ELSEVIER Journal of Petroleum Science and Engineering 16 (1996) 275-290 Evaluation of empirically derived PVT properties for Pakistani crude oils Mohammed Aamir Mahmood, Muhammad Ali Al-Marhoun * Department of Petroleum Engineering, King Fahd Uniuersity of Petroleum and Minerals, Dhahran 31261, Saudi Arabia Received 3 February 1996; accepted 12 June 1996 Abstract This study evaluates the most frequently used pressure-volume-temperature (PW) empirical correlations for Pakistani crude oil samples. The evaluation is performed by using an unpublished data set of 22 bottomhole fluid samples collected from different locations in Pakistan. Based on statistical error analysis, suitable correlations for field applications are recommended for estimating bubblepoint pressure, oil formation volume factor (PVF), oil compressibility and oil viscosity. Keywords: physical fluid properties; PVT tests; correlations; least-squares methods; statistics 1. Introduction Provision of pressure-volume-temperature (PVT) parameters is a fundamental requirement for all types of petroleum calculations such as determi- nation of hydrocarbon flowing properties, and design of fluid handling equipments. More importantly, vol- umetric estimates necessitate the evaluation of PVT properties beforehand. The PVT properties can be obtained from an experimental set-up by using repre- sentative samples of the crude oils. However, intro- duction of a PVT empirical correlation also extends statistical techniques to estimate the PVT properties effectively. For the development of a correlation, geological and geographical conditions are considered impor- * Corresponding author. tant as due to these conditions the chemical composi- tion of any crude may be specified. It is difficult to obtain the same accurate results through empirical correlations for different oil samples having different physical and chemical characteristics. Therefore to account for regional characteristics, PVT correla- tions need to be modified for their application. Be- cause of the availability of a wide range of correla- tions, it is also beneficial to analyze them for a given set of PVT data belonging to a certain geological region. This study examines the existing PVT correla- tions against a set of PVT data collected from different locations in Pakistan as shown in Fig. 1. All of the significant PVT correlations reported in petroleum literature are included in this study. The validity and statistical accuracy are determined for these correlations and finally the best suited correla- tions are recommended for their application to Pak- istani crude oils. In addition, this study can be used 0920-4105/96/$15.00 Copyright 0 1996 Elsevier Science All rights reserved PII SO920-4105(96)00042-3
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ELSEVIER Journal of Petroleum Science and Engineering 16 (1996) 275-290
Evaluation of empirically derived PVT properties for Pakistani crude oils
Mohammed Aamir Mahmood, Muhammad Ali Al-Marhoun *
Department of Petroleum Engineering, King Fahd Uniuersity of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
Received 3 February 1996; accepted 12 June 1996
Abstract
This study evaluates the most frequently used pressure-volume-temperature (PW) empirical correlations for Pakistani crude oil samples. The evaluation is performed by using an unpublished data set of 22 bottomhole fluid samples collected from different locations in Pakistan. Based on statistical error analysis, suitable correlations for field applications are recommended for estimating bubblepoint pressure, oil formation volume factor (PVF), oil compressibility and oil viscosity.
Provision of pressure-volume-temperature (PVT) parameters is a fundamental requirement for all types of petroleum calculations such as determi- nation of hydrocarbon flowing properties, and design of fluid handling equipments. More importantly, vol- umetric estimates necessitate the evaluation of PVT properties beforehand. The PVT properties can be obtained from an experimental set-up by using repre- sentative samples of the crude oils. However, intro- duction of a PVT empirical correlation also extends statistical techniques to estimate the PVT properties effectively.
For the development of a correlation, geological and geographical conditions are considered impor-
* Corresponding author.
tant as due to these conditions the chemical composi- tion of any crude may be specified. It is difficult to obtain the same accurate results through empirical correlations for different oil samples having different physical and chemical characteristics. Therefore to account for regional characteristics, PVT correla- tions need to be modified for their application. Be- cause of the availability of a wide range of correla- tions, it is also beneficial to analyze them for a given set of PVT data belonging to a certain geological region.
