Heavy feedstocks present difficult operational challenges for refiners that can add to safety risks and reduce profitability. Processing heavy crudes safely and profitably can require development of new equipment or major changes in operating conditions.
Innovative new methods, which model heavier feedstock processing more accurately, enable refiners to adapt their processes more easily.
Register now to learn more about this important new technology.
Who should attend: Plant Managers, Process Engineers, Engineering Managers, Operations Managers, Process Design Engineers
View OnDemand at: www.real-time-answers.com/refinery
4. Crude Oil
Yesterday and Today
• Initially oil was developed
based on it ease of access
and quality
– Light Sweet Crude
– Relatively shallow fields
• However…"The easy oil is
coming to an end“*
– Production in the North Sea
is on the Decline
– Major oil fields in the Gulf
region have pumped more
than half their oil
Drumbeat: October 5, 2009, The Oil Drum
*Alex Munton, Analyst for Wood Mackenzie
5. Crude Oil Demand
• Demand keeps increasing
–Growth in China and India
jumped by 2.3 MBPD barrels
in 2010
–Current consumption is +87
MBPD and expected to
increase 2-3% per year
–Prices per BBL remaining at
$80 to $100
EIA, International Energy Outlook 2011
6. Heavy Oil / Bitumen
Gaining Momentum
• Largest % of Reserves
• Aggressively being
sought out as prices
remain elevated.
• Resources are estimated
+8 Trillion Barrels Bitumen
30%
Conventional
30%
–That’s about 100 years of
global consumption at
current levels. Heavy 15%
Extra
Heavy 25%
Alboudwarej, Hussein, "Highlighting Heavy Oil“
Oilfield Review (Schlumberger).
7. Heavy Oil / Bitumen Deposits (MBBL)
529
3041
92
482
178
971
18
129
72
3359
Heavy Oil USGS, Heavy Oil and Natural Bitumen
Resources in Geological Basins of the World
Bitumen
8. Characteristics
• High asphaltene content
–Incorporates roughly 90 percent
of the sulfur and metals in the oil
• Dense, viscous nature
(similar to molasses),
• Contains impurities such as
waxes and carbon residue.
• API Gravity from 22° to
below 10°
3D Picture of
Venezuelan Crude
Asphaltene Molecule
9. Challenges – Technology &
Infrastructure
Production
• Location (e.g. Tar Sands)
• Extraction Processes
Transportation
• Limited Pipelines
• Flow capability
Processing
• Old Equipment Designs
• Impurity Removal
• Inconsistency in composition
10. Investment is Ongoing
• Extraction Technologies
– Expect $20 Billion USD in Western
Canada alone
– Mining Technologies / In Situ
development, e.g. SAGD
• Billions in Transportation
– Keystone XL Pipeline
• Billions in Refining Investments
– $1.6 Billion upgrade at Valero Port Arthur
– $3.7 Billion expansion Cenovus
/ConocoPhillips Wood River
11. The Role of Modeling
• Models serve a critical role in
the production, transportation
and processing of crude oil
– Design: Proper sizing and costing
of equipment
– Operate: Used to study changes
in the process, changes in the
feedstock, Operating Training
simulation
– Optimize: Reduce Energy,
Increase Product Value, Improve
Procedures
12. Getting It Right
• Years of experience on
light crudes
• Heavy Crudes are a
challenge
–Lack of experience
–Complex relationships
–Impurities
Without the proper models, companies run the
risk of not getting the most out of their
Investment!
14. Agenda – Heavy Oils Webinar
• Functionality currently available to improve simulations of processes
with heavy oils and bitumens
• Improved characterization of high boiling point petro components (Tb > 1,000K)
• Improved accuracy in prediction of liquid viscosity
• Improved accuracy in prediction of liquid thermal conductivity
• EoS binary interaction parameters for accurate prediction of Hg solubility in light
hydrocarbons
• EoS binary interaction parameters for accurate prediction of H2 and H2S
solubility in hydro-processing applications
• Functionality under development
• Characterization of asphaltenes and prediction of phase equilibrium
• Molecular-based characterization as alternative to petro-components
• New properties from structure method for improved accuracy of thermo-
physical properties
15. Heavy Oil Method for Characterization
• The original SIMSCI (Twu)
correlation was developed for
Normal Boiling Points (NBP)
of up to 1000 K (1340 oF).
• The Heavy Oil method
extrapolates critical properties
and molecular weight to NBPs
above 1000 K.
