Académique Documents
Professionnel Documents
Culture Documents
www.fuelfirst.com
Abstract
A molecularly explicit property prediction mode (MEPP) is developed for predicting the properties of light petroleum fractions from ASTM
D86 distillation data. The method uses a molecularly-explicit characterization model (MECM) based on optimization to define the molecular
composition of the petroleum fuel from which more than 20 thermodynamic and transport properties are calculated. The model is tested using a set
of 68 pure components to characterize and predict the properties for 30 different petroleum naphtha samples and is found to be in excellent
agreement with the current conventional prediction methods based on global properties. The method is suitable for incorporation into the new
generation of molecularly-explicit models for simulating the kinetics and dynamics of petroleum refining processes and has the flexibility to be
tailored to use any set of pure components as desired.
q 2005 Elsevier Ltd. All rights reserved.
Keywords: Characterization; Property estimation; ASTM D86 distillation
1. Introduction
Accurate characterization of petroleum fuels is an important
step in the application of kinetic and thermodynamic
calculations for the design, operation, and simulation of
petroleum refining processes. An insufficient description of
heavier hydrocarbons (e.g. pentane and heavier; CC
5 ) reduces
the accuracy of predictions. Ideally, fuels properties are
determined experimentally in the laboratory on actual fluid
samples taken from the process under study. Because of the
expense of the experimental determination of such data, there
is interest in their accurate prediction.
In order to speed up the execution of real-time simulation, a
series of simplified correlations has been proposed for the
evaluation of physical properties of petroleum fractions during
the last five decades [1]. Several charts and correlations were
developed many of which require as input parameters the fuels
bulk or global properties such as the average boiling point, the
specific gravity and some characterization factors. However,
wide boiling range fractions are mixtures of a large number of
hydrocarbon compounds the type of which varies along the
distillation curve, therefore, a single value for boiling point or
* Tel.: C965 664 0663; fax: C965 483 9498.
E-mail address: albahri@kuc01.kuniv.edu.kw.
URL: http://www.albahri.info.
0016-2361/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fuel.2005.09.005
MECM
Model
749
C3 Detailed
composition
of 68
predefined
C11 molecules
Aggregation
Model
API
MW
RI
H/C
Viscosity
SurfaceTension
Otherproperties
750
Table 1
Comparison of global properties of light reformate calculated from generalized correlations and from aggregation of pure components using MEPP model
Specific gravity
Vapor pressure @ 37.8 8C, bar
Reid vapor pressure, bar
Cubic average boiling point, 8C
Mean average boiling point, 8C
Volume average boiling point, 8C
Weight average boiling point, 8C
Mole average boiling point, 8C
Watson characterization factor
Molecular weight, gm/mole
Refractive index
Hydrogen content, wt frac
Kinematic viscosity @ 98.9 8C, mm2/s
Kinematic viscosity @ 37.8 8C, mm2/s
Surface tension @ 25 8C, dynes/cm
Aniline point, 8C
Critical temperature, 8C
Pseudo critical temperature, 8C
Pseudo critical pressure, bars
Critical compressibility factor
Paraffins content, mole%
Naphthenes content, mole%
Aromatics content, mole%
Acentric factor
Freezing point, 8C
Research octane number
Motor octane number
Heat of vaporization @ NBPb, J(abs)/gm
Net heat of combustion @ 25 8C, J(abs)/gm
Isobaric liquid heat capacity @ 15.6 8C, J(abs)/gm K
Isobaric vapor heat capacity @ 15.6 8C, J(abs)/gm K
Liquid thermal conductivity @ 25 8C, J(abs)/s m2 K/m
a
b
Experimental or predicted
from experimental
MECM simulationa
% Error (deviation)
0.7143
0.7033
0.655
87.7
84.4
89.3
90.3
81.6
12.09
93.5
1.3973
0.1458
0.32
0.48
20.44
37.8
265.6
254.4
31.03
0.2773
71.50
17.19
11.3
0.284
K102.3
72.3
70
322.08
44,245
2.437
1.882
0.1099
0.7083
0.7033
0.655
82.6
81.1
84.1
87.1
79.6
12.18
93.1
1.3960
0.1507
0.31
0.49
19.48
57.1
261.3
254.8
31.23
0.2664
70.88
17.43
11.7
0.285
K104.7
73.2
71.7
323.06
44,047
2.039
1.604
0.1161
K0.84
0.00
0.00
K1.44
K0.94
K1.45
K0.92
K0.57
0.77
K0.42
K0.09
3.34
K3.56
1.70
K4.70
6.21
K1.51
0.14
0.67
K3.91
K0.87
1.43
3.34
0.32
K1.44
1.23
2.37
0.31
K0.11
K16.38
K14.79
K9.08
751
START
Preseleced set of
pure components
ASTMD86
DistillationData
+ PNA+RVP
Pure Component
Data from Literature
or Predicted
Initial Estimation
of composition
Convert
to
TBPcurve
Mixing
Rules
END
Comparison&
ErrorAnalysis
Global
Properties
Optimization
Model
Objective
Function
Optimal Composition
of the Simulated
Petroleum Fraction
Mixing
Rules
Fig. 2. Chemical logic diagram for the methodology used to develop the property model.