This study examines the existing PVT correla- tions against a set of PVT data collected from different locations in Pakistan as shown in Fig. 1. All of the significant PVT correlations reported in petroleum literature are included in this study. The validity and statistical accuracy are determined for these correlations and finally the best suited correla- tions are recommended for their application to Pak- istani crude oils. In addition, this study can be used
0920-4105/96/$15.00 Copyright 0 1996 Elsevier Science All rights reserved
PII SO920-4105(96)00042-3
216 M.A. Mahmood, M.A. AI-Marhoun/Journnl of Petroleum Science and Engineering 16 (1996) 275-290
I COAL
Fig. 1. Location of mineral reserves in Pakistan.
as an effective guideline for correlation applications for all the other oil samples possessing similar com- positional characteristics.
2. PVT correlations
The frequently used empirical correlations for the prediction of bubblepoint pressure, oil FVF at bub- blepoint, two-phase FVF, undersaturated oil com-
pressibility, viscosity at and above bubblepoint, and dead oil viscosity are reviewed in the following sections.
2. I. Bubblepoint pressure correlations
Standing (1947) p resented a correlation for pre- dicting bubblepoint pressure by correlating reservoir temperature, solution gas/oil ratio, gas relative den- sity, and oil gravity. The gases in the oil samples contained CO, as the only non-hydrocarbon. The data used for this study were sampled from Califor- nia oil fields. Lasater (1958) for his correlation development acquired data without non-hydrocarbon gases. The oil samples were collected from Canada, the U.S.A., and South America. The aforesaid corre- lations were widely acclaimed and utilized for a considerably long time until Vazquez and Beggs (1980) reported their work for bubblepoint pressure prediction of a gas-saturated crude. They recom- mended a bifurcation for evaluating PVT parame- ters, and suggested two ranges ( yAp, < 30 and yAp, > 30) of oil samples. Glaso (1980) also presented a correlation for predicting bubblepoint pressure from a data set comprising of reservoir temperature, solu- tion gas/oil ratio, gas relative density, and oil grav- ity. The data for his study mainly belonged to the North Sea region. He also recommended a method for correcting a predicted bubblepoint pressure if a significant amount of non-hydrocarbon gases is pre- sent along with the associated surface gases. Al- Marhoun (1988) published his correlation for deter-
Table 1
Data ranges of existing correlations for oil FVF and bubblepoint pressure
M.A. Mahmood, M.A. Al-Marhoun/ Journal of Petroleum Science and Engineering 16 (1996) 275-290 217
mining bubblepoint pressure based on Middle East oil samples.
2.2. Oil FVF at bubblepoint pressure correlations
The very first correlation was developed by Standing (1947) utilizing the same data used for his bubblepoint pressure predication. Vazquez and Beggs (1980) reported their research recommending a bifur- cation in the data with two ranges of oil API gravity. Glaso (1980) also published a correlation which was based on Standing’s correlation with minor modifica- tions. He used 41 experimentally determined data points, mostly from the North Sea region. Al- Marhoun (1988) reported his correlation for which he acquired data from Middle East oil reservoirs. Al-Marhoun (1992) updated his correlation by ac- quiring a large data set of 4012 data points collected from all over the world. Table 1 shows the data ranges of the selected correlations discussed above.
2.3. Two-phase FVF correlations
Standing (1947) reported the first correlation for predicting two-phase FVF by correlating solution/gas oil ratio, temperature, gas relative den- sity, and oil gravity. Applying the same PVT param- eters used by Standing, Glaso (1980) published his correlation. Al-Marhoun (1988) reported his correla- tion using a data set collected from Middle East oil
The earliest research was conducted by Calhoun (1947) when he presented a graphical correlation for determining the isothermal compressibility of an un- dersaturated crude oil. Trube (1957) for his graphical correlation used pseudoreduced pressure and temper- ature to determine undersaturated oil compressibility. Vazquez and Beggs (1980) also presented a com- pressibility correlation using the available reservoir parameters.
2.5. Undersaturated oil viscosity correlations
Beal (1946) published his graphical correlations for determining the undersaturated oil viscosity of
crude oil by using a data set representing U.S. oil sample only. He used gas-saturated oil viscosity, bubblepoint pressure, and pressure above bubble- point as the correlating parameters. Vazquez and Beggs (1980) by using 3593 data points also pub- lished their correlation for undersaturated oil viscos- ity. Khan et al. (1987) published their correlation based on 75 bottomhole samples and 1503 data points obtained from Saudi oil reservoirs. The most recent correlation reported by Labedi (1992) for light crude oils is based upon Libyan crude oil data.