• Property at 1000K is used as
reference and extrapolation is
based on constant Watson K
to NBP of pseudo component
16. Method for Improved Accuracy in Prediction
of Liquid Viscosity
• Based on correlation published by Twu* [1985]
• Uses principle of corresponding states
– Normal alkane reference series
– Kinematic viscosities at 100oF and 210oF are estimated for n-alkane with same boiling
point as fluid of interest
– Kinematic viscosities at 100oF and 210oF are calculated for fluid of interest based on
perturbation from reference series, based on specific gravity and boiling point.
• Walther equation is fitted to estimate temperature dependence and
used to interpolate or extrapolate
*Twu, C.H., Ind. Eng. Chem. Process Des. Dev., 24 (1985)
1287-1293
Slide 16
17.
18. Effect of Accuracy of Viscosity Prediction on
Heat Exchanger Performance
Crude
20.98oAPI
735oF
MeABP
11.43
Watson K
103oF
400 psig
Vac Resid
4.6oAPI
1182oF
MeABP
11.35
Watson K
Slide 18
278oF
250 psig
19. Effect of Accuracy of Viscosity Prediction on
Heat Exchanger Performance
Viscosity(L) Measured API Procedure Heavy Oil
method Values 11A4.2 (2011) Prediction
Crude
Duty (106 BTU/hr) 3.654 8.976 5.432
20.98oAPI
735oF
MeABP LMTD (oF) 165 150 159
11.43 U (BTU/hr/ft2/oF) 4.657 12.62 7.144
Watson K Shell Side
103oF Tout (oF) 265 245 258
400 psig
ΔP (psi) 58 8 27
Tube Side
Tout (oF) 110 119 113
ΔP (psi) 9 13 11
Vac Resid
4.6oAPI
1182oF
MeABP
11.35
Watson K
Slide 19
278oF
250 psig
20. Effect of Accuracy of Viscosity Prediction on
Heat Exchanger Performance
Temperature, oF Crude Kinematic Viscosity, cSt
Measured API 11A4.2 Heavy Oil
Crude
20.98oAPI 60 49.75 111.6 70.65
735oF 80 27.95 58.77 39.69
MeABP 100 17.52 34.26 24.38
11.43
Watson K Vac Resid Kinematic Viscosity, cSt
103oF 210 6787 454.0 4784
400 psig
250 1261 147.5 858.5
300 274.6 50.83 176.8
Vac Resid
4.6oAPI
1182oF
MeABP
11.35
Watson K
Slide 20
278oF
250 psig
21. Conclusions: SimSci Heavy Oil Method for
Liquid Viscosity Prediction
• Method results in a significant improvement to accuracy of
prediction of liquid viscosities
– Marked improvement in the 100 cSt to 100,000 cSt range
– No reduction in accuracy in the 0.1 cST to 100 cSt range
• Temperature dependence, especially at low temperatures is
improved
• Observations:
– Method is very sensitive to value used for normal boiling point (method
uses SimDis 50% point)
– It appears best to treat undefined portion of fluid (petro fractions) as
single component and avoid applying mixing rules to individual component
viscosities
– Accuracy may be improved further if more heavy oil viscosity data are
made available
– Some correlation between viscosity and asphaltene content was expected,
but not supported by current data set
Slide 21
22. Prediction of Liquid Thermal Conductivity
• Negligible amount of published data for liquid thermal conductivity
for bitumens and heavy oils
– Data from Wiltec* for narrow boiling fractions from SRC-II coal liquids
• Fortunately, thermal conductivity (λL) for most organic liquids varies
between 0.10 and 0.17 W.m-1.K-1and varies linearly with temperature
– No successful theory, to date, for predicting λL
– Approximate techniques are recommended for engineering calculations
• Heavy Oil Method (modified Sato-Riedel)
λL = const1 + const2(1 – Tr)2/3
Heavy
Oil
*Gray, J.A., Brady, C.J., Cunningham, J.R., Freeman, J.R., Wilson, G.M., Ind. Eng. Chem. Process Des. Dev. 1983,
22, 410-424
23. Conclusions: Liquid Thermal Conductivity
• The best and the worst:
– Riazi & Faghri method predicts wrong sign of temperature gradient for
heaviest fractions
– Heavy Oil method (based on Sato-Riedel) requires only estimate of
critical temperature, Tc and appears to be the most accurate
• Very little data available
– No data on bitumens
– Some evaluated data on pure components from NIST via TDE
– Coal liquid data (used as surrogate for bitumen) are suspect
Slide 23
24. Adverse impact of Mercury
• Process impairment
• Process equipment failure (corrosion / cracking)
• Catalyst poisoning
• Adsorption bed malfunction
• Wastewater contamination
