Planning For The Integrated Refinery Subsystems PDF

Cranfield University
Edith Ejikeme-Ugwu
PLANNING FOR THE INTEGRATED REFINERY SUBSYSTEMS
SCHOOL OF ENGINEERING
PhD. THESIS
Cranfield University
SCHOOL OF ENGINEERING
PhD. THESIS
Edith Ejikeme-Ugwu
PLANNING FOR THE INTEGRATED REFINERY SUBSYSTEMS
Supervisor: Dr. Meihong Wang

Co-Supervisor: Prof. Hoi Yeung
June 2012
This thesis is submitted in partial fulfilment of the

requirements for the degree of Doctor of Philosophy
© Cranfield University, 2012. All rights reserved. No part of

this publication may be reproduced without the written
permission of the copyright holder.
Abstract
In global energy and industrial market, petroleum refining industry accounts for a
major share. Through proper planning and the use of adequate mathematical
models for the different processing units, many profit improving opportunities can
be realized. The increasing crude oil price has also made refining of crude oil
blends to be a common practice. This thesis aims to provide useful insight for
planning of the integrated refinery subsystems. The main subsystems referred to
are (1) The crude oil unloading subsystem (2) The production and product
blending subsystem and (3) The product distribution subsystem.
Aspen HYSYS® was first used to develop a rigorous model for crude distillation unit
(CDU) and vacuum distillation unit (VDU). The rigorous model was validated with
pilot plant data from literature. The information obtained from the rigorous model
is further used to develop a model for planning of the CDU and VDU. This was
combined with models (obtained from empirical correlations) for fluid catalytic
cracker (FCC) and hydrotreater (HDT) units to form a mathematical programming
planning model used for refinery production and product blending subsystem
planning. Since two different types of crude were considered, the optimum
volumetric mixing ratio, the sulphur content at that mixing ratio and the CDU flow
rate were determined.
The yields fraction obtained from the rigorous model were then used to generate
regression model using least square method. The sulphur composition of the crude
oil was used as independent variable in the regression model. The generated
regression models were then used to replace the regular fixed yield approach in a
refinery planning model and the results compared. From the results obtained, the
proposed method provided an alternative and convenient means for estimating
yields from CDU and VDU than the regular fixed yield approach.
The proposed aggregate model for the production and products blending
subsystem was integrated with the modified scheduling model for the crude
unloading subsystem developed by Lee et al. (1996) and products distribution
i
model developed by Alabi and Castro (2009) for refinery planning. It was found
that the regression model could be integrated in a refinery planning model and
that the CDU flow rate was maximised as compared to the non- integrated system.
Keywords: Refinery planning, linear programming, optimization, linear regression,

process modelling
ii
Acknowledgements
I hereby extend my innermost gratitude to Dr. Meihong Wang my supervisor.

Meihong, this thesis benefited immensely from your guidance, tolerance, your
excellent grasp of what was important and your attention to detail. You were
always available when I needed help and your support and encouragement kept
me going throughout the difficult times. I would also like to extend my
appreciation to my co-supervisor, Professor Hoi Yeung for his high level critics and
ingenuity all through the cause of this work. Hoi, you are not just a supervisor, but
a friend and a confidant. I really appreciate you.
My sincere appreciation goes to Dr Lazaros Papageorgiou and Dr Songsong Liu at

Department of Chemical Engineering, University College London (UCL) for their
candid contribution in making this work a huge success.
I would like to extend my appreciation to Dr ‘Keluo Chukwuogo, the Managing and

Technical Director of Boskel Nigeria Ltd. for all his support and encouragements.
‘Keluo, if not for your prod towards self-development, PhD would not have been
considered. I thank you for the motivation.
My personal gratitude is extended to my colleagues, Chet, Mubarak, Alhaji Aminu,

Aminu Hamisu, Habeeb, Yemi, and most importantly Adekola for your continuous
support and encouragement all through the period of this work.
The VC Bedford family, you will never be forgotten, especially my Mummy in the
Lord Pastor Christina Emmanuel-Odiachi for your sincere and continuous
encouragements and prayers all through the tough times. May the Lord continue to
pour his fresh oil upon you and your family will always remain in the hollow of his
palm.
Enough thank you cannot be said to my siblings and my mum for their endless
prayers. Special thank you to my sister fondly called Nulia and her family for all
effort made to see that this work was a success.
iii
My heartfelt appreciation to my darling and loving husband (Obim) for all the
sacrifices you offered for this journey to come true. You denied yourself of a lot of
things for this work to move on. Your encouragement was second to none; I am
saying thank you to the best Daddy in the world. Also to my beautiful daughters
Chidera Adaeze Nkeiruka Ejikeme and Fechi Kosisochukwu Chisom Ejikeme for all
their support and understanding all through the hard times. Your birthday
celebrations were denied so that mummy can meet up with her work. From my
heart, I say thank you and I love you all so much.
iv
Table of contents
Abstract..................................................................................................................................................... i
Acknowledgements ............................................................................................................................ iii
List of figures ........................................................................................................................................ xi
List of tables ........................................................................................................................................xiii
Abbreviations ...................................................................................................................................... xv
Nomenclature ................................................................................................................................... xvii
1 Introduction ................................................................................................................................. 1
1.1 Introduction to refinery ................................................................................................... 1
1.1.1 Crude oil unloading subsystem ............................................................................ 2
1.1.2 Production and product blending subsystem ................................................ 3
1.1.2.1 Atmospheric tower or Crude Distillation Unit (CDU) .......................................... 4
1.1.2.2 VDU ........................................................................................................................................... 5
1.1.2.3 Catalytic processes ............................................................................................................. 5
1.1.2.4 Reformer unit ....................................................................................................................... 5
1.1.2.5 Hydro treating Process ..................................................................................................... 6
1.1.2.6 Alkylation unit ...................................................................................................................... 6
1.1.3 Product distribution subsystem .......................................................................... 7

1.2 Refinery planning ............................................................................................................... 7
1.2.1 Strategic (long-term) planning ............................................................................ 8
1.2.2 Tactical (medium-term) planning ...................................................................... 8
1.2.3 Operational (short-term) planning .................................................................... 8
1.3 Refinery scheduling ........................................................................................................... 8
1.4 Motivations for the study ................................................................................................ 9
1.4.1 Modelling for refinery planning .........................................................................10
1.4.2 Modelling for production planning...................................................................10
1.4.3 Refinery planning for crude oil mix ..................................................................11
1.4.4 Planning for the integrated refinery subsystems........................................11
v
1.5 Research aim, objectives and Scope ..........................................................................12
1.6 Methodology.......................................................................................................................13
1.7 Novelties in this thesis ...................................................................................................13
1.8 Computational tools used .............................................................................................14
1.8.1 Introduction to GAMS ............................................................................................14
1.8.2 Introduction to Aspen HYSYS®...........................................................................15
1.9 Outline of the thesis .........................................................................................................15
2 Literature Review.....................................................................................................................17
2.1 Introduction .......................................................................................................................17
2.2 Mathematical optimization techniques ...................................................................17
2.2.1 LP ...................................................................................................................................18
2.2.2 NLP ................................................................................................................................20
2.2.3 MILP ..............................................................................................................................21
2.2.4 MINLP...........................................................................................................................23
2.3 Planning in oil refineries ...............................................................................................23
2.3.1 Types of planning models.....................................................................................24
2.3.1.1 Rigorous models ................................................................................................................ 24
2.3.1.2 Empirical models .............................................................................................................. 24
2.3.2 Modelling of process units for refinery planning ........................................24

2.3.2.1 CDU ......................................................................................................................................... 24
2.3.2.2 VDU ......................................................................................................................................... 25
2.3.2.3 FCC .......................................................................................................................................... 26
2.3.2.4 HDT ......................................................................................................................................... 26
2.3.3 Current methods for CDU modelling in refinery planning ......................27

2.3.3.1 Fixed yield approach ....................................................................................................... 27
2.3.3.2 Swing cut method ............................................................................................................. 28
2.3.3.3 Fractionation-index method......................................................................................... 29
2.3.3.4 Production and product blending subsystem ....................................................... 31
2.3.3.5 Integrated modelling approach for the subsystems ........................................... 32
2.3.3.6 Future trend in refinery planning .............................................................................. 32
vi
2.4 Scheduling in oil refineries ...........................................................................................33
2.4.1 Types of Scheduling models ................................................................................33
2.4.1.1 Discrete- time formulation ............................................................................................ 34
2.4.1.2 Continuous-time formulation ...................................................................................... 34
2.4.2 Modelling for refinery scheduling .....................................................................35

2.4.3 Recent advances in refinery scheduling .........................................................36
2.4.3.1 Crude oil unloading subsystem scheduling............................................................ 36
2.4.3.2 Production unit scheduling ........................................................................................... 37
2.4.3.3 Product distribution subsystem scheduling .......................................................... 38
2.5 Integration of planning and scheduling ...................................................................38