2.6. Gas-saturated oil viscosity correlations
Chew and Connally (1959) presented their work for predicting change in oil viscosity as a function of the solution gas/oil ratio. Their data set of 457 data points covered samples from South America, Canada, and the U.S.A. Beggs and Robinson (1975) acquired a large data set to obtain a correlation for predicting gas-saturated oil viscosity. Khan et al. (1987) re- ported their research using 150 data points obtained from Saudi crude oil samples. For light crude oils, Labedi (1992) presented his correlation using Libyan crude oil samples.
2.7. Dead oil viscosity correlations
Beal (1946) reported a correlation by applying 753 data points for his analysis. He correlated oil gravity, and temperature covering a range of lOO- 220°F. Beggs and Robinson (1975) presented their correlation using 460 dead oil observations. Glaso (1980) also developed a correlation using a tempera- ture range of 50-300°F for 26 crude oil samples. Ng and Egbogah (1983) presented their viscosity corre- lations by modifying the Beggs and Robinson corre- lation. Recently, Labedi (1992) has published a cor- relation for light crude oil sampled from Libyan reservoirs.
All of the correlations selected for this study are given in Appendix A.
3. PVT data acquisition for Pakistani crude oils
PVT reports of 22 bottomhole fluid samples were acquired from different locations in Pakistan for the evaluation purpose of this study. This unpublished
278 M.A. Mahmood, M.A. Al-Marhoun/Joumal of Petroleum Science and Engineering 16 (1996) 27.5-290
Table 2
PV’T differential data with the corresponding oil viscosity values
No. T P, B,, R, % “API /_q,
1 250 2885 2.916 2249 1.0608 56.5
2 248 1680 1.468 557
3 248 1415 1.432 486
4 248 1215 1.404 433
5 248 1015 1.378 381
6 248 815 1.352 328
I 248 615 1.322 273
8 248 415 1.292 215
9 248 227 1.246 144
10 248 133 1.214 96
II 248 15 1.092 0
12 245 3280 1.921 1340
13 188 4197 2.365 2371
14 248 1725 1.522 663
15 248 1515 1.493 603
16 248 1315 1.465 547
17 248 1115 1.438 490
18 248 915 1.409 432
19 248 715 1.380 376
20 248 515 1.350 316
21 248 315 1.314 251
22 248 183 1.278 192
23 248 113 1.248 152
24 248 15 1.098 0
25 229 1316 1.375 435
26 229 1065 1.350 379
27 229 865 I.329 335
28 229 665 1.306 288
29 229 465 1.282 239
30 229 265 1.250 182
31 229 163 1.227 145
32 229 15 1.087 0
33 222 2949 1.940 1321
34 222 2615 1.844 1210
35 222 2215 1.753 1074
36 222 1815 1.681 937
37 222 1415 1.610 802
38 222 1015 1.541 670
39 222 615 1.467 506
40 222 298 1.386 340
41 222 15 1.073 0
42 232 1525 1.460 550
43 232 1315 1.43 1 496
44 232 1115 1.403 446
45 232 915 1.376 395 46 232 715 1.348 342
47 232 515 1.320 288
48 232 315 1.286 228
49 232 185 I.253 180
50 232 15 1.097 0
51 217 1512 1.416 512 52 217 1315 1.391 468
53 217 1115 1.363 419
1.1955
1.2468
1.2955
1.3539
1.4272
1.5264
1.6611
1.8583
1.9810
0
1.0713
0.8253
1.3205
1.3692
1.424 1
1.4923
1.5775
1.6801
1.8180
2.0083
2.2297 2.4120
0
I .4030
I .4905
1.5762
1.6918
1.8545
2.0949
2.3000
0
1.2613
1.3003
1.3595
1.4356
1.5338 1.6640
1.8954
2.2520 0
1.3428
1.3898
1.4407
1.5022 1.5808 1.6839
I .8442 2.0370
1.1836 1.2194
1.2671
37.2
37.2
37.2
37.2
37.2
37.2
37.2
37.2
37.2
37.2
29.3
39.5
38.5
38.5
38.5
38.5
38.5
38.5
38.5
38.5
38.5
38.5
38.5
40.5
40.5
40.5
40.5
40.5
40.5
40.5
40.5
29.0
29.0
29.0
29.0
29.0
29.0
