You are on page 1of 4

Evaluation of Three New Methods

For Predicting Pressure Losses


In Vertical Oilwell Tubing
I. R. Vohra, SPE-AIME, U. of Tulsa
J. R. Robinson, SPE-AIME, U. of Tulsa
James P. Brill, SPE-AIME, U. of Tulsa

Introduction
Six of the several correlations available for predict- production. Rather, these data are extremely valu-
ing pressure losses during simultaneous, continuous, able because of the absence of mass transfer between
steady-state flow of oil, water, and gas in vertical pipes phases and the resulting improved accuracy of, for
were evaluated statistically by Lawson and Brill. 1 example, predicted in-situ phase velocities.
Three more promisillg correlations have recently been All three of the correlations were developed from
published - those of Beggs and Brill,2 Aziz et at. s, two-phase flow data, one of the phases being gas. A
and Ch4erici et at. 4 The purpose of this paper is to third phase (water) can be included if we assume that
extend the work of Lawson and Brill to cover the its presence does not change the physical phenomena
new correlations. of two-phase flow. Possible changes that could inva:li-
Details of the three new methods can be found in date results are slippage between oil and water, for-
the original sources listed in the references. The same mation of emulsions, and the influence of water on
flowing pressure surveys and related data for 726 gas bubble coalescence and the formation of gas slugs.
field and experimental wells were used as were re- The degree to which such changes are present in the
ported by Lawson and Brill. Fluid physical properties three-'phase data is unknown. .
were estimated using the· same correlations as em- We applied the three correlations to all data, know-
ployed by Lawson and Brill. ing full well that many of the data were beyond the
. Although every attempt was made by Lawson and range of variables used to develop the correlations.
Brill to screen data in the data bank, no doubt some However, correlations are frequently used indiscrim-
questionable data have been included. In a study such inately beyond their stated ranges of validity. There-
as this, the quality of the data is critical. It has been fore a test of their performance over a broad range
found, for example, that small errors in measured gas of data is of value.
volumes and oil formation volume factors can signifi-
cantly affect the calculated results. A valid criticism Validation of Programming
of the bank is that inadequate use was made of what Pressure losses calculated by a computer program of
limited measured PVT data were available. Another each method were compared with the corresponding
possible criticism is that many of the data used were pressure losses reported by each author for identical
gas-water data with low gas/water ratios, a situation well cases. This comparison indicated the degree of
seldom found in practice. No claim was made that agreement between the results of the two programs.
the gas-water data were meant to represent gas well Because of variations in the selection of fluid physical

The reliability and relative merits have been estimated for three recently published
multiphase-flow pressure-drop prediction correlations applicable to vertical tubing.
Data from 726 tests, embracing broad ranges of flow rate,. pipe size, API gravity,
gas/liquid ratio, and water/oil ratio were used in the evaluation. It was found that
no single correlation consistently performed best in every range.

AUGUST, 1974 829


TABLE I-STATISTICAL RESULTS FOR property correlations, convergence techniques, and
ALL WELL TESTS
specified tolerances, exact duplication was not always
Average expected.
Percent Standard
Correlation Difference Deviation Beggs's calculated pressure gradients for 27 uphill
Beggs and Brill -17.8 27.6
and 27 downhill vertical flow cases 5 were reproduced
Aziz et al. + ·8.2 34.7 by the program of this study. The no-slip friction
Chierici et al. -42.8 43.9 factor was calculated using the Colebrook equation
rather than the smooth-pipe equation suggested by
Beggs and Brill.2
Aziz et aI.3 published their computer program to-
gether with calculated pressure traverses for 48 oilwell
cases, of which 38 were from the study of Espanol
et at. 6 Of the 726 well test data in this study, 44 were
taken from Ref. 6. The results of this program agreed
within 5 percent with the Aziz et aI. calculated pres-
sure gradients for the well tests of Ref. 6.
Considerable difficulty was encountered in match-
ing resuHs for the 31 well cases reported by Chierici
et at. 4 The major differences were for their Wells 11
through 13. If these three wells were excluded, an
average error and standard deviation of -1.63 and
6.02 percent, respectively, were calculated, compared
z
with +0.30 and 5.37 percent by Chierid et aI. for
o
;::::
the 28 wells.
« 40
G:;
o
Discussion of Results
~ 20 Por 'statistical evaluation of each pressure-loss corre-
«o lation, three numerical measures of reliability were
z
;:: used. These were the percent difference, (PD), aver-
Vl o L._ - - - ' _ - - - l L - - - - l L - - - - ' _ - - - '_ _--'_---l.--c--_--l
age of percent differences (APD), and standard devi-
1.049 1.38 1.61 1.995 2.376 2.441 2.992
1311 (1751 11401 (2281 (4\1 1601 (46) ation of percent difference values from the average
VALUES OF INSIDE TUBING DIAMETER, INCHES percent difference (SD).
(NUMBER OF WELL TESTS IN EACH GROUP IN PARENTHESESI
Fig. I-Statistical results for well data grouped
by internal tubing diameter.
PD = APm - APe X 100 percent
APm
n
40 :z PD.
APD = ...;.,i_=.....I_ _
.....
u
n
ifi 20 n
.....
0:::
u. :z (PD, - APD)2
u.
i = 1
SD=
~ 0 n- 1
z
.....
U
0:::
where
lt -20 APe = calculated pressure loss,
.....
o
<C o APm = measured pressure loss,
.....
0:::
• AZIZ
;;C -40
A. BEGGS n = number of cases.
o CHI ERICI Statistical results for the three correlations are pre-
lI'-60 sented in Table 1.
z Table 1 indicates that t4e Beggs and BrilF Corre-