2.6 Summary ..............................................................................................................................39
3 Rigorous Modelling for CDU and VDU ..............................................................................41
3.1 Introduction .......................................................................................................................41
3.2 Process description for the refinery production ..................................................42
3.3 Modelling of CDU and VDU in Aspen HYSYS® .......................................................43
3.3.1 Crude oil characterization ....................................................................................43
3.3.2 Blending or mixing of crude oil ..........................................................................45
3.3.3 Pre-Heat train............................................................................................................45
3.3.4 CDU modelling ..........................................................................................................47
3.3.5 VDU modelling ..........................................................................................................50
3.3.6 Validation of CDU model .......................................................................................52
3.4 Process analysis for CDU and VDU under different crude mix ratios ..........56
3.4.1 25/75 volumetric mixing ratio ...........................................................................56
3.4.4 Analysis of the results on the different crude mix proportions ............59
3.5 Summary ..............................................................................................................................59
4 Refinery Production and Product Blending Subsystem Planning Based on
Fixed Yield Approach .......................................................................................................................61
4.1 Introduction .......................................................................................................................61
vii
4.2 Process description .........................................................................................................62
4.3 Mathematical programming model for CDU and VDU .......................................63
4.3.1 CDU material balance constraints .....................................................................63
4.3.2 VDU material balance constraint .......................................................................64
4.3.3 Capacity constraint for CDU and VDU .............................................................64
4.4 Empirical correlations used in mathematical programming model for HDT
and FCC units ..................................................................................................................................65
4.4.1 HDT unit and material balance...........................................................................65
4.4.2 FCC unit and material balance ............................................................................65
4.4.3 Capacity constraint for HDT and FCC units ...................................................67
4.5 CDU and VDU mathematical programming planning model validation .....67
4.6 Model for refinery production and product blending subsystem planning
in GAMS .............................................................................................................................................73
4.6.1 Mathematical model formulation ......................................................................73
4.6.2 Capacity constraint .................................................................................................74
4.6.3 Blending Constraint ................................................................................................74
4.6.4 Market requirement Constraint .........................................................................74
4.6.5 Crude oil composition constraint ......................................................................74
4.6.6 Objective function....................................................................................................74
4.6.7 Crude limit constraint ............................................................................................75
4.7 Refinery planning under different volumetric mixing ratios ..........................76
4.7.1 Solution strategy ......................................................................................................76
4.7.2 Problem statement and implementation .......................................................76
4.7.4 Decision variables ...................................................................................................77
4.7.5 Parameters used ......................................................................................................77
4.8 Results and analysis ........................................................................................................80
4.9 Summary ..............................................................................................................................83
5 Refinery Production and Product Blending Subsystem Planning with
Aggregate Model ................................................................................................................................85
viii
5.1 Introduction .......................................................................................................................85
5.2 Derivation of regression model for CDU and VDU with information from
rigorous model ...............................................................................................................................86
5.2.1 Determination of factor that is used as independent variable...............86
5.2.2 The derived regression models for CDU and VDU ......................................87
5.2.3 Validation with GAMS model for CDU and VDU ...........................................90
5.3 Derivation of regression model for FCC and HDT with information from
empirical correlation ...................................................................................................................91
5.4 Aggregate model development ...................................................................................92
5.5 Refinery production and product blending planning using aggregate
model .................................................................................................................................................92
5.5.1 Problem statement and implementation .......................................................93
5.5.3 Decision variables ...................................................................................................93
5.5.4 Parameters used ......................................................................................................93
5.6 Results and analysis ........................................................................................................94
5.7 Comparison between fixed yield method and the regression method ........95
5.8 Summary ..............................................................................................................................96
6 Planning for the Integrated Refinery Subsystems under Deterministic
Condition ...............................................................................................................................................99
6.1 Introduction .......................................................................................................................99
6.2 Planning model for the crude unloading subsystem ....................................... 100
6.2.1 Mathematical model formulation for crude unloading subsystem ... 101
6.2.2 Material balance equations for the vessel: ................................................. 102
6.2.3 Material balance equations for the storage tanks: .................................. 103
6.2.4 Material balance equations for charging tank:.......................................... 104
6.2.5 Rules that must be followed for charging of crude oil during operation
105
6.3 Planning model for the production and product blending subsystem ..... 107
6.4 Planning model for the products distribution subsystem ............................. 107
6.5 Planning model for the integrated refinery subsystems................................ 109
ix
6.6 Refinery planning for integrated subsystems .................................................... 110
6.6.1 Problem statement and implementation .................................................... 112
6.6.2 Objective function................................................................................................. 113
6.6.3 Decision variables ................................................................................................ 113
6.6.4 Parameters used ................................................................................................... 113
6.7 Results and analysis ..................................................................................................... 114
6.7.1 Variation in Crude oil Price ............................................................................... 117
6.7.2 Variation in Final Products Price .................................................................... 118
Summary................................................................................................................................... 118
6.8 ..................................................................................................................................................... 118
7 Conclusions and Recommendations for Future Work ............................................ 119
7.1 Conclusions ...................................................................................................................... 119
7.1.1 Rigorous modelling of CDU and VDU ............................................................ 119
7.1.2 Fixed yield method for refinery production and product blending
subsystem planning.............................................................................................................. 119
7.1.3 Aggregate model for refinery production and product blending
subsystem planning.............................................................................................................. 120
7.1.4 Planning for the integrated refinery subsystems..................................... 120
7.2 Recommendations for future work ........................................................................ 121
7.2.1 Consideration of more than two pre-mixed crude oils .......................... 121
7.2.2 Validation of CDU model .................................................................................... 121
7.2.3 Consideration of uncertainty ........................................................................... 121
7.2.4 Real life case study for Refinery Planning................................................... 121
References ......................................................................................................................................... 123
Appendices ........................................................................................................................................ 131
Appendix A (Ratawi and Brent Crude Assay) ................................................................. 131
Appendix B (Crude oil cost) ................................................................................................... 133
Appendix C (Charts used for Empirical correlations) ................................................. 135
Appendix D (Graphs of Yields versus API) ....................................................................... 145
Appendix E (Graphs of Yields versus Viscosity) ............................................................ 150
x
Appendix F (Process flow sheet) ......................................................................................... 152
Appendix G Conversion from ASTM to TBP (Li 2000) Polynomial Regression
Method ........................................................................................................................................... 153
Appendix H (Details of Rigorous runs in AspenHYSYS) ............................................. 153
Appendix I Publications ........................................................................................................... 159
xi
List of figures
Figure 1-1 Topology of a Typical Refinery (Courtesy of the Standard oil company) .......... 1
Figure1-2 Three Main Subsystems of a Refinery ............................................................................... 2
Figure1-3 Crude oil Unloading Subsystem View ............................................................................... 3
Figure1-4 A view of the Production and products blending subsystem vi.............................. 4
Figure 1-5 Complete activities of what happens in a typical refinery (Hydrocarbon
processing, 2012). ......................................................................................................................................... 4
Figure 1-6 Product distribution subsystem view (Courtesy of Pacific L.A. marine
terminal LLC, 2012) ...................................................................................................................................... 7
Figure 1-7 Schematic showing information and material flow in integrated system ....... 12
Figure 2-1 Fixed Yield Model (Adapted from Alattas et al. 2011) ............................................ 28
Figure 2-2 flow diagram of a Fixed yield structure representation ......................................... 28
Figure 2-3 Swing cuts model (Reproduced from Li, 2004) ......................................................... 29
Figure 2-4 The FI model (Reproduced from Alattas et al., 2011) ............................................. 30
Figure 3-1 Overview of Chapter Methodology ................................................................................. 41
Figure 3-2 Configuration of a Refinery (Arofonosky et al., 1978) ............................................ 42
Figure 3-3 TBP and ASTM D86 curve for the crude mix ............................................................... 48
Figure 3-4 Generic CDU and VDU model in Aspen HYSYS ........................................................... 52
Figure 3-5 Comparison of product yields from CDU model in Aspen HYSYS and Li et al.
(2005) .............................................................................................................................................................. 55
Figure 3-6 Conversion of ASTM to TBP (using Li, 2000) ............................................................. 56
Figure 4-1 Refinery Topology for the Process .................................................................................. 62
Figure 4-2 CDU yield from Aspen HYSYS model validation with GAMS ................................. 70
Figure 4-3 CDU yield from Aspen HYSYS model validation with GAMS ................................. 71
Figure 4-4 Plot of Volumetric ratios versus Profits ........................................................................ 81
Figure 5-1 Procedures for Developing Aggregate Model ............................................................. 85
Figure 5-2 Plots of predicted yields from CDU and VDU .............................................................. 88
Figure 6-1 Flow of crude oil for Crude unloading subsystem ................................................. 100
Figure 6-2 Schematics for the Integrated Refinery Subsystems............................................. 109
Figure 6-3 Flow rate of Final Products ............................................................................................. 115
xi
Figure 6-4 Refinery Profit and Variation in crude Price ............................................................ 117
Figure 6-5 Variation in Final product Price .................................................................................... 118
xii
List of tables
Table 1-1 Difference between planning and scheduling ................................................................ 9

