You are on page 1of 4

Efficiency and Pressure Recovery in

Hydraulic Jet Pumping of Two-Phase


Gas/Liquid Mixtures
Baohua .llao, * SPE, Roger N. Blais, SPE, and Zellmlr Schmidt, SPE, U. of Tulsa

Summary. Hydraulic jet pumping of gas/liquid mixtures was studied experimentally, and a mathematical model is proposed to ex-
tend the standard single-phase model for predicting efficiency and pressure recovery to suction fluids with gas/liquid ratios up to 2,200
scf/STB. The experimental progrruv comprises 616 low-pressure tests in a plastic model pump designed for flow visualization and meas-
urement of pressure profile along the throat and diffuser, and 373 high-pressure tests on a stock pump. For the high-pressure tests,
power fluid was supplied at 200 to 3,000 psi and at 200 to 860 B/D; air was supplied from 0 to 185 MscflD. Discharge pressures
ranged from 800 to 2,000 psi.
The mathematical model extends a previous model that describes single-phase performance from mass and energy conservation. The
empirical loss coefficients for the nozzle and throat/diffuser are replaced by a nondimensional expression that varies as three dimension-
less parameters: nozzle-to-throat area ratio, discharge-to-power-fluid pressure ratio, and air/water ratio (which usually is in convention-
al units of cubic feet per stock-tank barrel but is, of course, basically dimensionless). The loss coefficient for the nozzle is constant,
but for the throat/diffuser it is a constant plus a product of a constant times the three parameters, each to a power.
Compared with the standard model, which always overpredicts pressure recovery and thus efficiency, the new model reduces the
standard error of the estimate to 18% of its former value.

Introduction The goal is to determine the power-fluid pressure, PP ' necessary


The accepted theory of jet-pump operation is derived from single- to lift fluid at a given pump-intake pressure, Pi' to the surface by
phase assumptions. Power fluid and suction fluid are assumed to ensuring that the pump-discharge pressure, Pd' is sufficient.
be similar liquids. Since Rankine 1 developed the basic theory of These pressures are related by a dimensionless pressure-recovery
operation in 1870, using concepts of mass and energy conserva- ratio, Fp:
tion, most investigators have grappled with realistically assessing Fp=(PrPi)/(p p -Pd)' ............................ (1)
frictional losses, not with addressing operation when suction fluid
is a multiphase mixture. Notable among these early studies are those The conservation laws imply that Fp will depend on a geomet-
of Gosline and O'Brien 2 and Cunningham. 3 Petrie et al. 's4 stan- ric factor of the pump and on the mass flow rates of the suction
dard installation design model cautions users to apply it only when and power fluids. These factors are FA =An/At' the ratio of noz-
free gas is limited to < 10 scf/STB. Corteville et al. 5 recently pub- zle cross-sectional area to throat area, and the dimensionless mass-
lished results of a study on two-phase performance using kerosene flow ratio, Fm' defined as
and N2 at power-fluid pressures up to 1,160 psi. Fm=w;lwp' ...................................... (2)
The two-phase model reported here is based on experiments con-
ducted with water as a power fluid and with water and air as the Therefore, we seek a function of the form Fp =Fp(Fm,FA . .. ).
suction fluid in a surface test loop operated at field-scale pressures In practice, the simple theory based on conservation laws must
and flow rates. This model extends the applicability of the stan- acknowledge the real work by including frictional loss factors for
dard design model by adjusting an empirical loss coefficient for energy irreversibly disspated in the nozzle and in the throat/diffuser
the throat and diffuser. Rather than being a dimensionless constant, section. These two loss factors, fn and ftd' respectively, can be
this coefficient becomes a function of three dimensionless parame- evaluated only empirically. The form of the model given by Petrie
ters - one describing pump geometry; another, the operating pres- et at. is written most easily if we define a geometric factor to sim-
sures; and a third, the gas/liquid ratio. Because all tests were plify the algebra:
conducted with air and water, no empirical adjustment for physi- B=(1-2FA )Fl/(1-FA )2 . .......................... (3)
cal properties of the suction or power fluid is attempted. The work
reported here is drawn from Refs. 6 and 7, which provide more Then,
detailed information.