tf~
0
i= lation yielded the lowest standard deviation of 27.6
<C 40
a.
Cl
Cl
0:::
<C
Cl
Z
20
•r
A.:
0'--::;
.~~
~t

percent, and the Aziz et at. 3 correlation gave the
lowest average error of + 8.2 percent. Of all the
methods analyzed in both this study and that of Ref.
1, only the Aziz et at. method yielded a positive aver-
<C
l-
age percent difference.
V>
0 Besides determining the cumqlative statistical
13-20 25-31 31-35 35-43 43-56 figures for the 726 well cases, we evaluated the three
(201 (54) (117) (52) (49)
correlations within specific ranges of important flow
RANGES OF PRODUCED OIL: API GRAVITY
(NUMBER OF WELL TESTS IN EAC.H GROUP IN PARENTHESES) variables. The five' variables used in Ref. 1 were
Fig, 2-Statistical results for well data' grouped selected and the performance of each method was
by produced-oil API gravity. tested within set limits of each variable. Pigs. 1
830 JOURNAL OF PETROLEUM TECHNOLOGY
through 5 show the resulting performance curves. 20
Fig. 1 shows that all three correlations followed the
same general trend with increasing pipe size and that ...
u
all the methods overpredicted pressure losses for well i5 0
cases in which the inside tubing diameter was between ...
De
u.
u.
2.6 and 3.2 in. Only five test cases had tubing diam- Cl -20
eters above 3.2 in. and therefore no reliabl~ estimates I-
i5
on performance could be determined for these larger ...
'-'
De

pipe sizes.
Performance curves for the 292 well cases in which
... -40
CL

<.!>
• AZiZ
there was no water production are plotted in Fig. 2 ...::c -60
<C
De
.to BEGGS
as a function of produced oil API gravity. For oils in o CHI ERI C1
0 __0

the range of 20° to 31 ° API gravity, all three corre-


lations underpredicted pressure losses. The Aziz et ai. !!f'. 60
method performed better for the heaviest (below 20° z
0
API) and lightest (above 43° API) oils, whereas the ;::
<C 40
Beggs and Brill method yielded better results for oils ~
of intermediate API gravity. . Cl
Cl
. Fig. 3 depicts the behavior of the statistical curves De
<C 20
Cl
as a function of the produced gas/liquid ratio. Both Z
<C
the Beggs and Brill and the Chierici et aU schemes l-
V>
0
overestimated calculated down-hole pressures in all o .004-.036 0.1-1.0 1.0-20 20-00
(292) (81 (91 (92) (3251
ranges of gas/liquid ratio. However, the Aziz et ai. RANGES OF PRODUCING WATER/OIL RATIO
correlation underpredicted pressure drops for gas / (NUMBER OF WELL TESTS IN EACH GROUP IN PARENTHESES I
liquid .ratios below 5,000 scf/STB; above 5,000 scf/ Fig. 4-Statistical results for well data grouped
STB, this correlation overestimated flowing bottom- by producing water/oil ratio.
hole pressur~s and exhibited the least scatter.
No general trend could be ascertained in the varia-
tion of performance with produced water / oil ratio
(Fig. 4). All three methods had the highest standard
deviation for well cases with no water phase present
and with water/oil ratios above 20. The lack of con-
sistency in the results of all three correlations could

20 . - - - - . - - - - . - - - - , - - - . - - - - - . . . ,
.....
30
___ e .----.
u
t5 10 •