Table 3-1 Crude oil types and properties ........................................................................................... 44
Table 3-2 Sulphur concentration, API and Viscosity in the crude oil mixture ..................... 44
Table 3-3 Pseudo Components and Physical Properties of crude oil mix ............................. 46
Table 3-4 TBP and volume percent distilled ..................................................................................... 49
Table 3-5 The mixed crude Fractions TBP (oF) cut point Specification using Aspen
HYSYS prediction ......................................................................................................................................... 49
Table 3-6 The mixed crude cut point Specification at ASTM D86using Aspen HYSYS
prediction ........................................................................................................................................................ 50
Table 3-7 Vacuum Tower Fractions TBP (oF) cut point Specification predicted by Aspen
HYSYS................................................................................................................................................................ 51
Table 3-8 Cut point Specification at ASTM D86 for Vacuum Tower predicted by
AspenHYSYS ................................................................................................................................................... 51
Table 3-9 Bulk properties of Tia Juana crude oil (Venezuela), (Watkins, 1979) ................ 53
Table 3-10 TBP ranges of CDU fractions (Li et al., 2005) ............................................................. 54
Table 3-11 Comparison of product yields from CDU model in Aspen HYSYS and Li et al.
(2005) .............................................................................................................................................................. 54
Table 3-12 Conversion of ASTM D86 to TBP ..................................................................................... 56
Table 3-13 Product yields from CDU and VDU in HYSYS for 25/75 mix ................................ 57
Table 3-14 Product yield from CDU and VDU from HYSYS for 75/25 crude mix ............... 58
Table 3-15 Products yield from CDU and VDU in HYSYS for 50/50 mix ................................ 58
Table 4-1 Products cuts from Hydro treating of the oils .............................................................. 65
Table 4-2 Yields fraction from FCC ....................................................................................................... 66
Table 4-3 The Coefficients generated for CDU and VDU in Aspen HYSYS at 50/50 crude
mix ..................................................................................................................................................................... 68
Table 4-4 Validation results for CDU and VDU ................................................................................. 69
Table 4-5 Validation of CDU model with yield from GAMS.......................................................... 69
Table 4-6 Validation of VDU model with yield from GAMS ......................................................... 70
Table 4-7 Product yield of 25/75% crude oil mix from HYSYS ................................................. 71
xiii
Table 4-8 Validation of yield results for 25/75 volumetric mixing ratio from Aspen
HYSYS................................................................................................................................................................ 72
Table 4-9 Capacity of processing units in the Refinery ................................................................. 78
Table 4-10 Product price for Final Products (EIA 2011) ............................................................. 78
Table 4-11 Process Units Operating Cost (Ajose, 2010)............................................................... 78
Table 4-12 Production limit (Ajose, 2010) ........................................................................................ 79
Table 4-13 Crude oil and Butane data (EIA 2011).......................................................................... 79
Table 4-14 Summary of the profit generated from different volumetric mixing ratios ... 80
Table 4-15 Summary of results obtained from the refinery planning model using Fixed
yield method .................................................................................................................................................. 82
Table 5-1 Different volumetric mixing ratios and their properties ......................................... 86
Table 5-2 Regression model derived for CDU and VDU yields ................................................... 89
Table 5-3 Regression model derived for CDU and VDU units yield.......................................... 89
Table 5-4 GAMS results for comparison with Aspen HYSYS model ......................................... 90
Table 5-5 Comparison of Yields obtained from GAMS Model and results from
AspenHYSYS in 50/50 crude mix. .......................................................................................................... 90
Table 5-6 Comparison of Yields obtained from GAMS Model and results from
AspenHYSYS in 75/525crude mix. ........................................................................................................ 91
Table 5-7 Regression model for FCC .................................................................................................... 92
Table 5-8 Summary of results obtained from the refinery planning model on Aggregate
Model ................................................................................................................................................................ 94
Table 5-9 Comparing Fixed yield and Aggregate method ............................................................ 95
Table 6-1 Data used for the 30 day horizon (Lee et al., 1996) ................................................ 106
Table 6-2 Final products quality specification (AMD Refinery data) ................................... 114
Table 6-3 Summary of results for the integrated refinery subsystem ................................. 114
xiv
Abbreviations
Atmospheric Gas Oil
Crude Distillation Unit
Vacuum Distillation Unit
Fluid Catalytic Cracking
Hydrotreater
Vessels
Charging Tank
American Petroleum Institute
True Boiling Point
American Society for Testing and Materials
Barrels
Mixed integer linear programming
Mixed integer non-linear programming
Non-linear programming
Linear programming
End Boiling Point
Initial Boiling Point
Energy Information Administration
Fuel Gas
Fuel-Oil
xv
Gross Overhead
Heavy Naphtha
Heavy Distillate
Light Naphtha
Light Distillate
Bottom Residue
General Algebraic Modelling Systems
Gasoline Blend
Gasoline product
High Vacuum Gas Oil
Low Vacuum Gas Oil
Liquefied Petroleum Gas
Normal Boiling Point
Process Industry Modelling Systems
Refinery and Petrochemical Modelling Systems
xvi
Nomenclature
Set/indices
VE = {v =1, 2…V/ crude oil vessel}
ST = {i =1, 2…l/ storage tanks}
CT = {j =1, 2…J/charging tanks}
COMP = {k =1,2…K/ crude oil components}
CDU = {l = 1,2…L/crude distillation units}
Time = {t =1,2…T/ Time Horizon}
Time = {m =1,2…m/ Time Horizon}
: Volume of crude oil in storage tank i at time t
: Storage tank i maximum capacity.
Volume of crude oil in charging tank j at time t
: Storage tank minimum capacity.
: Charging tank maximum capacity.
: Charging tank minimum capacity
: Unloading cost for vessel v
: Charging tank changeover cost
Inventory cost for storage tanks i per unit time per unit volume
Inventory cost for charging tanks j per unit time per unit volume
: Minimum concentration of component k in the blended crude oil of

charging tank j.
: Maximum concentration of component k in the blended crude oil
of charging tank j.
: Minimum concentration of component k in storage tank i.
: Maximum concentration of component k in storage tank i.
xvii
: Demand of blended or mixed crude oil j by CDU within the
scheduling horizon
Volumetric flow rate from vessel v to storage tank i at time t.
Maximum crude oil rate from vessel v to one storage tank i.
: Minimum crude oil rate from vessel v to one storage tank i. In this
situation its assumed 0
Volumetric crude oil rate from storage tank i to one charging tank
j.
Maximum crude oil rate from storage tank i to one charging tank j.
Minimum crude oil rate from storage tank i to one charging tank j
Volumetric flow rate from charging tank j to CDU l at time t
Maximum crude oil rate from charging tank j to CDU l
Minimum crude oil rate from charging tank j to CDU l
Volume of crude oil in vessel v at time t
Binary variable to denote if storage tank i is receiving crude from

vessel v.
Binary variable to denote if charging tank j is receiving crude from
storage tank i
Binary variable to denote if CDU l is receiving mixed or blended
crude j,g from charging tank j
Volumetric flow rate of component k from vessel v to storage tank
i at time t.
Volumetric flow rate of component k from storage tank i to one
charging tank j.
Volumetric flow rate of component k from charging tank j to CDU l
at time t
Volume of component k in charging tank j at time t
xviii
Time for vessel v to begin unloading
Time for vessel v to finish unloading
Vessel v arrival time
Vessel v departure time
Binary variable to denote storage tank I holding at time t
Penalty for holding tank i
Total optimal operational cost
Parameter for demand violation for mixed crude j
Parameter defined by the user that determines the interval to

interval variation for CDU l throughput.
Binary variable to denote that CDU l switches from charging tank j
to charging tank g.
Binary variable that denotes that vessel v starts unloading at time t
Binary variable that denotes that vessel v finishes unloading at

time t
Production and Product Blending Subsystem
Depot
Mixed crude into the CDU
Time interval for the planning horizon
All commodities
Intermediate materials from process units
Purchased material which include butane
Final products
Blending possibilities
xix
Attributes for quality
Process Units
Upper and lower limits
Coefficient of yield from unit u based on feed stream flow rate f

and output stream f’ that depends on the type of crude mix cm
Coefficient of intermediate product flow streams intm, from unit u
which depends on the type of crude mix cm
Attribute for blending the intermediate products
Maximum final product specification
Minimum final products specifications
Level of intermediate material intm from crude mix cm that

blended into final product fp during time period t
Cost of refinery processes
Cost of crude mix cm during time period t
Capacity of process units
Cost of purchased material pm during time period t
Cost of refining or refinery operations at time period t
Total sales revenue from final products at time period t
Sales revenue from final products
Raw material cost at time period t
Flow of final product fp
Level of process u from crude mix cm at time period t
Purchased crude mix entering the unit (CDU)
Flow of intermediates from process units
xx
Flow rate of purchased material
Maximum production requirement

Products distribution
Total Transportation cost for final products to depot at time t
Total Product tanks inventory cost
Volume of final product fp from product tank pt to depot dp at time

t
Inventory of final product tank pt
Volume of final product fp from product tank to depot dp at time t
Maximum volume of product tank pt
Minimum volume of product tank pt
Flow of final product fp from product tank to depot dp at time t
Demand of final product fp from product tank pt at the depot dp at

time t
Flow of final product fp to product tank pt at time t
Initial volume of product fp in the product tank pt at time t
Volume of final product fp at the depot dp at time t
Initial volume of final product fp at the depot dp at time t
Transport cost for final products to depot at time t
xxi
1 Introduction
1.1 Introduction to refinery
The quest to develop the petroleum refining industry came from several changes in life-
styles. The increased needs for illuminants, for fuel to drive the factories of the
industrial revolution, for gasoline to power the automobiles, as well as the demand for
aviation fuel, all contributed to the increased use of petroleum. The world prediction is
that oil consumption will increase from 77.1 million barrels/day in 2001 to 118.8
million barrels/day in 2025 (Li, 2004).
A refinery is essentially a group of chemical engineering unit processes and unit

operations for the purpose of converting crude oil into products of value while
considering the qualities and quantities stipulated by the market (Tung-Hsiung and
Chuei-Tin, 2008).
Figure 1-1 is a typical refinery topology, which is made up of the storage tank farm (e.g.
storage tanks for crude oils, storage tanks for intermediate products, charging tanks)
and the processing plants which is the most complex part of the refinery.
Figure 1-1 Topology of a Typical Refinery (Courtesy of the Standard oil company)
1
The refinery can roughly be divided into three main subsystems as shown in Figure 1-2,
where Ci, Pi and DPi are crude oil types, final products and depot or gas stations:
 Crude oil unloading subsystem