• Mercury carryover in flare systems
25. Mercury Solubility in n-Alkanes at 30°C
1.8
Mercury Solubility (ppm-molar)
1.6
1.4
1.2
1 Measured
ZERO Kij's (SRK EOS) ‐ Uncorrected Mercury Vapor Pressures
0.8 CUSTOM Kij's (SRK EOS) ‐ Uncorrected Mercury Vapor Pressures
ZERO Kij's (SRKM EOS)
0.6 CUSTOM Kij's (SRKM EOS)
0.4
0.2
5 6 7 8 9 10
Carbon Number of n-Alkane
26. Results
Mercury — n‐Alkane (nC1 – nC10) binary interaction parameters available for
low pressure, low temperature (0-65 °C) applications
SRK, SRKM, PR and PRM Equations of State
• Predicted solubility validated against thermodynamic and plant data
SRKM and PRM Equations of State
•Additionally, EOS α-parameters fine-tuned to match mercury vapor
pressures to 1500 K
27. Hydrogen Solubility Prediction
• Heavy Oils require increased Hydroprocessing to remove trace
impurities (Sulfur, Nitrogen etc)
• Supply and Management of Hydrogen is critical to Refinery
optimization and meeting stringent Sulfur specification
• Current techniques are inaccurate as T & P dependence not
considered
– Very little published data with Heavy Fractions
• Invensys SimSci-Esscor is taking to lead in this emerging area
– Improve prediction of H2 solubility in hydrocarbons
– Improve prediction of H2S solubility
– Improve prediction of NH3 solubility
30. Solubility of hydrogen in n-decane – no kijs
Invensys proprietary &
Slide 30
confidential
31. Solubility of hydrogen in n-decane – with kijs
Invensys proprietary &
Slide 31
confidential
32. Observations and Conclusions
• For H2 – hydrocarbon binaries, kij appears to be a linear function of
temperature
– kij should take values greater than unity at high temperatures
• For H2 – hydrocarbon binaries, kij required for an aromatic
component appears to be larger than kij for similar molecular weight
paraffin
• For H2S – hydrocarbon binaries, kij is poorly correlated with
temperature
‒ Very low R2 values
• Work is continuing for NH3 – hydrocarbon binaries
• Work is continuing to generalize kij for undefined narrow-boiling
fractions (petro components), based on molecular descriptor(s)
– Boiling point, API gravity and/or molecular weight
34. Asphaltenes
• Bitumen and heavy oils are rich in Structures of Hypothetical Asphaltene
asphaltenes Components*
• Heaviest, most polar fraction
• Insoluble in n-alkanes, soluble in aromatics
• Self-associating, forming aggregates
• Contribute significantly to viscosity and
coking tendency
• Asphaltenes are precipitated to lower
viscosity and make more easily refined
product (e.g. VAPEX processes)
• Heavy oils are diluted with solvent (e.g.
condensate) to lower viscosity for
transport
• To optimize processes, it is necessary
to have accurate prediction of amount
of asphaltene precipitated as function
of amount of solvent temperature and
pressure *From University of Utah
35. Calculation of Asphaltene Phase Behavior
• Bitumen is divided into four SARA fractions, Saturates (S),
Aromatics (A), Resins (R), Asphaltenes (A)
• Asphaltene fraction is further divided into fractions with different
molar masses based on gamma distribution
• Correlations from Akbarzardeh* et al. provide estimates of solubility
parameters and molar volumes of asphaltene fractions
• Flory-Huggins version of Regular Solution Theory is applied to
predict activity coefficients of asphaltene fractions
• Assume liquid-liquid equilibrium
• Heavy, asphaltene-rich phase with only asphaltenes and resins
• Light oil-rich phase including all components
• Parameters of gamma distribution are adjusted to match onset of
precipitation conditions
*Kamran Akbarzadeh, Hussein Alboudwarej, William Y. Svrcek, Harvey W. Yarranton,
Fluid Phase Equilibria 232 (2005) 159-170
36. Asphaltenes Characterized as Low Molecular
Weight Polymers
Shultz-Zimm distribution
(special case of Gamma distribution)
f(x,α,β) = βα/Γ(α). x(α-1).e-βx, mean = α/β
x = (r – 1)/(ravg – 1)
r = number of monomer units
ravg = average number of monomer units
11/7/2012
36
38. From: Narrow Boiling Fraction Representation
of Middle Distillate
NBP 540 (Tb = 540K, S = 0.8876)
Tpc = 745.78K
Ppc = 2093.07 kPa