Theory and Definitions Fp= (1+fn)-2FA-BF~+(1+ftd)Fl(1+Fm)2 ........ (4)


The principal component of a jet pump (Fig, 1) is a nozzle fitted
to a throat/diffuser section. Power fluid, usually clean crude from Petrie et al. give values of fn =0.03 and ftd=0.2 for a pump
a surface pump, is injected into the nozzle under high pressure, operating on pure liquid. Note that the model is relatively insensi-
from which it issues at high speed. The resulting high-speed jet tive to small variations in fn' The approach in this study is to re-
is at low pressure by Bernoulli's principle. Thus, it entrains suc- lax the constraint that ftd be constant and instead regress on
tion fluid, and the combined mixtured is allowed to decelerate in additional parameters, including FA' Fpdp=Pd/P p ' and Faws' The
a cylindrical throat and then passes through an expanding, conical results of this regression are given in the Data Analysis and Model-
diffuser. Pressure is recovered as the slowing fluid swaps kinetic ing section.
energy for pressure. First, however, it is necessary to clarify how Fm is to be defined
The quantitative theory of jet-pump performance is based on mass when gas is present in the suction fluid. Letting qi remain the in-
and energy conservation. Cunningham 3 gives a complete deriva- take volumetric liquid flow rate, we must add a gas flow rate. Be-
tion, and Petrie et at. 4 provide an abridged form. This steady-state cause Fm is defined as a ratio of mass flow rates, the custom of
model assumes uniform properties of a single-phase, incompressi- writing it as a ratio of volumetric flow rates applies only when the
ble discharge fluid resulting from complete mixing of power and densities of the power and suction fluids are identical. We define
suction fluids in an axially symmetric, simplified pump geometry. qiaw as the volumetric flow rate of water that equals the mass flow
rate of air. Then, F m can be expressed as
'Now at IntI. Training and Consulting Inc. F m =(qi +qiaw)/qp' ................................ (5)
Copyright 1990 Society of Petroleum Engineers

SPE Production Engineering, November 1990 361


~ t-------------------l"'*T---1
I . . !~I I
~ I
II
I
I
Fig. 1-Schematlc showing nomenclature. .r-~I
I
::r1~'"'!"I!1!'IP1
Assuming API standard conditions (14.7 psia and 60F), if qja I
is the volumetric flow rate of air at standard temperature and pres- .. t
~-----,qr
sure, then we can derive, from the ideal-gas law and appropriate
conversion factors,
qjaw=0.2178qja' .................................. (6)
I. CIMTIlIf'UGA.t. t"I.ItIIr IJ. PULUTION OotoM'I" . OftI'ICr fl..0WtII1ET11I 15. POLl
Thus, Fm =(qi+0.2178qia)/qp' ........................ (7) I. TAN. t. TUflaIN( 'LOWIUTEft to. TUT '"U.... 14. WINCH
II.L.p.mpUlt T.IIIXVI: II SI'jlRATOft
... ",,.. TIIIPt.()( '.COYPftUSOft 12.MUfEXCH.lMIPt

We must still consider pump efficiency, defined as


E==P/P j (8) Fig. 2-Jet-pump test facility: two-phase flow loop.