29.0
29.0
29.0
39.9
39.9
39.9
39.9 39.9
39.9 39.9
39.9 39.9
41.0 41.0
41.0
0.318
0.337
0.352
0.367
0.389
0.406
0.430
0.207
0.308
0.320
0.334
0.349
0.364
0.379
0.397
0.438
0.47 1
0.327
0.333
0.34 1
0.350
0.365
0.397
0.416
0.896
0.252
0.263
0.277
0.294
0.314
0.340
0.38 1
0.460
0.589
0.380
0.386
0.394
0.404 0.417
0.435 0.458
0.486 0.748
Table 2 (continued)
No. T P, B,,, R,
54 217 915 1.324 369
55 217 715 1.300 316
56 217 515 1.278 259
57 217 315 1.248 196
58 217 183 1.217 145
59 217 15 1.088 0
60 188 1717 1.394 556
61 188 1515 1.373 509
62 188 1315 1.354 462
63 188 1115 1.335 419
64 188 915 1.318 378
65 188 715 1.298 330
66 188 515 1.275 280
67 188 315 1.247 225
68 188 170 1.215 165
69 188 15 1.067 0
70 296 2883 2.619 1977
71 296 2615 2.475 1757
72 296 2315 2.331 1536
73 296 2015 2.203 1340 74 296 1715 2.092 1169
75 296 1415 1.995 1018
76 296 1115 1.910 884
77 296 815 1.832 760
78 296 515 1.747 628
79 296 249 1.633 470
80 296 152 1.599 379
81 296 104 1.504 317
82 296 15 1.142 0
83 281 4975 2.713 2496
84 281 4115 1.981 1458
85 281 3315 1.777 1074
86 281 2615 1.658 827
87 281 1915 I.552 615
88 281 1215 1.449 407
89 281 615 1.351 248
90 281 15 1.104 0
91 237 1226 1.418 470
92 237 1065 1.401 433
93 237 915 1.385 398
94 237 765 1.369 362
95 237 615 1.35 325
96 237 465 1.330 285
97 237 315 1.305 241
98 237 I83 1.275 190
99 237 114 1.253 I58 100 237 79 1.238 130
101 237 15 1.090 0
102 237 I295 I .349 357
103 237 I I65 1.335 330
104 237 1015 1.318 299
105 237 865 1.303 268
106 237 715 1.287 236
107 237 565 I.268 202
“API P”
1.3260
I .4037
1.5126
1.6882
1.8670
1.2595
1.3058
1.3614
1.423 1
1.4938
1.5954
1.73 1 1
1.9298
2.2450
1.407 I
1.4613
1.5337
1.6191
1.7167
1.8277
I .9523
2.095 1
2.281 I
2.5585
2.7812
2.9800
0
1.1545
1.1888
1.4410
1.6839
1.9220
2.5098
3.4445
0
1.5337 I .5922
1.6561
1.7323
1.8241
I .9424
2.0908
2.2778
2.4141
2.5500 0
I .2435 1.2758
1.3184
1.3687
I .4307 1.5137
41.0
41.0
41.0
41.0
41.0
41.0
42.6
42.6
42.6
42.6
42.6
42.6
42.6
42.6
42.6
42.6
39.9
39.9
39.9
39.9
39.9
39.9
39.9 39.9
39.9
39.9
39.9
39.9
39.9
31.9
31.9
31.9
31.9
31.9 31.9
31.9
3 1.9
39.4
39.4
39.4
39.4 39.4
39.4
39.4
39.4
39.4 39.4
39.4 39.5
39.5
39.5
39.5 39.5 39.5
0.301
0.310
0.318
0.328
0.338
0.352
0.367
0.386
0.411
0.878
0.222
0.232
0.243
0.254
0.266 0.278
0.292
0.309
0.332
0.365
0.386
0.402
0.769
0.205
0.245
0.275
0.310
0.350 0.405
0.482
0.914
0.330 0.338
0.345
0.356 0.372
0.388
0.4 IO
0.380
0.392
0.406 0.425
0.452 0.485
M.A. Mahmood, M.A. Al-Marhoun/ Journal of Petroleum Science and Engineering 16 (1996) 275-290 219
data set consists of 166 data points for evaluating bubblepoint pressure and oil FVF at bubblepoint pressure correlations. These data points are the re- sults of standard differential liberation tests con- ducted on bottomhole fluid samples collected di- rectly form oilfields. Table 2 shows the differential data set in detail, whereas Table 3 depicts the com- position and statistical analysis of the Pakistani crude data. The number of data points used for oil com- pressibility, two-phase FVF, oil viscosity (above, and at bubblepoint pressure), and the dead oil viscos-
Table 3
Data ranges of Pakistani crude oils
Parameter Range Parameter Range
FVF@P, 1.20-2.916 Y0 0.753-0.882
‘b 19-4915 &b 0.25-0.38
R, 92-2496 PO 0.206-0.548
API 29.0-56.5 0.581-1.589
C0 lo-5-10m4 2 0.23-1.4 P > P, 1115-6029 N, (mole%) 0.51-1.54