 Production and product blending subsystem
 Product distribution subsystem
Crude oils arrives in either a tanker, vessels or a truck and are stored in storage tanks
for charging or blending before being refined and blended to meet the target set by the
planners. These blended final products are then sent to depots or gas stations for
retailing.
Figure1-2 Three Main Subsystems of a Refinery
1.1.1 Crude oil unloading subsystem
The crude oil unloading subsystem consists of unloading berths or docking stations,
storage tanks, and charging tanks with pipelines connecting them. In some situations
this subsystem is located near the sea for ease of loading and or unloading into the
tanks for storage as shown in Figure 1-3. As tankers or vessels arrive at the refinery’s
docking stations/berths for unloading, the refinery ensures that there is sufficient
capacity for storage.
2
Figure1-3 Crude oil Unloading Subsystem View
The Coryton refinery located in Essex on the Thames Estuary, southeast UK
Image courtesy of Terry Joyce (Courtesy ofHydrocarbonstechnology.com)
In the crude oil unloading subsystem of a refinery, various types of crude oil are
purchased and processed. These crude oils are distinguished by their various
compositions (Wu et al., 2006). The processing flow scheme of a refinery is largely
determined by the composition of the crude oil and the chosen cuts of the petroleum
products (Trzupek, 2002).
The compositions of processed crude have great influence on refinery margins. This
gave rise to control of the quality of charged crude by refiners and the use of advance
technologies to plan and schedule crude oil changes (Reddy et al., 2004).
1.1.2 Production and product blending subsystem
The production and product blending subsystem of the refinery consists of the Crude
Distillation Unit (CDU), Vacuum Distillation Unit (VDU), Fluid Catalytic Cracker (FCC)
and the Hydrotreater (HDT) units and other process units used to convert the crude to
more valuable products. A view of the production and products blending subsystem is
shown in Figure 1-4.
3
Figure1-4 A view of the Production and products blending subsystem vi
(Courtesy of Building green dot com)
Figure 1-5 shows the general processing arrangement of the production and product
blending subsystem of a refinery. The design of a refinery is tailored to a particular
crude oil specifications or properties; however, there are some basic process operations
that are carried out in most refineries. Some of these processes are.
Figure 1-5 Complete activities of what happens in a typical refinery (Hydrocarbon processing, 2012).
1.1.2.1 Atmospheric tower or Crude Distillation Unit (CDU)
The crude is heated in a furnace and charged to an atmospheric distillation tower,

where it is separated into butanes and lighter wet gas, unstabilized full-range gasoline,
heavy naphtha, kerosene, heavy gas oil, and topped crude. The topped crude or
4
atmospheric residue from the bottom of the CDU is sent to the vacuum tower and
separated into a vacuum gas oil, overhead stream and reduced crude bottoms (Gary and
Handwerk, 1975).
1.1.2.2 VDU
The feed into the VDU is the atmospheric residue from the CDU. The stream into the
VDU is treated in the same way as the CDU but under high pressure. This VDU helps to
further remove the distillate from the crude oil before being sent to the catalytic cracker
unit. The reduced crude bottoms from the vacuum tower called the vacuum residue, is
thermally cracked in a delayed coker to produce wet gas, coker gasoline, gas oil, and
coke.
1.1.2.3 Catalytic processes
These are the conversion operations that further process the heavy and less valuable
products from the distillation unit into high octane products. There are two types of
such processes in refinery which are: Fluid Catalytic Cracking Unit (FCCU) and Catalytic
Reforming unit (CRU) (Al-Qahtani and Elkamel, 2010). The atmospheric and vacuum
crude unit gas oils and coker gas oils are used as feedstocks for the catalytic cracking or
hydrocracking units.
1.1.2.4 Reformer unit
The gasoline streams from the crude tower, coker, and cracking units are fed to the
catalytic reformer to improve their octane numbers. The products from the catalytic
reformer are blended into regular and premium gasoline for sale.
The wet gas streams from the crude unit, coker, and cracking units are fractionated in
the vapour recovery section into fuel gas, liquefied petroleum gas (LPG), unsaturated
hydrocarbons (propylene, butylene, and pentenes), normal butane and iso-butane. The
fuel gas is burned in refinery furnaces and the normal butane is blended into gasoline or
LPG. The unsaturated hydrocarbons and iso-butane are sent to the alkylation unit for
processing (Gary and Handwerk, 1975).
5
1.1.2.5 Hydro treating Process
Contaminants in liquids petroleum products are removed by means of hydro treating

operations. The major concern is the presence of sulphur, nitrogen, oxygen and heavy
metals in products distillation such as kerosene, naphtha, and diesel on which if not
removed, will cause deactivation of catalyst and the yield of products which will fall
short of environmental standards. The bottoms from CDU are also hydro treated before
being fed in to the FCC unit in hydrodesulphurization process (HDS). Operating
principle employed is hydrogenation reaction which takes place by reacting hydrogen
with the feed at high temperature to form H2S, NH3, and H2O. Generally, the nature of
the feed and the level of treatment required determine the operating condition of hydro
treating process. Treated liquid naphtha is separated and stabilized in a column to strip
off H2S and light gases before being fed to CRU (Al-Qahtani and Elkamel, 2010).
1.1.2.6 Alkylation unit
The alkylation unit uses either sulphuric or hydrofluoric acid as catalyst to react olefins
with iso-butane to form iso-paraffins boiling in the gasoline range. The product is called
alkylate, and is a high-octane product blended into premium motor gasoline and
aviation gasoline.
The middle distillates from the crude unit, coker, and cracking units are blended into
diesel and jet fuels and furnace oils.
In some refineries, the heavy vacuum gas oil and reduced crude from paraffinic or
naphthenic base crude oils are processed into lubricating oils. After removing the
asphaltenes in a propane deasphaltene unit, the reduced crude bottoms are processed
in a block operation with the heavy vacuum gas oils to produce lube-oil base stocks.
The heavy vacuum gas oils and de-asphalted stocks are first solvent-extracted to
remove the heavy aromatic compounds and are de-waxed to improve the pour point.
They are then treated with special clays to improve their colour and stability before
being blended into lubricating oils.
Each refinery has its own unique processing scheme which is determined by the
equipment available, operating costs, and product demand. The optimum flow pattern
of any refinery is dictated by economic considerations and no two refineries are
6
identical in their operation (Gary and Handwerk, 1975). Few if any refinery carry out all
of these processes.
1.1.3 Product distribution subsystem
After the final products are produced, they are stored in the product tanks awaiting to
be distributed to the various regions or depots that are located close to the consumers
or retailers. The primary means of this transportation are ships, pipeline and rails while
the secondary means are trucks as shown in Figure 1-6. In this thesis, trucks are
assumed for transportation of the finished products to depots.
Figure 1-6 Product distribution subsystem view (Courtesy of Pacific L.A. marine terminal LLC, 2012)
1.2 Refinery planning
Planning is concerned with making available an adequate manufacturing capacity for

broad classes of products (Edgar et al., 2001). Planning activities involves optimization
of raw material supply, refining and subsequent commercialization of final products
over one or several time periods (Pinto et al., 2000).A product is any item that the
selection and quantity to be manufactured or produced is decided by top management
(Alonso-Ayuso et al., 2005). In contrast, Scheduling is centred on details of material
flow, manufacturing and production. Planning is forecast driven while scheduling is
order driven.
Time horizon is the set of time periods within which planning is targeted (Alonso-Ayuso
et al., 2005).The characteristics of planning is the time horizon of months or weeks,
whereas scheduling tends to be of shorter duration, that is, weeks, days or hours
7
depending on the cycle time from raw materials to final product (Edgar et al., 2001).
Planning can be classified as follows (Kong, 2002):
 Strategic (long term) planning
 Tactical (medium term) planning
 Operational (short term) planning
1.2.1 Strategic (long-term) planning
This has to do with the decision on production network, plant sizing, product selection
and product allocation among plants. It is also linked closely to a company’s strategic
business plan and direction. The objective is maximization of the expected benefit given
by the net profit along the time horizon (Kong, 2002).
1.2.2 Tactical (medium-term) planning
This is concerned with deciding on the best utilization of limited or available resources
including vendor, factories, depots, and dealer centres along the time horizon such that
given targets are met at a minimum cost or profit maximized (Kong, 2002).
1.2.3 Operational (short-term) planning
This is concerned with determining operations assignment to process plant and also the
sequencing and scheduling of jobs and operations along a time horizon, given a
production network, a specific job demand and operations target (Alonso-Ayuso et al.,
2005). This is subject to the constraints in current inventory of crudes, blending
feedstock and a set of forecasted crude deliveries. The updates are at regular intervals
e.g. monthly, weekly or in response to changes in level of inventory, receipt of any
changes in raw material delivery schedule or sudden changes in the style of demand.
Short-term or operational planning is called scheduling at the production stage (Kong,
2002).
1.3 Refinery scheduling
Oil refineries are increasingly concerned with improving the scheduling of their
operation (Pinto et al., 2000). In the refinery supply chain, crude oil scheduling is vital.
8
Efficient crude scheduling manages plant shutdown, excessive flaring, demurrages and
stock-out (Hartmann, 1997). It also determines the type of blending recipe to use that
satisfies planners’ quantity and quality targets.
In crude oil scheduling operations, the decision makers are faced with factors
concerning the vessels arrival time, the unloading of the vessels, the time for brine and
water to settle and the charging of the CDU with the crude oil. Handling this manually
could be complicated and inefficient as these conditions changes all the time. It will not
also guarantee that the operation is carried out at minimal cost and hence a scheduling
model is necessitated.
Planning and scheduling can be summarised as follows in Table 1-1: (Edgar et al.,
2001).
Table 1-1 Difference between planning and scheduling
planning scheduling
 tactical  operational
 aggregate data  detailed data
 describe large segments of  describe smaller