API Procedure 4D3.1 (1987)
( )
T pc = 10.6443 exp − 5.1747 × 10 − 4 Tb − 0.5444S + 3.5995 × 10 − 4 Tb S Tb0.81067 S 0.53691
API Procedure 4D4.1 (1987)
( )
Ppc = 6.162 × 10 6 exp − 4.725 × 10 −3 Tb − 4.8014 S + 3.1939 × 10 −3 Tb S Tb−0.4844 S 4.0846
where :
Tb = mean average boiling point in oR
S = specific gravity, 60 o F/60 o F
T pc = pseudocritical temperatur e in oR
Ppc = pseudocritical pressure in psia
Slide 38
40. To: Molecular-Based Representation of
Middle Distillate
Tc Pc
(K) (kPa)
butyldecalin 729.78 1906
butylindan 749.25 2507
pentadecane 701.08 1471
41. Linear Combination of Bond Orbitals (LCBO) Method
Applied to Properties from Structure
• Based on work of R.D. Brown (1953): for molecules with σ bonds, using a set of
orbitals situated in each bond (e.g. methane’s 8 σ electrons move in orbitals formed
by linear combinations of localized σ orbitals in each C-H bond
• Total electronic energy by orders of overlap integral:
– At zeroth order, just depends linearly on number of C and H atoms
– At first order, just depends on number of CH3, CH2, CH and C atoms
– At second, just depends on number of each possible Ci-Cj bonds (9 of them) and of above carbon
atoms
– At third order, just depends linearly on number of possible Ci-Cj-Ck triplets (35 of them) and of the
above carbon atoms and bonds
• The higher the degree of expansion, the more distinguishable will be the positions of
isomers
• LCBO method accurately predicts ΔH0f, ΔG0f, enthalpies of alkyl radicals, C-H and
C-C dissociation energies
• It is possible to add CH2, CH and C naphthenic carbon atoms to alkyl atoms and
corresponding bond combinations
• Provides additional accurate prediction of:
– Activation energies required for kinetic model of thermal cracking (initiation, propagation, evolution
and termination reactions)
42. Example: Decylbenzene, Normal Boiling Point
Atoms Bonds
Type Count Δi Type Count Δi
CH3 1 0.10340 CH3-CH2 1 -0.04364
CH2 9 -0.28352 CH2-CH2 8 0.23626
aromCH 5 -0.05881 CH2-aromC 1 0.12346
aromC 1 0.02533 aromCH-aromCH 4 0.00341
aromCH-aromC 2 -0.03111
Multi - linear regression of tranformed equation used to initialize non - linear form :
⎡⎛ N desc
⎞ ⎛ N desc
⎞⎤
Tb = c3 ⎢⎜ c1 + ∑ ni ∆ i ⎟
⎜ ⎟ ⎜ c 2 + ∑ ni ∆ i ⎟ ⎥
⎜ ⎟
⎣⎝ i =1 ⎠ ⎝ i =1 ⎠⎦
N desc
∑n ∆
i =1
i i = 1 * 0.10340 + 9 * −0.28352 + 5 * −0.05881 + 1 * 0.02533
+ 1 * −0.04364 + 8 * 0.23626 + 1 * 0.12346 + 4 * 0.00341 + 2 * −0.03111 = −0.79569
Tb = 1195[(− 0.10206 − 0.79569) (− 1.10206 − 0.79569 )]
Tb = 565.31K
Slide 42
44. Conclusions and Further Work
• The Linear Combination of Bond Orbitals (LCBO) method improves
accuracy of prediction of thermo-physical properties
• Number of “atom” and “bond” descriptors is manageable for hydrocarbons and
molecules with hetero-atoms found in crude oil
• Linear and hyperbolic equation forms are sufficient to represent almost all properties
• Linear regression can be used to ascertain statistical significance of regressed
coefficients and confidence intervals around predictions
• Equations to predict properties are independent of other property predictions
• Molecular representation may be a viable alternative approach to
the characterization of petroleum fluids
• Accuracy in prediction of thermo-physical properties of mixtures can be improved
• More knowledge of chemical structure is retained for reaction pathways and kinetics
calculations
• Work is in progress to:
• Create a data bank of thermo-physical property data for components with boiling
points up to 800K and above
• Implement a Gibbs energy minimization method to infer composition in terms of a
subset of defined surrogate components from basic assay data
45. Wrap Up
• Heavy Oil / Bitumen is in Our Future
• Current Modeling Technology is Limited
• Invensys SimSci-Esscor is on a journey
– Heavy Oil Consortium Providing Input
– Working Through the Issues
– Improving Model Robustness and Accuracy
• Driven to ensure proper design, operations, and optimization of
facilities
• Ultimately translates into to cost savings for our clients through the
design, operate, and optimize life cycle.