=[(Pd-Pi)(qj+qiaw)]l[(pp -Pd)qp] ................. (9)


reported in Ref. 6. Low-pressure tests were not included in the
=FpFm . ....................................... (10) model reported here to avoid giving undue weight to exploratory
Thus, pump efficiency is merely the product of the pressure- results not in the normal range of field operations.
recovery and mass-flow-rate ratios.
Data Analysis and Modeling
Experimental Facility and Procedure The goal of modeling was to produce an empirical expression for
The hydraulic-jet-pump experimental facility 8 consists of an 85- ltd for use in Eq. 4. Predicted or calculated values of Fp were com-
ft-taIl mounting tower and a multiphase flow loop (see Fig. 2). The pared with measured values determined from the defining expres-
pump is mounted aboveground to facilitate instrumentation access. sion, Eq. 1. Then, ltd was adjusted to minimize the sum of the
Transducers on the pump and flow loop are connected to data- square deviations between calculated Fp' Fpc, and measured Fr
acquisition devices housed in a portable building beside the tower. After numerous analysis-of-variation computations to determine the
significant parameters on which ltd can be modeled, three were
Two-Phase-F1ow System selected, as described in the Introduction: FA' Fpdp, and Faws'
Fig. 2 shows the flow diagram of the test system. The power fluid Petrie et ai. reported constant values of fn =0.03 and ltd =0.2. We
is water, while the simulated production fluid is a water/air mix- adjustedfn to 0.04, which is the best fit to our data but does not
ture. A centrifugal pump (1) supercharges water from the tank (2), substantially differ from the Petrie et ai. model in predicting Fp
supplying up to 3,400 BID to both the production-fluid (3) and the because the model is insensitive tofn' It wasfttl, however, that was
power-fluid (4) triplex pumps. Water from the low-pressure (max- transformed from a constant to a function. The pure-liquid data were
imum 1,480 psi) triplex pump (3) passes through a pulsation damper segregated (Faws=O), and a constant value for ltd was found by
(5a) and 1- or *-in. turbine flowmeter (6a) before entering the mixer regression to be 0.1. The two-phase model was then forced to pro-
(7), where it merges with high-pressure air from the compressor vide this value when Faws=O by regressing on iftd-0.1) when
(8). Water from the high-pressure (maximum 3,730 psi) triplex Faws ;CO. The model was of the form
pump (4), which serves as the power fluid, likewise passes through Itd=O.1 +a(Fpdp)b(Faws)C(FA)d. . ................... (11)
a pulsation damper (5b) and 1- or *-in. turbine meter (6b) before
entering the test pump (10). Ambient air is compressed (8) and By taking logarithms of each variable and using regression pro-
stored in a 264-ft3 reservoir (8a) before being sent through an grams in the SAS Statistical Package,TM Version 6.03, values for
orifice flowmeter with a 3-in. meter run (9) and a screen plate at a, b, c, and d were found:
the tee, just upstream from the mixer (7). The two-phase bubble- a=4.1 Xl0- 3 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12)
flow production fluid from the mixer (7) enters the suction of the
test pump (10). The metered high-pressure power fluid is injected b= -2.6, ....................................... (13)
directly into the test pump. Data are obtained from pressure and c=0.62, ........................................ (14)
temperature transducers (P and T) at various points on the flow
system. The spent power fluid from the pump's lower end, together and d=0.53 . ...................................... (15)
with the produced gas/liquid mixture, returns to the separator (11). The constant a presumes that Faws is measured in standard cu-
The separated liquid returns to the tank (2) after passing through bic feet per stock-tank barrel. (If Faws truly is expressed in dimen-
the heat exchanger (12), while the air is exhausted to the atmosphere sionless units, the exact conversion 539 scf==96 STB yields
from the separator. a=1.2XlO- 2 .)
Instruments were recalibrated regularly with, for example, a dead- In this model, we elected to use the air/water ratio at standard
weight tester for pressure. The accuracies attributed to the instru- conditions because that is typically what the field operator knows
ments are 0.25% for pressure, 0.25% for the digital thermom- by direct measurement. The case can be made, however, that for
eter, 0.5% for the turbine flowmeters, and 0.7% for the orifice the model to be physically correct, it must use the in-situ air/water
meter. Steady-state conditions were maintained during each test, ratio in the throat/diffuser. Even if the field operator knows the
with a maximum allowed temperature drift 5 F. necessary pressure in situ, it is very unlikely that he or she will
know the temperature of the turbulent fluid. If it is known, as in
Experimental Program. Some 373 tests were run with pumps hav- our laboratory measurements, the standard error of the estimate
ing four different throat/nozzle combinations (Table 1). The pump can be reduced slightly-e.g." for our 263 two-phase measurements,
was manufactured by Natl. Production Systems (NPS); thus, the from 0.067 to 0.057. This is a 15% improvement over the U. of
NPS nozzle and the throat numbers are given. Table 2 shows the Tulsa Artificial Lift Projects model at standard conditions and
range of operating variables. Of the 373 tests, 263 were two-phase reduces the standard error to 15% of Petrie et ai. 's error, as op-
and 110 were pure-liquid tests. Preliminary modeling based on the posed to 18 % for the standard-condition field model. For this case,
616 tests performed at low pressure in the plastic model pump were if Faw replaces Faws in Eq. 11, then the coefficients become