production environment segments
 planning horizon: months  planning horizon of

or weeks few hours to several
days/weeks
1.4 Motivations for the study
Due to the increasing demand for petroleum products, the cost of crude oil and the set
environmental regulations need to be observed and monitored strictly, tools required
for production planning and management decision making become inevitable for the
refinery economic evaluation (Guyonnet et al., 2009).
9
In modern refineries, the planning department is relied upon for decision making such
as final product demand forecasting, crude oil procurement, production and product
blending and product distribution.
1.4.1 Modelling for refinery planning
Experience has shown that it is more practical and profitable to use mathematical
models for the purpose of planning (Wagner, 1975). This has been encouraged due to
initial successes recorded, and has also pioneered linear optimization methods to a
wide variety of decision areas, which led to the demonstration of techniques for making
this scientific approach realizable in a competitive business environment (Wagner,
1975).
There are many mathematical programming models that exist on refinery planning. For
proper model development and selection of suitable mathematical modelling
techniques good understanding of the process interactions, experience and practice is
required (Al-Qahtani and El-kamel, 2010).
Refinery planning models are mostly linear. However, non-linearity can arise which is
mainly associated with product yields from the refinery units, blending and the pooling
of intermediate product.
1.4.2 Modelling for production planning
In the production and products blending subsystem of the refinery, the refining area is
the place where most of the production activities are performed, activities such as
converting crude oil into more valuable products, to meet customers demand and
quality specifications. The profitability of a refinery will be maximized if these
operations are run as efficiently as possible. There has been so much argument on the
accuracy of using linear model for CDU, VDU, FCC unit and HDT unit in refinery
planning. Refining processes are nonlinear (Slaback, 2004), but the tabulated values of
yield and quality of intermediates products produced are linearized and used in an
optimization planning model (Brooks and Walsem, 1999).The results obtained from
using rigorous models for processing units are more accurate (Li et al. ,2005).
10
1.4.3 Refinery planning for crude oil mix
Refining of crude oil blends is increasing due to the increase in crude oil price, meeting
stringent regulations, making profit and the need to meet the demand of petroleum
products that has been programmed for the refinery. This has resulted in most refiners
buying different crude oil (sweet or sour, heavy or light) and blending them to meet the
processing unit specifications. Accurate modelling to obtain information on yields from
these units for refinery planning is essential.
In the past, several methods have been developed for modelling CDU and VDU to obtain
information on product yields and then used for refinery production planning.
The major challenge faced by refiners is to enhance the distillate yields by optimizing
the mixed crude oil to make them profitable (Shaoping et al., 2010). The cost of crude
accounts for about 80% of the refinery expenditures, which means that when cheaper
crude is processed, it has a positive impact on the refinery profit margins (Stratiev et al.,
2010).
1.4.4 Planning for the integrated refinery subsystems
Planning and scheduling models for refineries should look at modelling each of this
subsystems and realizing the material flow by the integration of all subsystems (Li,
2004).
An integrated modelling approach would provide a better link between the main
subsystems of the refinery, and management of inventory will be achieved while
resolving the issues between the crude oil supplies.
Recent advances to enhance refinery planning model have centred on proper

integration of models i.e. integrating the three main subsystems with the refinery
planning and the idea of integrating multiple refineries for proper production
management (Guyonnet et al., 2009).
Some researchers gave an extensive review on the integration of models on production

and product planning and product distribution model. The reason is that, this
integration could be achieved by feeding the crude oil unloading subsystem by
information from production and product planning or vice versa as in Figure 1-7and
11
their conclusion is that a heuristic based Lagrangian decomposition could be used for
the integration. In spite of the arguments, no integrated model exists for the planning of
the three subsystems (Guyonnet et al., 2009).
Figure 1-7 Schematic showing information and material flow in integrated system
1.5 Research aim, objectives and Scope
This thesis is aimed to deal with tactical (medium term) planning for the integrated
refinery subsystems using Mixed Integer Linear Programming (MILP). The following
objectives are meant to be achieved:
 To develop a rigorous model for CDU and VDU. Validate the model and analyse.
 To develop an aggregate planning model for the processing units for refinery
planning for the production and product blending subsystems.
 To carry out a case study with the aggregate model and implement using
mathematical programming.
 To modify a short term MILP scheduling model by Lee et al. (1996).
 To integrate the developed aggregate mathematical planning model with crude
oil unloading subsystem model, and product distribution subsystem model and
implement in under deterministic condition in refinery planning implement in
GAMS.
 To carrying out Case studies from a refinery in West Africa to demonstrate the
applicability of the proposed model and solution approach.
12
The scope of this research: In this work, two different types of crude were considered
namely: Brent and Ratawi crude. The crude property considered is the sulphur content
in crude oil mix. This is because the sulphur content of crude affects the price of the
crude and the profit margin of the refinery. Only LP and MILP were investigated for
refinery planning. NLP or MINLP were not studied in this thesis since the relationship
between the product yields and the sulphur content in the crude mix was found to be
linear.
1.6 Methodology
Figure 1-8 is used to illustrate the methodology used to achieve the set objectives in this
work.
Figure1-8 Overview of research methodology
1.7 Novelties in this thesis
 Use of mixed crude rather than the single crude generally used for refinery
planning models.
13
 Development of a rigorous model for CDU and VDU and information transfer
from the rigorous model for CDU and VDU into simplified CDU and VDU model
for planning.
 Development of an aggregate model for planning of the production and product
blending subsystem of a refinery.
 Integration of the aggregate model for production and product blending
subsystem with modified scheduling model for crude oil unloading subsystem
and products distribution subsystem for refinery planning.
1.8 Computational tools used
1.8.1 Introduction to GAMS
Substantial progress was made in the 1950s and 1960s with the development of
algorithms and computer codes to solve large mathematical programming problems.
The number of applications with these tools in the 1970s was less than expected,
however, because the solution procedures formed only a small part of the overall
modelling effort. A large part of the time required to develop a model involved data
preparation and transformation and report preparation. Each model required many
hours of analyst and programming time to organize the data and write the programs
that would transform the data into the form required by the mathematical
programming optimizers. Furthermore, it was difficult to detect and eliminate errors
because the programs that performed the data operations were only accessible to the
specialist who wrote them and not to the analysts in charge of the project (Brooks et al.,
1997).
GAMS was developed to improve on this situation by:
 Providing a high-level language for the compact representation of large and

complex models
 Allowing changes to be made in model specifications simply and safely
 Allowing unambiguous statements of algebraic relationships
 Permitting model descriptions that are independent of solution algorithms.
GAMS which represent general algebraic modelling system is specifically designed for
modelling linear, nonlinear and mixed integer optimization problems. The system is
14
especially useful with large, complex problems. GAMS is available for use on personal
computers, workstations, mainframes and supercomputers. GAMS allows the user to
concentrate on the modelling the application problem by making the setup simple. The
system takes care of the time-consuming details of the specific machine and system
software implementation (Brooks et al., 1997).
1.8.2 Introduction to Aspen HYSYS®
Aspen HYSYS® is used to create rigorous steady state and dynamic models for plant
design, performance monitoring, troubleshooting, and operational improvement and
asset management in process industries.
The reasons for choosing Aspen HYSYS are as follows:
 Aspen HYSYS provides the accuracy, speed and efficiency required for process
design activities.
 The level of detail and the integrated utilities available in Aspen HYSYS allows for
skilful evaluation of design alternatives, with fast and efficient process
simulation.
 It is an interactive, project-oriented programme.
 It has the facility for interactively interpreting commands, as they are entered
one at a time (Seider et al., 1998).
1.9 Outline of the thesis
Chapter 2 will review current practice on planning of the individual subsystem,

integrated subsystems under deterministic condition.
Chapter 3 will focus on developing a rigorous model for CDU and VDU. The model will
be validated with pilot plant data from literature and further analysed.
The information from the rigorous modelling is used to develop a simplified

mathematical programming planning model and implement in a refinery production
and product blending subsystem planning model in Chapter 4. Empirical correlation by
Gary and Handwerk (2001) is used for the FCC and HDT units with consideration of the
crude characterization, products yields and qualities. 53 scenarios or volumetric mixing
15
ratios from Aspen HYSYS model is carried out to determine the volumetric ratios that
gives the highest profit.
In chapter 5, an aggregate model for predicting the yield from CDU and VDU is
developed, validated and applied in refinery production and products blending
planning. The developed model is applied to a case which is used for comparison with
the regular fixed yield counterpart.
Chapter 6 will detail on the integration of the modified crude unloading subsystem
model, the developed aggregate model in chapter 4 and already existing products
distribution planning model. The model for the three subsystems shall be applied to a
case study for the planning of a refinery. This refinery planning model is based on
deterministic approach.
Chapter 7 will then be conclusions and recommendations for future works.
16
2 Literature Review
2.1 Introduction
In this chapter, the techniques used for mathematical optimization and the benefits of
using LP are described. Planning in oil refineries and recent practice in modelling
process units for planning in oil refinery is reviewed. The review will also cover
scheduling and its models for the refinery, and recent advances in refinery scheduling.
Finally, the integration of planning and scheduling model is reviewed. The future trend
for refinery planning is included.
2.2 Mathematical optimization techniques
Optimization makes use of efficient quantitative method to select the best among the
entire set of solutions. The aim of optimization is to determine the values of the
variables in the process that yield the best value of the performance criterion (Kallrath
and Wilson, 1997).
Usually a mathematical programming model in optimization consists of four key objects
 Objective function
 Decision Variables (continuous, semi-continuous, binary, integer)
 Constraints( equalities, inequalities), and
 Parameters.
Mostly, objective function focuses on either maximising the profit that is generated for a
particular plan, or minimizing operational cost for scheduling, blending and
transportation subject to constraints that describes the operating conditions (Kallrath
and Wilson, 1997). Identifying the possible decisions or problem variables to be made
is the first step taken in any decision problems. Next is to identify the decisions that are
admissible which leads you to the set of constraints that needs to be determined
according to the nature of the problem in question. The information or values available
are called the parameters (Enrique et al., 2002).
In the past, manual calculations were used in solving optimization problems for
industries and other establishment with inaccuracies existing in the solutions and time
17
consuming factors. This progressively changed to the use of spreadsheet which has
proven to be more accurate and less rigorous compared to the manual calculation. But
the use of mathematical programming and different applications has long taken over
the petroleum industry. The invention of both the simplex algorithm and digital
computers was what propelled the spread of LP applications in the industry followed by
many other applications in the area of refinery planning (Al-Qatahni and Elkamel,
2010).
When developing refinery mathematical optimization models, they usually lead to

