to quantifying multiphase
head degradation. Under assumotions of no aas-liauid
slip, the model will not predict anv head dearadation.
CORRELATIONAL MODEL (Model 21
The main disadvantage of the dynamic model is that it is
complicated to solve for. To overcome this, approximate
correlations were developed correlating the pressure
increase per stage, pump inlet pressure, pump inlet void
fraction and the liquid flow rate. Various parameters were
tried and the best results were obtained from the
following:
AP = K (P$ (ain)= (QJ
Here AP is in psi per stage, P, is the pump stack inlet
pressure in psig, a, is the pump stack inlet void fraction
(not percent) and Q, is in gallons/min. For the radial I-
42B pump (287 points):
K = 1.154562, El = 0.943308, E2 = -1.175596 and E3 = -1.300093
For the radial C-72 pump (173 points):
K= 0.1531026, El = 0.875192,E2 = -1.764939, E3 = -0.918702
Finally, for the axial K-70 pump (326 points):
K = 0.0936583, El = 0.622180, E2 = -1.350338, E3 = -0.317039
6 MULTIPHASE FLOW THROUGH CENTRIFUGAL PUMPS
The regression analysis coefficients for the l-42, C-72
and K-70 pumps were 70.63%, 87.66% and 85.12%
respectively. All data are diesel-CO, data.
In predicting the pressure increase in ESPs, this
correlation performed substantially worse than the
dynamic model.
Table (1) compares errors from the dynamic and the
correlational models. Excepting for the C-72 pump, errors
from the correlational model (Model 2) are about 2 to 3
times those obtained from the dynamic model. However,
the correlational model is extremely easy to use and an
engineer can use this model for approximate estimates.
PUMP DESIGN (AXIAL verses RADIAL) AND
ORIENTATION
The axial K-70 and the radial C-72 pumps are similar in
dimensions and their performance will be compared
here. In general, the axial pump suffers less degradation
than the radial pump. This trend can be seen in Figures
(5) through (12). The reader can compare Figures (5)
and (9) Figures (6) and (10) and so on until Figures (8)
and (12) and see that in general, the axial pump
performs better. The dynamic model predicts this trend.
Another useful parameter to compare is the C,,/rb value
for the K-70 and C-72 pumps. For the same values of a,
P, and Q,, the value of Cdr,, is lower for the radial C-72
pump compared to that for the K-70 pump. This implies
that the gas-liquid velocity lag will be lower in the axial K-
70 pump. Thus, the liquid phase will be accelerated less
in the K-70 pump and less head will be lost as velocity
head. This explains why the radial pump of comparable
size performs worse than an axial pump. The higher
C,Jrb value is also reflective of a pumps tendency to
have bubbly rather than churn-turbulent flow. Pump
manufacturers should strive for hiaher G/Jr, values. This
can be done by avoiding or breaking up the churn-
turbulent regime. Reference (14) is an example of a
patent application for a two-phase pump that tries to
break up the bubbles and mix a churn-turbulentflow
regime into a bubbly regime.
Note that the ratio of centrifugal to gravitational forces is
about 150 or more even in ESPs with diameters of 2-4
inches. This ratio will obviously be even higher for
pipeline pumps. This suggests that (i) the pump
orientation (vertical versus horizontal) does not effect the
model, (ii) The model can be used to model the
performance of multiphase pipeline pumps currently
under development. Indications are that C,,/r,,will be the
only correlating parameter required.
CONCLUSIONS
1)
2)
3)
4)
5)
Dynamic model is developed and successfully tested
against data from three centrifugal pumps of both
radial and axial geometries.
Flow physics in impellers is extremely complex and
simplistic, correlational approach (eg. Model 2) will
have only limited success.
C,,/rb seems to be the only parameter required to
complete the dynamic model for gas-liquid flow
through any centrifugal pump.
The model agrees with the photographic studies in
terms of:
a) void fraction distribution in an impeller
b) gas-liquid velocity ratio behavior
c) explaining pump instability at lower liquid rates
(left of the best-efficiency point, bep)
The model gives the manufacturers a tool to design
multiphase centrifugal pumps. Conventional single-
phase flow dynamics is obviously totally inapplicable
to model multiphase centrifugal pumps.