362 SPE Production Engineering, November 1990


TABLE 1-PUMPS USED IN TESTS 1.2r-----------------~

Number in.2 Number in.2 FA(An/At}


1.0
.: .....

8
8
0.0131
0.0131
10
8
0.0562
0.0346
0.2331
0.3786
..*.. .. ...
4 0.0050 5 0.0167 0.2994
4 0.0050 4 0.0131 0.3817 0.8

TABLE 2-PUMP OPERATING VARIABLES u...'l..0.6

Power fluid
Pressure, psi 2,000 to 3,000
Rate, B/D 200 to 860
Suction-fluid pressure, psi 100 to 1,400
Liquid rate, BID 40 to 860
Gas rate, Mscf/D o to 185
F BWS' scf/STB o to 2,200
Discharge-fluid pressure, psi 800 to 2,000

0.2 0.6 0.8 1.0 1.2


a=0.23, ........................................ (16)
Fp
b= -2.0, ....................................... (17)
c=0.67, ........................................ (18) Fig. 3-Fpc from Petrie et al.'s model vs. measured F p
pressure-recovery ratio: single-phase model.
and d=O.44 . ...................................... (19)
For practical application, this refInement is unnecessary. It is use-
ful to compare this model to the standard Petrie et at. model when The objective is to estimate production effIciency for a selected noz-
gas is present. Figs. 3 and 4 show the disagreement between pre- zle/throat combination. A variety of geometries are evaluated and
dicted and measured values of pressure recovery, Fp, for the two the optimum design is selected. For fIeld applicability, we use the
models. If a model were perfect, predicted values, Fpc, would lie air/water ratio at standard conditions. The steps involved are as
on the diagonal line. Clearly, the standard model overpredicts Fp follows.
when gas is present. Expressed quantitatively, the standard error 1. Compute the pump-intake fluid gradient.
of the estimate over the 263 data points in two-phase conditions 2. Compute the minimum throat annulus to avoid cavitation.
is 0.372 for the standard Petrie et ai. model and 0.067 for the pro- 3. Select the nozzle and throat sizes from manufacturers' infor-
posed field model. Thus, the new model reduces the error of the mation and calculate FA'
estimate to 18% of its former value. 4. Select a surface power-fluid pressure and estimate the resulting
Fig. 5 shows the degradation of pump efficiency as gas is added nozzle pressure and flow rate. First, compute the Reynolds num-
to the suction fluid. Assumed values for FA and Fp are given. ber in the power tubing, determine whether the flow is laminar or
turbulent, and calculate the frictional loss.
Application 5. Characterize the returning fluid, which consists of the suction
A brief outline of how to apply this model interactively in an in- and the power fluids. The gradient, water cut, and gas/liquid ratio
stallation design follows. A casing-type open system is assumed. are needed.
6. Determine the frictional loss of the return flow through the
casing/tubing annulus.
1.2.----------------""71 7. Estimate a reasonable pump-discharge pressure.
8. Setfn=O.04, and computeftd with Eqs. 11 through 15.
9. Compute Fm with Eq. 7.
1.0