structured problems such as
 LP problems
 Nonlinear programming (NLP) problems
 Mixed Integer Linear Programming (MILP) problems
 Mixed Integer Non-Linear Programming (MINLP) problems.
2.2.1 LP
LP involves problems in which both the objective function and the constraints are
linear. The word programming in this context means optimization. It is the most widely
used optimization techniques. For example,
(2-1)
(2-2)
(2-3)
(2-4)
(2-5)
18
Equation (2-1) represents the objective function, Equation (2-2), (2-3), (2-4) and (2-5)
are the constraints while x, y and z are the variables.
The features of LP problems for these examples are:
 There is an objective function which is to be maximized. There is no

multiplication of variables such as x*y or introduction of mathematical functions
such as power, exponential, trigonometric functions of variables generally. It is
the objective function which drives the process of optimization.
 The constraints are also linear in expressions on the left hand side. They are
linked to a constant at the right hand side by a relation which will be one either
of . No other relations are permitted.
 There is also a requirement that the variables cannot take negative values.
 In the linear expressions the constants are usually rational real values, whole
numbers and decimal fractions are included but not expressions such as pi or
roots, which can only be included by approximating them to a given number of
digits.
 Inequalities < or > are not permitted in the constraints, but this turns out not to
be a drawback. Usually practical models will only require constraints
(Kallrath and Wilson, 1997).
Limitations of LP programming techniques
 LP can only handle a single objective which may be unsatisfactory

 In the constraints and objective function of an LP problem, the assumption has to
be made that all modeling can be made linear.
Advantages of LP programming techniques
Problems associated with process that are inherently nonlinear are simplified when
modeled with LP i.e. products properties and feed variance as well as changes in
19
process conditions of operation and their relationship. LP provides a reliable value
structure especially the issue of marginal prices for products.
In solving optimization problems in refinery planning, LP approach is mostly preferred

to other optimization algorithm due to its convenience (Zhang and Zhu, 2006).
The ease of formulation of LP problems and its use for approximating nonlinear model
around its steady state made it a preferred tool. Since LP model is one of the oldest
techniques for optimization, it has been widely applied in the area of planning for
manufacturing and processing industries for obtaining global optimum and reliable
solutions. Hence there is vast information and experiences on LP which is readily
available; this made LP an easy alternative for planners to make use of. Also, the
simplified, robustness and fast nature of the solution time of LP in comparison with
other techniques makes it a quick decision making tool when profit is the ultimate
focus.
Operationally, LP establishes operating strategies and goals of real operations. In as

much as the choice for LP is a tradeoff between simplicity and robustness with accuracy
of the solution obtained especially when nonlinear processes are approximated, the
solutions obtained are optimum with accurate data. LP provides the means of
interpreting the optimum results in simple and concise manner for better
understanding in many industries for better grasp of business or industrial problem
(Simon and Azma, 1983).
Finally, most of the existing commercial software for refinery production planning such
as RPMS (Refinery and Petrochemical Modeling System) is based on a very simple LP
model which mainly is composed of linear relationships (Pinto et al., 2000).
2.2.2 NLP
This is a programming problem with some nonlinear constraints, or a nonlinear

objective function, or both. It can be represented with the following equation:
(2-6)
20
(2-7)
(2-8)
(Edgar et al., 2001)
Equation (2-6) is the objective function, while Equation (2-7) and (2-8) are the
constraints.
Due to the fact that refinery operations exhibit some inherent non-linear characteristics
when modelled, they are often approximated into linear models for simplicity and ease
of computation. NLP model was developed by Moro et al. (1998) for planning of the
entire refinery with all the process units represented and non-linear blending relations.
Pinto and Moro (2000) presented a nonlinear model for each unit and the combination
of the whole refinery. Neiro and Pinto (2004) transformed the same model framework
into a multi-period and multi-scenario model in their attempt to generate models that
truly covers and represents all the features of a refinery.
Companies are developing in-house commercial software packages for NLP models but
are faced with challenges of convexities and difficulty of convergence which is the
reason why NLP are often referred to as ‘NP hard’, hence, the industries are ultimately
left with the option of LP based software for solving their refining planning models.
Though, some achievement are made in building NLP models but it has short comings of
the solution being local and specific to a particular plant and situation which makes it
difficult for bench marking for global solution (Alattas, et al., 2011).
2.2.3 MILP
When an LP contains integer variables, or certain other type of variables, it becomes an

MILP problem, usually containing both integer and continuous variables.
21
MILP retains the linear objective and constraints but adds an integrality attribute to the
non-negativity requirement on some, or all of the variables. Models that can be
formulated using MILP contains the following features and properties,
 Counting
 Representation of states and yes –no decisions
 Logical implications
 Simple non-linear functions
For example:
(2-8)
(2-9)
(2-10)
(2-11)
where c and p are integer (Kallrath and Wilson, 1997)
Advantages and Limitations of MILP
One of the advantages of MILP approach is that it provides a general framework for
modeling a large variety of problems such as multi-period planning, job shop scheduling
and supply chain management problems.
However, the major difficulty lies in the computational expenses that may be involved in
solving large scale problems, which is due to the computational complexity of MILP
problems which are non-deterministic polynomial time hard (NP-hard).
22
2.2.4 MINLP
A technique to solve optimization problems which allows some of the variables to take
on binary, integer, semi-continuous or partial integer values, and allows nonlinear
constraints and/or objective functions (Kallrath and Wilson, 1997).
2.3 Planning in oil refineries
Petroleum refineries are increasingly concerned with the improvement of their

planning operations (Pinto et al., 2000).Well developed, fairly standard and widely
understood tools and technologies are used for refinery planning. However,
evolutionary changes are expected e.g. refinery models need to be more accurate and
improved by the combination of rigorous model runs plant test, vendor data e. t. c.
(Pelham and Harris, 1996).
Applications of planning models in the refinery and oil industry include crude selection,
crude allocation for multiple refinery situations, and partnership model for negotiating
raw material supply and operations planning. And with appropriate planning, new
technologies for process operations can be integrated for profitability.
Due to increase in scope and size of optimization problems over the years, refiners are
forced to engage the services of optimization software for quick and efficient solution
for profit maximization. Activities in planning involve optimization of raw material
supply, processing and subsequent commercialization of final products over one or
several time periods (Pinto et al., 2000).
The commercial software packages, such as Refinery and Petrochemical Modeling

System (RPMS) (Shah et al., 2011), General Algebraic Modeling System (GAMS), Process
Industry Modeling System (PIMS) (Zhang and Zhu, 2006), Petroplan and host of others,
are developed to solve the wide range of practical problems that arise in refineries.
RPMS is based on a very simple model which mainly is composed of linear relationships,
the production plans generated by these tools are interpreted as general trends (Pinto
et al., 2000). Li et al. (2005) presented a refinery planning model that utilizes simplified
empirical nonlinear process models. He demonstrated how CDU, FCC and product
23
blending models are formulated and applied to refinery planning. The refinery planning
model was formulated in GAMS.
2.3.1 Types of planning models
There are two main types of planning models, these are:
 Rigorous models
 Empirical models
2.3.1.1 Rigorous models
These are models that are based on the theoretical understanding of the system and
process variables interactions. Application of conservation principles are the bases of
this method (i.e. energy and material balances) and equilibrium phase relationships.
One of the major advantages of this model is the ability to formulate them before
putting the system into operation (Al-Qatahni and Elkamel, 2010).
2.3.1.2 Empirical models
These are data driven models, also known as black box models. These models are found
to be useful when rigorous models cannot be implemented due to limited resources or
are found to be complex. In this type of planning model, the systems are seen in terms of
inputs, outputs, and their relationships without any knowledge of the systems internal
mechanism.
2.3.2 Modelling of process units for refinery planning
2.3.2.1 CDU
CDU is one of the major processing units in the refinery downstream operation. Two
types of model are usually used in CDU modelling, namely; Empirical and Rigorous (Li,
2004).
Empirical models for CDU
Empirical models make use of empirical correlations to determine material and energy
balances for CDU. It was initially proposed by Packie (1941), and Watkins (1979)
24
expanded on it. These correlations are also very much ideal for preliminary design
purposes (Li, 2004).
Rigorous models for CDU
Rigorous models simulate a CDU taking into account phase equilibrium, heat and mass
balance all through the column. The advantage of such rigorous model includes results
on flow rates, compositions of streams internally and externally and optimal operating
conditions. Research work has continuously been carried out just to improve rigorous
models.
Due to the challenges involved in rigorous modelling, people do work on numeric