T 182-296 CT (mole%) 30.99-55.76
7, 0.825-3.445
280 M.A. Mahmood, M.A. Al-Marhoun/Journal of Petroleum Science and Engineering 16 (1996) 275-290
ity correlations are 246, 352, 104, 16 and 16, respec- tively.
In general, this data set covers a wide range of bubblepoint pressure, oil FVF, solution gas/oil ratio, and gas relative density values; whereas the tempera- ture and oil gravity belong to relatively higher values attributed to regional trends prevailing in Pakistani crude oils. This comprehensive data bank offers a good opportunity for further studies in this area.
4. Evaluation procedure
Statistical and graphical error analyses are the criteria adopted for the evaluation in this study. Existing PVT correlations are applied to the ac- quired data set and a comprehensive error analysis is performed based on a comparison of the predicted value with the original experimental value. For an in-depth analysis of the accuracy of the correlations tested, error analysis based on different ranges of oil API gravity is also carried out graphically. An error analysis based on oil API gravity ranges is consid-
70.00
60.00
6 5 F 40.00
‘Z
m 2 a 4 a 30.00
P 0 3 b 20.00
k
10.00
0.00
ered an effective tool for determining the suitability of the correlation for heavy, medium, or light oil.
The following statistical means are used to deter- mine the accuracy of correlations to be evaluated.
4.1. Average percent relatiue error (Er)
The average percent relative error is an identifica- tion of relative deviation of the predicted value from the experimental value in percent and is defined by:
E, = -!- 5 Ei
Izd i=l
(1)
where
E, = x 100 (i=1,2, . . . . n) (2)
The lower the value the more equally distributed is the error between positive and negative values.
4.2. Average absolute percent relative error (Ea)
The average absolute percent relative error indi- cates the relative absolute deviation of the predicted
+ Al-Marhoun 66
I I I I I
API434 34cAP1~38 38cAPk42 API>42
(16) (17) 6’8) (35)
Ranges of oil API gravity (with corresponding data points)
Fig. 2. Statistical accuracy of bubblepoint pressure correlation grouped by oil API gravity.
M.A. Mahmood, M.A. Al-Marhoun / Journal of Petroleum Science and Engineering 16 (1996) 275-290 281
Table 4
Statistical accuracy of bubblepoint pressure correlations
Correlation E,
Standing (1947) -43.5 49.18 0.43 391.05 68.37
Lasater (1958) -20.61 31.31 0.04 273.65 49.36
Vazquez and -52.07 55.31 0.16 403.99 70.30
Beggs (1980)
Glaso (1980) -24.82 32.08 0.04 247.00 45.64
Al-Marhoun 27.97 31.50 0.30 81.96 20.24
(1988)
value from the experimental values in percent. A lower value implies a better correlation. It is ex- pressed as:
'd i=l (3)
cent relative errors. The minimum and maximum
values are determined to show the range of error for each correlation and are expressed as:
Emin = $n I Ej I i= 1
and
E max = r&xlEil i= 1
4.4. Standard deviation (s)
(4)
(5)
The standard deviation is a measure of dispersion of predicted errors by a correlation, and it is ex- pressed as:
4.3. Minimum and maximum absolute percent rela- tive errors (Emi, and E,,,,,
S= (6)
Both the minimum and maximum values are de- termined by analyzing the calculated absolute per-
A lower value implies a smaller degree of scatter around the average calculated errors.