40r---.---._---.---.--~r---._--._--_r--_,

0.8 Faws=OSCF/STB
35

()
30
~0.6
25

0.4 ~ 20
UJ

15

0.2 FA=2331
fn =.04
Fpdp=7

0.2 0.4 0.6 0.8 1.0 1.2


0~~0~.3~-0~.~6--~0~.9r--+1.2~~1.~5--~~-2~.~I--~2~A--~2~
Fp Fm
Fig. 4-F pc from this model vs. measured Fp pressure-
recovery ratio: two-phase model. Fig. 5-Efficiency at various air/water ratios.

SPE Production Engineering, November 1990 363


Authors ltd = friction factor at throat and diffuser, dimensionless
Faw = in-situ air/water ratio, ft 3/STB
Faws = air/water ratio, scf/STB
FA = ratio of nozzle area to throat area, dimensionless
Fm = mass-flow-rate ratio, dimensionless
Fp = measured pressure-recovery ratio, dimensionless
Fpc = pressure-recovery ratio computed from model,
dimensionless
F pdp = ratio of discharge pressure to power-fluid pressure,
dimensionless
Pd = pump-discharge pressure, psi
Pi = pump-intake pressure, psi
Jlao Blais. Schmidt
Pp = pump power-fluid pressure, psi
Baohua "lao, currently employed by Inti. lhllning and Con- Pi = consumed input power
sulting Inc. In Tulsa, has 20 years' field experience In China.
He holds a BS degree In mechanical engineering from Qlng
Pt = transferred useful power
Hua U. and an MS degree In petroleum engineering from the qi = liquid-phase flow rate at pump intake, BID
U. of lUisa, where he has also completed a PhD dissertation. qia = air flow rate through pump in standard conditions,
Roger Blais has been associate director of the U. of Tulsa MscflD
Artificial Lift Projects since Its founding In 1983. He chaired qiaw = equivalent-volume flow rate of water, converted
the Physics Dept. during 1986-88 and currently Is Interim
from qia' B/D
provost and academic vice president. Blais holds a BA degree
qp = flow rate of power fluid through pump, BID
from the U. of Minnesota and a PhD degree from the U. of Ok-
lahoma, both In physics. Zellmlr Schmidt Is a professor of Wi = inlet mass flow rate, Ibm/sec
petroleum engineering at the U. of Tulsa and Director of the wp = power-fluid mass flow rate, Ibm/sec
U. of lUlsa Artificial Lift Projects. He spent 10 years as a pro-
duction engineer with INA-Naftaplln In Yugoslavia and has Acknowledgments
served as a consultant to various companies. Schmidt, who
was a 1987 Distinguished Lecturer, holds a BS degree In en- We thank NPS for donating many of the test-facility components,
gineering from the U. of Zagreb and MS and PhD degrees In including the two triplex pumps and two jet pumps. Hal Petrie and
petroleum engineering from the U. of lUlsa. Fred Christ of NPS also gave much-needed advice during the study,
although any errors are purely our responsibility. The 16 sponsor-
ing members of the U. of Tulsa Artificial Lift Projects also deserve
10. Compute Fp with Eq. 4. recognition for their support. Raymond Carr and Ed Barefoot helped
11. Recompute the nozzle and intake pressures from F p and Pd' with construction and operation of the facilities; Bill Coberly and
12. Recompute the surface pump operating pressure and maxi- Peyton Cook assisted with statistical modeling; and Carol Antle
mum noncavitating power-fluid flow rate. typed the manuscript.
13. Calculate efficiency, E, with FmFp in Eq. 10.
14. Determine the required horsepower ofthe surface pump from References