models. Some researchers applied the θ method of convergence in modelling of the
main column to obtain a steady state solution that is equal to the field data (Huang,
2000). Other researchers like Hess et al. (1977) and Holland (1983) made use of 2N
Newton -Raphson method on the tower for steady state solution. Russell (1983)
implemented an ‘inside out’ approach to model a CDU, this method uses a combination
an existing technique, which allows a very wide specification. Very little information is
required at start and the speed of processing is better than other techniques.
Also commercially available software packages, such as Aspen Plus (Aspentech), PRO/II
(SimSci-Esscor) and design IITM (ChemShare), have been developed and are commonly
used. Rigorous simulation model has proven reliable and flexible since the steady state
results can be matched with field data.
2.3.2.2 VDU
In crude oil distillation, one of the goals is to maximise the extraction of the distillate
liquid from the raw crude. These distillates also called reduced crude sometimes serve
as feedstock to other process or processes where values are added to the material.
Crude distilled in the atmospheric tower still contains some more distillable oils that
can be further distilled under vacuum. For 30 degree API crude, the vacuum distillates
can get up to 30 percent volume of the whole crude. This justifies why refiners need to
recover as much more distillates from given crude (Watkins, 1979).
25
2.3.2.3 FCC
FCC unit is another important process unit in a refinery. Currently, the ability of a
refinery to be profitable in a competitive market also depends on the successful
operation of the FCC unit (Slaback, 2004). 40 percent of the feedstock to the gasoline
blending process is from the FCC (Lin, 1993). This implies that FCC unit model should
be included in a detailed study of a refinery.
The purpose of the FCC unit is to breakdown the high-boiling, heavy components in the
crude oil into lighter, more valuable products. Previously, thermal cracking was used to
convert the heavy gas oil to lighter products. However, the catalytic cracking process
has now replaced thermal cracking due to its ability to produce more gasoline with a
higher octane rating (Gary and Handwerk, 2001). Zeolite catalyst is commonly used in
FCC unit operations (Slaback 2004).
Pinto et al (2000) applied a linear model to FCC unit. On the other hand, due to the
nonlinear nature of FCC, a linear model may not give reliable values on the yield and
properties of FCC distillates (Li 2004). Some researchers like Guyonnet et al. (2009),
Pongsakdi et al. (2006) and Neiro and Pinto (2005) used the same FCC unit model by
Pinto et al (2000) may give inaccurate yield of the unit.
Methods have been developed by Gary and Handwerk (2001) to obtain the yields of FCC
unit from simple feed properties and known conversion from charts and figures. These
methods are useful for obtaining typical yields for preliminary studies (Li, 2004) and
shall be implemented in this work.
2.3.2.4 HDT
Hydro treating is a process removing sulphur and or any objectionable elements from
products or feedstock by reacting them with hydrogen to meet product specification
(Gary and Handwerk, 2001).
The HDT receives feed from the CDU and VDU unit. To calculate the fraction of the
products from the HDT, the correlation by Gary and Handwerk (2001) was
implemented i.e. the product yield from the HDT is about 98 volume percent of the feed
(Same boiling range as feed). Yields of light product were assumed (2 volume percent).
This is due to the fact that very little literature on yields of HDT has been published.
26
2.3.3 Current methods for CDU modelling in refinery planning
CDU is one of the processing units that determine the profitability of the refinery. The
raw material for other processes is generated in the CDU. The operation of the CDU is
classified into modes, such as gasoline or diesel mode, which depends upon the
properties of the crude, the process constraint etc. in each of this modes, there exist a
set of predetermined cutpoints.
CDUs of a refinery can be modelled for planning purposes in different ways, including
the product yields and property of the crude oil distillation in the planning model. The
approaches reported include a fixed yield representation model, swing cut model (Li ,
2004), and fractionation index (FI) Alattas et al. (2011). Most of these methods relied
on the distillation behaviour of each crude oil pre-determined by crude assay outside
and distillation simulation program where crude oil are cuts at designated
temperatures and the resultant yields and properties information are passed to an LP
model.
2.3.3.1 Fixed yield approach
Fixed yield method is one of the widely used methods for CDU modelling and it is the
oldest method practised in most refineries. Though efforts has been made to improve or
develop new ways of modelling the CDU. However, the resulting approaches have
always resulted to fixed yield with nonlinear or more complex algorithm.
Fixed yield model is linear based approach that is employed for predicting yield. This
approach uses rigorous or empirical models for the determination of the material and
energy balance for the unit. In the fixed yield model, the cuts fractions as shown in
Figure 2-1 at various designated temperature on the True Boiling Point (TBP) curve
would be determined by the simulator based on the phase equilibrium, material and
energy balances in the whole column under steady state using the crude assay and the
result of the cuts and property information passed on to the LP planning model. Figure
2-2 describes the structure.
27
Figure 2-1 Fixed Yield Model (Adapted from Alattas et al. 2011)
Figure 2-2 flow diagram of a Fixed yield structure representation
Some authors applied a method known as “Adherent Recursion” to optimize cut points.
Here the result obtained from LP model (new cut points) were resent to the simulator
for yields and property updated.
The fixed yield method is rigorous and simple and does not introduce non-convexity
because it uses linear equation for CDU representation in the refinery planning model.
2.3.3.2 Swing cut method
In swing cut approach, physically non-existent cut are defined in the LP model. This is
illustrated in Figure 2-3. GO (Gross overhead) and HN (Heavy Naphtha) are the two
distillates of a CDU. For the LP to have the flexibility of adjusting the volume transfer
ratio of the two distillates, two adjustable pseudo-cuts, shown as the two rectangles in
28
Figure 2-3 are added. The range of a swing cut is defined as a certain ratio on the crude
feed bounded by limits, e.g. segment B-D defined the quantity of a cut (say 5% of the
total fed crude) that could go to either of the two distillates. The final volume ratio of GO
is depicted as segment A-C. In the same way, as soon as the HN swing cut is apportioned,
the final volume transfer ratio HN can be shown as segment C-E (Li, 2004).
In swing cut approach, distillation behaviour is not accurately represented due to the
non-linear nature of the distillate properties. Also swing cuts into adjacent draws are
assumed to be linear whilst the actual property distribution in crude oil is highly non-
linear.
Figure 2-3 Swing cuts model (Reproduced from Li, 2004)
2.3.3.3 Fractionation-index method
Another practice is the Fractional Index (FI) method. In FI practice, the CDU is modelled
as a series of fractionating units as shown in Figure 2-4. It uses the relative volatility of a
component in relation to the top and bottom product streams in the CDU that is
expressed as molar fraction. The top products are fed to the next unit while the bottom
products are withdrawn as CDU products, except for the last unit where the top is
withdrawn as overhead.
29
The FI method was initially proposed by Geddes, (1958) and subsequently Gilbert et al.,
(1966) extended it to CDU. Given the temperature ranges of the cuts, feed crude oil
assay, feed rates and FI values, for each unit the total and component mass balance and
the FI equation for the equilibrium calculation is applied.
Figure 2-4 The FI model (Reproduced from Alattas et al., 2011)
In FI approach, the yield purity is jeopardized because the bottom product being
collected as yield at a particular temperature T1 in PB1 may not have been completely
condensed, so the temperature of the vapour going into the next stage may enter with
entrained liquid of the bottom product. This method introduces non-convexities.
Recent advances in refinery planning
Due to increase in scope and size of optimization problems over the years, the solving of
refinery problem has taken a new aspect as petroleum refiners engage in the use of
software for quick solutions in an attempt to achieve maximum profit out of the
available resources within a particular time horizon.
These software are used to simulate the processing units to generate operating
conditions (e.g. temperature and pressure), yields and their properties to be optimized
in an existing planning model. Some refineries and petrochemical industries modify or
customise the simulators to solve problems based on their specific need. e.g. problems
on production planning, supply chain planning, material requirement planning,
30
blending, crude oil selection and evaluation, uncertainties in the operation environment
and to forecast future trend and opportunities. In the design and planning of refinery
processes to generate cuts which are further optimized, empirical correlations are used
(Watkins, 1979)
2.3.3.4 Production and product blending subsystem
The production and product blending area of the refinery consists of the CDU, FCC,
catalytic reformer (CR) and the HDT and other process units used to convert crude to a
more valuable product while the products qualities such as sulphur, octane number,
vapour pressure and product quantities are not compromised.
Usually the objectives of the individual units are conflicting and thus contribute to a
sub-optimal and many times infeasible overall operation. Optimization of process units,
online optimization of CDU has been carried out by Basak et al. (2002). Optimization of
secondary processing units such as CC and hydrocracking units are also an area of
importance due to the role it plays in the oil refinery. Lee et al. (1970) used dynamic
programming to determine the operating condition in a CC that maximized the profit
function.
Li et al., (2005) went further to integrate the CDU, FCC and product blending models
into refinery planning. What they did was to develop a refinery planning model that
utilizes simplified empirical linear process models with considerations on crude
characteristics, product yields and qualities.
Planning models for production and products blending subsystems have been
developed by Pinto et al. (2000), Moro et al. (1998), Joly et al. (2002), Nero and Pinto
(2005), Li et al. (2004), Khor et al (2008), Alattas et al. (2011). Li et al. (2004)
considered uncertainty in their model; they derived a loss function and applied it in
calculating the expected plant revenues. Uncertainty was considered in crude supply
and product demand. Pongsakdi et al. (2006) addressed the issue of uncertainty and the
financial risk aspect for refinery operations. Jeongho et al. (2010) also considered
financial risk management in refinery planning. In their work, they considered
uncertainty in crude price and product price in the production and product subsystem
of the refinery using LP. Chufu et al. (2008) developed a hybrid programming model for
31
refinery production planning. In their work they considered uncertainty in product
demand using LP.
Very little information is available in the published literature on overall optimization of