NOEMENCLATURE
Symbol Description
A
b
c
CLY
El,E2,E3
f
i
M
;
Q
r
u
V
W
Area, @
body forces
apparent mass coefficient, dimensionless
drag coefficient, dimensionless
exponents
leftover forces
gravitational constant
constant
mass rate (Ibm/s)
exponent
pressure (psi)
flow rate
radial coordinate
peripheral velocity (Ws)
absolute fluid velocity (ft/s)
fluid velocity relative to the impeller (Ws)
GREEK
Symbol Description
e
;
a
Y
density, Ibm/ft!
angular impeller velocity, radians/s
blade angle, degrees
void fraction (fraction)
blade angle (r-z plane), = 0 for radial
pumps adiabatic exponent
SUBSCRIPTS
Symbol Description
1
2
am
drag
Y
r
Z
inlet
outlet
apparent mass
drag
gas
liquid
radial coordinate
streamline coordinate (parallel to blades)
ACKNOWLEDGEMENTS
The author wishes to thank the Society of Petroleum
Engineers for allowing copyright release of the material
of this paper previously presented in SPE papers
SPE22767 and SPE24328. Thanks are due to Ms. Vira
Estrada for preparing the manuscript.
REFERENCES
1.
2.
3.
Sachdeva, R., Doty, D. R. and Schmidt, Z.:
Performance of Electric Submersible Pumps in
Gassy Wells, SPE 22767, presented at the 66th
Annual Technical Conference and Exhibition of SPE,
Dallas, TX, Oct. 6-9, 1991, accepted for publication.
Lea, J. F. and Bearden, J. L.: Effect of Gaseous
Fluids on Submersible Pump Performance, JPT,
December 1982 and SPE 9218.
Sachdeva, R.: Two-Phase Flow Throuqh Electric
Submersible Pumps, Ph.D. dissertation, University of
Tulsa, Tulsa, OK, 1988.
4.
5.
6.
7.
8.
9.
10.
11
12.
13.
14.
Sachdeva, R. et. al.: Performance of Axial Electric
Submersible Pumps in a Gassy Well, SPE24238,
presented at Casper, WY, May 18-21, 1992.
Sachdeva, R.: Understanding Multiphase Dynamics
in ESPs for Better Multiphase Pump Design
presented at the Electric Submersible Pump
workshop in Dallas, TX, 1992.
Zakem, S.: Determination of Gas Accumulation and
Two-Phase Slip Velocity in a Rotating Impeller,
Journal of Fluids Engineering, Vol. 102, 446-455,
(December, 1980).
Furuya, 0.: An analytical Model for Prediction of
Two-Phase (non-condensable) Flow Pump
Performance, Journal of Fluids Engineering, Vol.
107, 139-147, (March, 1985).
Stepanoff, A. J.: Centrifuaal and Axial Flow Pumps,
published by John Wiley and Sons, (1957).
Patel, B. R. and Runstadler, P. W.: Investigations
Into the Two-Phase Behavior of Centrifugal Pumps,
ASME Symposium on Polyphase Flow in
Turbomachinery, (December lo-15 1978) San
Fransisco.
Runstadler, P. W. and Dolan, F. X.: Two-Phase Flow
Pump Data for a Scale Model NSSS Pump, ASME
Symposium on Polyphase Flow in Turbomachinery,
65-73, (December 10-15, 1978) San Fransisco.
Hench, J. E. and Johnston, J. P.:Two-Dimensional
Diffuser Performance with Subsonic, Two-Phase, Air-
Water Flow, Journal of Basic Engineering, (March,
1972), 105120.
Wallis, G. B.: One-Dimensional Two-Phase Flow,
McGraw Hill Book Co., (1969).
Craver, M. B.: Numerical Computatuon of Phase
Separation in Two-Phase Flow, Journal of Fluids
Engineering, 147-153, Vol. 106, (June, 1984).
U K Patent Application, 982193533A, Application
Number 8718564, (February, 1988). Application filed
by Nuovopignone-lndustrie Meccaniche E Fonderia,
S.p.A., Italy.