-1/--0..----0
o a:::
~ 0

c
I-

u
t5
a:::
-20
U
~-10
z
UJ

a:::
~ -30 f-
0; .to

~ -40
.....
UJ

~
a:::
... /
0/
c.:l UJ
<C
a::: ;;;: -50
~-60 • AZIZ
<C e AZIZ
.to BEGGS .to BEGGS
'Ffl. -80 o CHIERICI z o CHIERICI
0
z i=
o <C 80
~ 60 o

:~o
~
/0_ c
~
~
c
c a::: -
---:::t~i~
0 / -e <C
40
~
<C
c
40
.lit
~l~.to~e c
Z
<C
l-
~ V') I

t;:; 20L---=c1...:-::-=--:c::-::--'=::-:----::-::c~=~=_=-=:__---' 0-.4 .4-LO LO-2.0 ) 2.0


5000-80,000 (219) (221) (154) (132)
(22m (2261 11671 (113) RANGES OF SUPERFICIAL VELOCITY OF PRODUCED
RANGES OF PRODUCING GAS/LIQUID RATIO L1QUI DS. FTISEC
(NUMBER OF WELL TESTS IN EACH GROUP IN PARENTHESES) (NUMBER OF WELL TESTS IN EACH GROUP IN PARENTHESES)
Fig. 3-Statistical results for well data grouped Fig. 5-Statistical results for well data grouped by
by producing gas/liquid ratio. superficial velocity of produced liquids.

AUGUST, 1974 831


be attributed to the possible formation of oil-water tive performance of the various correlating schemes
einulsions. For such mixtures, averaging separate oil within certain limitations of producing well flow pa-
and water properties does not yield appropriate values rameters. No single method consistently performed
for the various liquid properties required in each best in every range.
correlation.
Analysis of Fig. 5 shows the steady improvement References
in standard deviation with increasing superficial liquid 1. Lawson, J. D., and Brill, J. P.: "A Statistical Evaluation
of Methods Used To Predict Pressure Losses for Multi-
velocities at the wellhead. As stated by Lawson and phase Flow in Vertical Oilwell Tubing," J. Pet. Tech.
Brill/ "The poorer predictions at lower liquid super- (Aug. 1974) 903-914; Trans., AIME, 257.
ficial velocities could be an indication of a more 2. Beggs, H. D., and Brill, J. P.: "A Study of Two-Phase
cyclical or noncontinuous type of flow, which would Flow in Inclined Pipes," J. Pet. Tech. (May 1973) 607-
617; Trans., AIME, 255.
be expected to be more elusive of being predicted by 3. Aziz, K., Govier, G. W., and Fogarasi, M.: "Pressure
correlating techniques based on continuous flow." Drop in Wells Producing Oil and Gas," J. Cdn. Pet.
Tech. (July-Sept. 1972) 38~48.
Conclusions 4. Chierici, G. L., Ciucci, G. M., and Sclocchi, G.: ''Two-
Phase Flow in Oil Wells-Prediction of Pressure Drop,"
The reliability and relative merits were estimated for J. Pet. Tech. (Aug. 1974) 927-937; Trans., AIME, 257.
three recently published multiphase-flow pressure- 5. Beggs, H. D.·: "An Experimental Study of Two-Phase
drop prediction correlations applicable to vertical Flow in Inclined Pipes," PhD dissertation, U. of Tulsa,
Tulsa, Okla. (1972).
tubing. Well test data from several sources were
6. Espanol Herrera, J. H., Holmes; C. S., and Brown, K. E.:
selected to embrace broad ranges of flow rate, pipe "A Comparison of Existing Multiphase Flow Methods for
size, API gravity, gas/liquid ratio, and water/oil ratio. the Calculation of Pressure Drop in Vertical Wells,"
When applied to the entire data set of 726 tests, paper SPE 2553 presented at the SPE-AIME 44th Annual
Fall Meeting, Denver, Colo., Sept. 28-0ct. 1, 1969.
the correlation of Beggs and Bri1l2 was found to yield JPT
the least value of standard deviation. However, the
Aziz et at. 3 estimation technique resulted in the low- Original manuscript received in Society of Petroleum Engineers
office July 19, 1973. Revised manuscript received June 18, 1974.
est average difference between measured and calcu- Paper (SPE 4689) was first presented at the SPE·AIME 48th
lated pressure drops. Annual Fall Meeting, held in Las Vegas, Nev., Sept. 30·0ct. 3,
1973. © Copyright 1974 American Institute of Mining, Metal-
Figs. 1 through 5 reveal in greater detail the selec- lurgical, and Petroleum Engineers, Inc.

832 JOURNAL OF PETROLEUM TECHNOLOGY

You might also like