Table 2 and the bar plots are also shown in Fig. 3. The fit
between the experimental and stimulation results as shown by
the average percentage errors and correlation coefficients is
quite good. The average absolute deviation for all the
properties in Table 2 is about 3.5% and the average correlation
coefficient is 0.96.
Table 2
Error analysis for some of the properties investigated.
No.
Property
Value range
Av. % error
Correlation coefficient
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
API gravity
Cubic average boiling point, 8C
Mean average boiling point, 8C
Volume average boiling point, 8C
Molar average boiling point, 8C
Weight average boiling point, 8C
Watson characterization factor
Molecular weight, gm/mole
Refractive index
Hydrogen content, wt fraction
Motor octane number
Kinematic viscosity at 98.9 8C, mm2/s
Kinematic viscosity at 37.8 8C, mm2/s
Surface Tension of liquid at 25 8C, dynes/cm
Aniline point, 8C
True critical temperature, 8C
Pseudocritical temperature, 8C
Pseudocritical pressure, bar
Heat of vaporization at NBP, J(abs)/gm
Net heat of combustion at 25 8C, J(abs)/gm
Freezing, 8C
Acentric factor
Critical compressibility factor
Flash point, 8C
Liquid thermal conductivity at 25 8C, J(abs)/s m2 K\m
3594
34134
32133
35134
30134
35134
10.6113.06
68.2110.8
1.35461.4764
0.10630.1716
6097
0.250.43
0.320.7
14.5127.96
2065
196338
191338
26.235.9
311.6363.7
43,99645,362
(K119)(K72)
0.2440.312
0.27500.2805
(K74)(23)
0.10200.1211
2.14
1.34
0.99
1.34
0.83
1.07
0.80
2.06
0.21
2.57
4.15
4.04
5.41
2.67
1.74
0.93
0.80
2.22
1.17
0.18
5.38
3.13
0.25
5.16
1.78
0.990
0.995
0.995
0.995
0.995
0.996
0.970
0.990
0.993
0.965
0.850
0.960
0.972
0.995
0.830
0.991
0.989
0.900
0.948
0.960
0.946
0.832
0.924
0.976
752
(b) 40.0
(a) 140
35.0
120
30.0
100
25.0
80
20.0
60
API
MeABP
MW
MON
40
20
Surface Tension
10.0
Pc (pseudo)
5.0
0
0
(c)
Kw
15.0
0.8
0.7
15
25
35
1.6
1.55
0.6
1.5
0.5
1.45
0.4
H content
Viscosity 37.8C
Viscosity 98.9C
Acentric factor
Zc
Thermal cond.
0.3
0.2
0.1
0
0
(e)
0.2
0.4
0.6
1.4
RI
?Hc/30,000
1.35
1.3
1.3
0.8
(f)
1.4
1.5
1.6
400
40
350
20
0
300
20
250
40
60
80
100
100
50
Tc (true)
Tc (pseudo)
?Hv
200
Flash Point
50
150
150
200
250
300
350
400
Fig. 3. Bar plots (x-axis is property value calculated from generalized correlations, y-axis is property value calculated by MEPP model).
753
46000
45000
44000
Gorenkov et al.
API method
43000
43000
44000
45000
46000
Heat of Combustion from generalized
correlations (J/kg)
Fig. 4. Bar plot comparison for the net heat of combustion using the methods of
API and Gorenkov et al.
the net heat of combustion in Fig. 4. The figure shows the bar
plot for using both the API method [1] and the method of
Gorenkov et al. [8]. It is evident that the latter method provides
better correlation with the molecular ensemble and can be
considered more accurate than the API correlation.
4. Conclusion
4.1. Summary
It was an objective of this work to develop a model for
predicting the properties of undefined multicomponent
mixtures and light petroleum fractions like naphtha and
gasoline from ASTM distillation data alone using a characterization method that is more suitable for incorporation in the
molecularly-explicit simulation models than the current
methods and to enhance the prediction performance of
chemical process simulation packages.
The developed technique combines routine analytical tests
and a molecularly-explicit modeling approach to provide
quantitative insight to the petroleum fractions structure which
permitted easy accounting of molecules and enabled the direct
estimation of the thermodynamic and transport properties
thereof. The method was used to predict the properties of 30
petroleum naphtha samples with good agreement with the
experimental data. The said method can calculate the properties of light petroleum fractions with good accuracy when at
least one bulk property is available (namely ASTM D86 or
TBP distillation temperatures). However, these properties can
be predicted more accurately, if in addition RVP and PNA are
available. Although the technique is shown to predict properties of petroleum naphtha as an example, the method in
principle is applicable to any petroleum fraction.
This work further demonstrates that the complex nature of
petroleum fuels may be modeled by a limited set of
representative molecules using non-linear-regression optimization models. Considering the difficulty and complexity of
accounting for the thousands of compounds in petroleum fuels
and the limitations of the conventional pseudo-component
754