15.00
10.00
5.00
0.00
+ Standmg + Vaz RBegg
++- N-Marhoun 88
+ Al-Marhoun 92
I I I I
API<34 34<APk38 38cAPk42 API>42
(16) (17) (98) (35)
Ranges of API gravity (with corresponding data points)
Fig. 3. Statistical accuracy of oil FVF at bubblepoint pressure correlation grouped by oil API gravity.
282 M.A. Mahmood, M.A. Al-Marhoun/Journal of Petroleum Science and Engineering 16 (19961275-290
5. Results and comparison
Average absolute relative error is an important indicator of the accuracy of an empirical model. It is used here as a comparative criterion for testing the accuracy of existing correlations. After applying the existing correlations to the acquired data set, results in the form of average absolute relative error, aver- age percent relative error, minimum and maximum absolute percent relative error, and standard devia- tion are summarized in Tables 4-10. Another effec- tive comparison of correlations is performed through graphical representation of errors as a function of oil API gravity ranges. Figs. 2-8 represent correlation errors for four oil API gravity ranges.
Table 5 Statistical accuracy of oil FVF at bubblepoint pressure correlation
Correlation E, E, E mln Em s
Standing ( 1947) 1.39 2.31 0.05 7.96 2.36
Vazquez and 12.84 12.84 5.99 24.83 4.37
Beggs (1980)
Glaso (1980) 3.65 3.88 0.08 12.78 2.23
Al-Marhoun (1988) 2.27 2.34 0.01 13.0 2.55
Al-Marhoun (1992) 0.76 1.23 0.01 9.09 1.54
gravity; whereas the maximum error is obtained for a higher gravity range of 42 oil API gravity and above as depicted by Fig. 2.
5. I. Bubblepoint pressure correlations 5.2. Oil FVF at bubblepoint pressure correlations
Lasater (1958) together with Al-Marhoun (1988) Al-Marhoun (1992) exhibited a significantly uni-
showed least errors for the data used as shown in form error for all oil API gravity ranges as shown in Table 4. The least error of all the tested correlations Fig. 3. Corresponding to the least error obtained for
is obtained for a medium range of 34-38 oil API this correlation, a least value of standard deviation is
25.00 -
3 m w b 20.00 -
5 al .z g P 15.00 - al 4 E
z k% 10.00 - P F Q
5.00 -
+ Standmg
+ Glaso
-+- Al-Marhoun
“. VW
API434 34<AP1<38 38<AP1<42 API>42
(40) (29) (200) (83)
Ranges of oil API gravity (with corresponding data points)
Fig. 4. Statistical accuracy of two-phase FVF correlation grouped by oil API gravity.
M.A. Mahmood, M.A. Al-Marhoun / Journal of Petroleum Science and Engineering 16 (1996) 275-290 283
Table 6 Table 7
Statistical accuracy of two-phase FVF correlations Statistical accuracy of undersaturated oil compressibility correla-
shown in Table 5. This is also supported by Petrosky and Farshad (1993) when they showed that Al- Marhoun (1988) obtained better accuracy for Gulf of Mexico data.
5.3. Two-phase FVF correlations
Glaso (1980) obtained reasonable result with a least error as shown in Table 6. However, this correlation overestimates the predicted value com- pared to the experimental value. Fig. 4 shows the same trend of errors for Standing (1947) and Glaso (1980) for all oil API gravity ranges.
Calhoun (1947) showed a good harmony with the data used, but this correlation tends to underestimate the predicted compressibility value as shown in Table 7. This correlation gives least error for the medium oil API gravity range of 34-38, as shown in Fig. 5. This result is also favored by Sutton and Farshad (1990) through their research conducted on Gulf of Mexico data.