its operating pressure and power-fluid flow rate. 1. Rankine, J.M.: Proc., Royal Soc. of London (1870) No. 123,80-83.
15. Compare the efficiency and horsepower requirement to other 2. Gosline, J.E. and O'Brien, M.P.: The Water Jet Pump, U. of Califor-
nozzle/throat combinations. nia Publications in Engineering (1942) 3, No.3, 167-90.
As an example, let us select a pump at 6,000-ft depth with 51,6- 3. Cunningham, R.G.: "The Jet Pump as a Lubrication Oil Scavenge Pump
in. casing and 27's-in. tubing, a bottomhole pressure of 1,000 psi, for Aircraft Engines," Technical Report 55-143, Wright Air Develop-
ment Center, Dayton, OH (1955).
a temperature of 145F, a GOR of 1,000 scf/STB, a water gra-
4. Petrie, H.L., Wilson, P.M., and Smart, E.E.: "Jet Pumping Oil Wells,"
dient of 0.455 psi/ft, and a water cut of 0.25. The suction oil and World Oil (Nov. 1983) 51-56.
power fluid are both 35 API oil of 5-cp absolute viscosity. Set the 5. Corteville, J .C. et al.: "Research on Jet Pumps for Single and Multiphase
wellhead backpressure of 100 psi and the desired production rate Pumping of Crudes, "paper SPE 16923 presented at the 1987 SPE An-
of 600 BID. Choosing an NPS No.9 nozzle and No. 10 throat nual Technical Conference and Exhibition, Dallas, Sept. 27-30.
(FA =0.0167/0.0562=0.2972), we findltd=0.4523, Fm =0.5018, 6. Jiao, B.: "Behavior of Hydraulic Jet Pumps When Handling a Gas-Liquid
Fp =0.4514, and E=23%. These values for pressure recovery and Mixture," MS thesis, U. of Tulsa, Tulsa, OK (1985).
7. Jiao, B.: "Performance Model for Hydraulic Jet Pumping of Two-Phase
efficiency appear reasonable.
Fluids," PhD dissertation, U. of Tulsa, Tulsa, OK (1988).
8. Blais, R.N. et al.: "Test Facilities for Gas-Lift Valves and Hydraulic
Conclusions Jet Pumps," Advances in Test Measurements, Instrument Soc. of America,
The pressure recovery and efficiency of hydraulic jet pumps de- Research Triangle Park, NC (1985) 22, 525-30.
grade when free gas is present in the suction fluid, but predicting
SI Metric Conversion Factors
pump performance is possible when gas is present if three dimen-
sionless parameters are used to model the frictional losses associated API 141.5/(131.5 + API) g/ cm 3
with the throat and diffuser. Such modeling reduces the standard bbl x 1.589873 E-Ol m3
error of the estimate by more than a factor of five compared with cp x 1.0* E+OO mPa's
the unadjusted Petrie et aZ. model. ft x 3.048* E-Ol m
ft3 x 2.831 685 E-02 m3
Nomenclature OF (OF-32)/1.8 =oC
in. x 2.54* E+OO cm
a = regression coefficient, dimensionless
in. 2 x 6.451 6* E+OO cm 2
An = nozzle area, in. 2 psi x 6.894757 E+OO kPa
At = throat area, in. 2 scf/S,TB x 1.801 175 E-Ol std m 3/stock-tank m 3
b,c,d = regression exponents, dimensionless
B = parameter used in computing Fp ' dimensionless
Conversion factor is exact. SPEPE
Original SPE manuscript received for review Oct. 2, 1988. Paper accepted for publication
E = jet-pump efficiency, dimensionless June 14, 1990. Revised manuscript received June 7, 1990. Paper (SPE 18190) first present
in = friction factor at nozzle, dimensionless ed at the 1988 SPE Annual Technical Conference and Exhibition held in Houston Oct. 2-5.

364 SPE Production Engineering, November 1990

You might also like