8 MULTIPHASE FLOW THROUGH CENTRIFUGAL PUMPS
SI METRIC CONVERSION FACTORS
bbl x 1.589 873 E-01
= m3
ft x 3.048 E-01 =m
in x 2.540 E-02 = m
Ibm x 4.535 924 E-01 = kg
psi x 6.894 757 E+OO = kPa
TABLE 1
Error Comparisons for Dynamic and Correlational Models
DYNAMIC CORRELATIONAL
MODEL MODEL
(MODEL 2)
ABSOLUTE PRESSURE
AVERAGE AVG. ERROR STANDARD ABSOLUTE STANDARD
RISE AT BEP
ERROR
PUMP
(PSI)
DEVIATION AVG. ERROR DEVIATION
FOR 100%
(PSI) (PSI) (PSI) (PSI)
DIESEL
l-42 -0.29 1.09 1.35 3.19 4.64 14.20
C-72 -0.34 1.61 1.66 1.49 2.02 11.43
K-70 1.06 1.27 2.07 5.03 5.52 15.93
I
ACTUAL
I
vu2
Figure I: Effect of slip on outlet velocity triange
A
THEORETICAL PRESSURE
B
RISE (LIQUID)
-- -m--
THEORETICAL PRESSURE
RISE (29PHASE)
PUBLISHED PRESSURE
RISE (LIQUID)
PREDICTED PRESSURE
-PHASE)
FLOW RATE
Figure 2: Dynamic model formulation
I
I
I = IMPELLAR 1NLE.T
2= IMPELLAR OUTLET
Figure 3: Noemencalture
L42B
C-72
K-70
Figure 4: Pump geometries
- -
0
0 1000
2000 i 3000 4000 5000
FLOW RATE (bbl/d)
- 100% LIQUID - 100% LIQUID
2=PHASE, ACTUAL 2=PHASE, ACTUAL
0 2=PHASE, PREDICTED 0 2=PHASE, PREDICTED
Figure 5: Dynamic Model Predictions (K-70): Pin = 60 psig, 9.92% gas
20
. 0
I U L-PHr
-
-
0
A
A 7o LQUD
7 :I ASE, ACTUAL
0 2=PHASE, PREDICTED
-
1
I,_ I,
n
u
1000 2000
FLOW RATE (bbl/d)
4000 5000
Figure 6: Dynamic Model Predictions (K-70): Pin = 308 psig, 19.83% gas
2000 3000
FLOW RATE (bbl/d)
4000 5000
Figure 8: Dynamic Model Predictions (K-70): Pin = 350 psig, 49.58% gas
25
20
F
s
W 15
ctl
3
%I0
W
Qi
n
5
0
- 100% LIQUID
n 2-PHASE, ACTUAL
0 Z-PHASE, PREDICTED
0 2000 3000 4000 5000
FLOW RATE (bbl/d)
Figure 9: Dynamic Model Prediction (C-72): Pin = 55 psig, 9.92% gas
25
20
e
5
e_ 15
w
OL
3
z -lo
W
111:
e 5
0
- 100% LIQUID
n 2=PHASE, ACTUAL
0 Z-PHASE, PREDlCTED
8
-
-
I I * I I
0 1000 2000 3000
FLOW RATE (bbl/d)
4000 5000
Figure 10: Dynamic Model Prediction (C-72): Pin = 300 psig, 19.83% gas
25
-
n
100% LIQUID
2=PHASE, ACTUAL
20 --
f
0 2=PHASE, PREDICTED
0 1000
2000
3000
FLOW RATE (bblld)
1000
5000
Figure 11: Dynamic Model Prediction (C-72): Pin = 220 psig, 29.75% gas
25
20
0
2000 3000
FLOW RATE (bblld)
- 100% LIQUID
n 2=PHASE, ACTUAL
0 2=PHASE, PREDICTED
Figure 12: Dynamic Model Prediction (C-72): Pin = 410 psig, 49.58% gas
Figure 13:
- 100% LIQUID
ZPHASE, ACTUAL
0 2-PHASE, PREDICTElI
0 500 1000 1500 2000 2500 3000
FLOW RATE (bbl/d)
Model Predictions: I-42B pump, Pin=60 psig, in=5.67%, Diesel-CO2
- 100% LIQUID
n 2=PHASE, ACTUAL
0 2-PHASE, PREDICTED
FLOW RATE (bblld)
Figure 14: Model Predictions: I-42B pump, Pin=95 psig, in=30%, Diesel-CO2
loo0 1500 2000
FLOW RATE (bblld)
2500
Figure 15: Model Predictions: L42B pump, Pin=280 psig, in=30%, Diesel-C@
20
15
10
5
0
Figure 16:
- 100% LIQUID
2-PHASE, ACTUAL
2=PHASE, PREDICTED
-
0 500 loo0 1500 2000 2500 3000
FLOW RATE (bbl/d)
Model Predicitons: I-42B pump, Pin=280 psig, in=39.94%, Diesel-CO
a STAGE 1 d
I
I = IMPELLER
D = DIFFUSER
4 STAGE 2 4r STAGE 3 --~
a- b---++---D-w-----L---w+-
D---c~~---I -c-)+----D -w
Figure 17: Multistage pump behavior