45.00 -
40.00 -
g m Y 35.00 -
b 5 LZ 30.00 - ‘Z 1 e! a, 4 25.00 -
a
D 8 20.00 -
E k 15.00 -
10.00 -
API<34 (25)
34cAPlc38
(22)
38<AP1<42 (139)
API>42 (60)
Ranges of oil API gravity (with corresponding data points)
Fig. 5. Statistical accuracy of undersaturated oil compressibility correlation grouped by oil API gravity
284 M.A. Mahmood, M.A. Al-Marhoun/Joumal of Petroleum Science and Engineering 16 (1996) 275-290
Table 8 Table 9
Statistical accuracy of undersaturated oil viscosity correlation
Correlation E, E, Em,, Em s
Beal ( 1946) - 2.94 4.52 0.03 14.89 4.71
Vaaquez and - 14.01 14.15 0.08 46.39 12.54
Beggs (1980)
Khan et al. (1987) -7.61 7.91 0.10 26.59 6.64
Labedi (1992) -5.82 7.45 0.02 47.56 8.98
Statistical accuracy of gas saturated oil viscosity correlation
Correlation E, E., Em Em s
Beggs and -24.43 26.71 2.56 57.16 21.70
Robinson (1975) Chew and -3.41 12.21 1.27 25.31 13.62
Connally (1959) Khan et al. (I 987) - 18.60 29.92 1.19 64.80 30.81 Labedi (1992) -29.65 37.53 I .56 268.98 70.04
5.5. Undersaturated oil viscosity correlations
Beal (1946) showed better results than the other correlations tested. Table 8 shows a least standard deviation value for this correlation. This correlation is best suited to a low oil API gravity as shown in Fig. 6. Prediction by Labedi (1992) is also reason- able for a high oil API gravity range. All of the correlations unanimously overestimated the viscosity
values.
corresponding least scatter. This correlation is equally good for all oil API gravity ranges as shown in Fig. 7. With the exception of Labedi (1992) all correla- tions showed least error for high oil API gravity ranges but overestimated the viscosity values.
5.7. Dead oil viscosity correlations
5.6. Gus-saturated oil ciscosiv correlations
Chew and Connally (1959) is the best among others as shown in Table 9 with a least error and a
The Glaso (1980) correlation is found relatively better for gravity higher than 34 oil API gravity as shown in Fig. 8. All of the correlations obtained
large errors for low oil API gravity. Except Beal ( 1946) all of the correlations overestimated dead oil viscosity values as shown in Table 10.
6. Conclusions
The following conclusions can be drawn by this
evaluation study. (1) Although high errors are generally obtained
for the prediction of bubblepoint pressure, the error
obtained was extremely high in this case. This stresses the need of a new bubblepoint pressure correlation representing the chemical and geological difference of this region. Both Lasater (1958) and Al-Marhoun (1988) showed nearly equal errors but the latter exhibited a least standard deviation. Any one of these correlations may be used for Pakistani crude oils.
(2) For oil FVF correlations at bubblepoint pres- sure, all of the selected correlations showed a good degree of harmony towards the data used. All of the correlations underestimated FVF values, i.e. the pre- dicted value is less than the actual experimental value. Due to its least error and least standard devia- tion Al-Marhoun (1992) correlation is recommended for this type of PVT data. This correlation is also favored as it covers the same range of oil FVF, bubblepoint pressure, and temperature found in the Pakistani crude oil data.
(3) For two-phase FVF, all of the correlations are best applicable to the medium range of oil API gravity. Glaso (1980) is recommended for crude oil having this type of characteristics.
175.00 -
g 150.00 -
m % b 6 125.00 -
al .z m -F 100.00 -
a, s 5 B 75.00 -
8
; k 50.00 -
25.00 -
-.- Beggs & RobInson
+ Chew & Connally
I- Khan et al
0.00 ’ I I I I
API<34 34<APl<38 38<AP1<42 API>42
(2) (2) (9) (3)
Ranges of oil API gravity (with corresponding data points)
Fig. 7. Statistical accuracy of gas-saturated oil viscosity correlation grouped by oil API gravity.
286 M.A. Mahmmd, M.A. Al-Murhoun /.loumal of Petroleum Science and Engineering 16 (1996) 275-290
I * 250.00
t \ t- \
E a, a, .z 150.00 - m P + Glaso al ‘j - Ng & Egbogah
5
4 100.00 -
%
c $
50.00 -
0.00
API<34 34<AP1<38 38<AP1<42 API>42
(2) (2) (9) (3)
Ranges of oil API gravity (with corresponding data points)
Fig. 8. Statistical accuracy of dead oil viscosity correlatiun grouped by oil API gravity.
(4) Most of the compressibility correlations are good for medium and low oil API gravity ranges and showed large errors towards light oils. The evalua- tion process shows that Calhoun (1947) is a better choice than the other correlations.
(5) Most of the correlations for viscosity above bubblepoint pressure are good for heavy oils and exhibit large error for medium ranges of oil API gravity. Beal (1946) is recommended for the oil samples used, as it gives the least error and least scatter. This correlation is also suitable due to its comparable range of pressure above bubblepoint with the data used.
(6) Most of the viscosity correlations at bubble- point pressure performed better for heavy oils. Chew and Connally (1959) is recommended the best corre- lation as it gives least error for all oil API ranges.
(7) All of the dead oil viscosity correlations are found relatively more accurate for medium to high oil API gravity ranges. Based on a least error analy- sis, Glaso (1980) is recommended for application. This correlation is suitable also as its temperature
range matches with that of the samples used for this study.
(8) In conjunction with the standard error analy- sis, an error analysis based on oil API gravity ranges proved to be an effective tool for determining the suitability of the correlation for heavy, medium, or light oil. Thus, this type of analysis is strongly recommended for all evaluation studies of this na- ture.
M.A. Mahmood, MA. Al-Marhoun/Journal of F’etroleum Science and Engineering I6 (1996) 275-290 287
log = -
Ifp:
P, = R, =
S= T= x=
YAPI =
Yg =
% =
%b =
log10 number of data points pressure, psi &Pa) bubblepoint pressure, psi &Pa) solution gas/oil ratio, SCF/STB
(m3/m3> standard deviation temperature, “F (K) variable representing a PVT parameter stock tank oil gravity, “API gas relative density (air = 1) oil relative density (water = 1) bubble point oil relative density (water = 1)
pod = dead oil viscosity, CP
&b = gas-saturated oil viscosity, CP
CL, = undersaturated oil viscosity, CP
Subscripts:
c=
critical pr = pseudoreduced est = estimated from the correlation exp = experimental value
8. SI metric conversion factors
“API 141.5/(131.5 + “API) = g/cm3 bbl bbl X1.589837. 10-l = m3 CP CP x 1.0. 1o-3 a = Pa s “F (“F - 32)/1.8 = “C psi psi X 6.894757 = kPa “R “R/1.8 =K scf/bbl scf/bbl X 1.801175 . 10-l = std m3/m3
a Conversion is exact.
Acknowledgements
We thank the management of Oil and Gas Devel- opment Corporation (OGDC, Pakistan) for providing
(A-1)
the data for this research. We are also grateful to the Department of Petroleum Engineering at King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia, for its excellent research and comput- ing facilities, made available for this study.
Appendix A. Existing PVT correlations
The PVT correlations evaluated in this study are given below.
A.1. Bubblepoint pressure correlations
A.l.l. Standing (1947)
P, = 18( Rs/y,)0~8310y~
where
Y, = 0.00091T - 0.0125yAp,
A.1.2. Lasater (1958) I
Yp = (R,,‘379.3),‘[( RJ379.3) + (35Oy,,‘M,)]
(A-2a)
‘b = [(Pt-)(Tf460)]/y, (A-2b)
A.1.3. Vazquez and Beggs (1980)
p, = {(~,Rs~y,)~~l~c~~~PI/~~+~~~~l}1’C2 (A-3)
for yAp, I 30:
C, = 27.64
c, = 1.0937
C, = 11.172
for y > 30:
C, = 56.06
c, = 1.187
c, = 10.393
A.1.4. Glaso (1980)
P,= 10. I 7669t 1.7447 log A$-0.3021X(log Np$ (A-4)
’ Refer to the figures presented in the original work.
288
where
M.A. Mahmood, M.A. Al-Marhoun/ Journal of Petroleum Science and Engineering 16 (19961 275-290
where
Np, = (~,/yg)"~~'670.17' _ yO 989
A. I .5. Al-Marhoun (I 988)
3 0 715082 P, = 5.38088 x IO--R; y -1.X77840 3.143700 8 x