oil refineries, but many studies have been reported on the optimization of process units.
The optimization of the production units does not achieve the global economic
optimization of the plant (Pinto et al., 2000), hence the need to integrate the refinery
subsystems.
2.3.3.5 Integrated modelling approach for the subsystems
Since integrated modelling approach provides a better link between the three main
subsystem of the refinery, Neiro and Pinto (2005) presented a nonlinear model for
refinery planning which deals with the supply chain management for multiple refinery
sites, they considered uncertainty in crude and product price as well as demand using
scenario based approach with two time periods, however they did not consider crude
oil operations and distribution in detail. Pitty et al. (2008) presented a dynamic model
of refinery supply chain operation, they considered suppliers and customers, functional
departments in the refinery, production units and refinery economics. However, their
model did not provide explicit details on crude unloading operations and product
distributions. They demonstrated that a dynamic model of an integrated supply chain
can serve as a valuable quantitative tool that aids in such decision-making. Alabi and
Castro (2009) solved a large scale integrated refinery planning using Dantzig-Wolfe and
Block Coordinate-descent Decomposition method. Guyonnet et al. (2009) presented a
simple MINLP model for an integrated refinery subsystem. However, in their model they
considered nonlinearity in the crude oil mix entering the CDU.
2.3.3.6 Future trend in refinery planning
One of the future trends in refinery planning is to find a path towards an automated
intelligent refinery that yields profit at its maximum. Increase in available integrated
information systems that will improve the comparison of plan versus actual with
adjustments for specific real changes from planned feed (Pelham and Harris, 1996).
The need for detailed components and properties characterization of crude oil in order
for the amount of products generated from the processing units to be maximized has led
32
to innovation of new technology processes. In this present times, the models and
simulators makes use of rigorous computations to characterize and obtain sets of
pseudo components which represents the bulk chemical constituents and properties of
the crude oil to be processed (Briesen and Marquardt, 2004), this serves as a basis for
the commercial simulation software.
The rate of change in the refinery planning and operations was observed by Katzer et al.
(2000) to increase in the next 20 years other than the last 70 years. It was discovered
that future refinery planning will undertake the molecular characterization of the crude
oil before refining. This information will create an increase in the modelling details for
adequate processing condition that will reduce sulphur concentration and other
environmental emissions to acceptable threshold at low cost (Briesen and Marquardt,
2004).
2.4 Scheduling in oil refineries
Refinery scheduling has attracted an increasing amount of attention in the past decade.
The reasons for this are (1) to improve productivity and reduce costs (2) substantial
advances of related modelling and solution techniques, as well as the rapidly growing
computational power.
Scheduling operations in refinery operations are complex. This has led to developments
of optimization models and solution for sub-systems. The following are the three main
functional subsystems in a refinery scheduling operations:
 The crude unloading subsystem scheduling
 The production units scheduling
 Products distribution scheduling
The crude oil unloading, storage, blending and charging of CDU is the most essential
part of the entire system and it affects the other two subsystems.
2.4.1 Types of Scheduling models
Scheduling of crude oil operations involves two major approaches. They are discrete-
time formulation and continuous-time formulation. There also exists a third approach
33
but this is not very popular, it is called the mixed time formulation. In the mixed time
formulation, the discrete-time and continuous-time is mixed together.
2.4.1.1 Discrete- time formulation
In discrete time formulation, the scheduling horizon is split into number of intervals of
pre-defined duration. However, the duration may not always be equal. All activities such
as start and end of the task are compelled to happen within the set boundary of the time
horizon. Binary variables are used to enforce a decision as to when a task is carried out
within the set boundary. This enables the model to be solved easily. When the problems
in the model increase, this increases the size of the binary variables and thereby making
the model complex. Problems are usually intractable or require uneconomical
computational time and effort when they are large and are represented using discrete
time formulation. The accuracy of models depends on size of time interval because
discrete- time representation are mere approximation of the real problem. The finer the
time interval, the better the results at the expense of computational time and effort,
while the larger the time interval, the more the results are suboptimal that are
operationally infeasible due to oversimplification of the problem. Bassett et al (1996)
described the complexity of scheduling problem that depends on the length of the
scheduling time interval, the number of units/equipment involved and the number of
tasks and resources available using complex cube.
2.4.1.2 Continuous-time formulation
Continuous- time representation activity start and end time are included explicitly as
optimization variables. Continuous time models allows event to take up any time along
the scheduling horizon which leads to smaller sizes mathematical models with lesser
effort and computational time. This type of representation can be classed into global
and unit-specific event based model (Al-Qatahni and Elkamel, 2010).
The optimization algorithm determines the size of the interval. In global event based
model, a uniform grid applies to the entire event. While in unit specific event based
model, a non-uniform grid where each unit has its own set of time intervals is employed
such that task corresponding to the same event point in different units occurs in
34
different times. Most works presented in the literature are global event based (Al-
Qatahni and Elkamel, 2010).
Jia and Marianthi (2003) developed a MILP model for gasoline Blending and
Distribution Scheduling based on continuous representation of the time domain. They
assumed constant recipes for the blending stage. GAMS/CPLEX was used for the
solution of the resulting MILP formulation. Again Jia and Marianthi (2004) also
developed a comprehensive mathematical model for efficient short –term scheduling of
oil refinery operations based on a continuous time formulation.
Moro and pinto (2004) developed a scheduling model for oil refinery operations based
on unit specific event point formulation using the state task network representation
that was introduced by Kondili et al. (1993).
Comparing the two approaches used for scheduling, continuous time models which
allows event to take up any time along the scheduling horizon while in discrete-time
formulation the scheduling horizon is split into number of intervals of pre-defined
duration which may not always be equal. However, discrete-time based formulation is
still popularly used for solving industrial problems (Maravelias and Grossmann, 2006).
2.4.2 Modelling for refinery scheduling
Knowing and including what is relevant and neglecting the issues that are irrelevant for
the specific decisions is the key issue that lies in building models of refinery scheduling.
MILP has been widely used for scheduling problems because it is rigorous and flexible,
and it has extensive modelling capacity. The application of MILP based scheduling
methods ranges from the simplest single- stage single-unit multiproduct processes to
the most general multipurpose processes. These process scheduling problems are
inherently combinatorial in nature because of the many discrete decisions involved,
such as equipment assignment and task allocation over time (Floudas and Lin, 2005).
For large MILP models, the most widely applied technique employed to solve the
computational problems is decomposition method (Floudas and Lin, 2005). In this
approach, large and computational complex problems are broken down into smaller
and easier sub problems which are then solved to global optimality. The solution to the
main or parent problem is obtained by integrating optimal schedule of the individual
35
sub problems. Decomposition strategies are classified into time-based decomposition
and spatial-decomposition. Researchers have applied the method to decompose large
scale MILP scheduling problems. Harjunkoski and Grossmann (2001) applied spatial-
decomposition technique for scheduling large steel production model. Basset et al.
(1996) adopted the combination of the two strategies based on process recipe. Shah et
al. (2009) presented a general novel decomposition scheme which spatially breaks
down the refinery scheduling problem into sub problems, which when solve to
optimality are then integrated to obtain optimal solution for the whole problem.
They decomposed the whole refinery operations into three categories namely, crude oil
unloading and blending, production unit operations and product blending and delivery.
There are commercially available software for scheduling operations in the refineries,
some are as follows:
 Process Industries Modelling Systems (PIMS) from AspenTech,
 Ominisuite and PetroPlan: These two are for short term scheduling (Li, 2004).
Commercial tools for production scheduling are few and these do not allow a rigorous
representation of plant particularities. For this reason, refiners are developing in house
tools strongly based on simulation (Pinto et al., 2000).
2.4.3 Recent advances in refinery scheduling
2.4.3.1 Crude oil unloading subsystem scheduling
Supply of crude to the refinery takes place at the crude oil unloading area, which
consists of vessels, storage tanks, charging tanks and docking station.
In the crude oil unloading subsystem of the refinery, models have been developed, this
include that of Lee et al.(1996) who addressed the issue of inventory management of a
refinery that imports several types of crude oil which are delivered by different vessel
with MILP. In their paper they reported a short-term scheduling of crude oil inventory
management issue that has to do with crude oil unloading from vessel to storage and
from storage to charging tanks and finally to the CDU. They linearized the bilinear
formulation for the mixing equation by replacing it by individual component flow. They
have also assumed sulphur to be the only specific key component in their blended or
36
mixed crude before being fed to the CDU. This is due to the fact that viscosity and other
properties of mixed crude are hard to obtain.
Shah (1996) reported an MILP model for crude oil scheduling. In his formulation, the
problem was decomposed into an upstream sub problem considering off loading and
storage in portside tanks, and a downstream sub problem involving charging tanks and
CDU operation. The objective of the work was to minimise tank heels. Sulphur content
was also considered as the major key component of the crude.
Wu et al. (2006) included that a short-term scheduling operation in oil refinery should
be viewed from control theory and hence, should be solved by combining enumeration
and heuristic instead of using mathematical programming formulation.
Magalhaes (2004) and Pinto et al. (2002) focused on crude scheduling using MINLP.
Magalhaes (2004) used both continuous and discrete time formulation. In their work,
they developed and integrated model for refinery wide scheduling for current
operational practise to be improved.
Li et al. (2002) proposed a solution algorithm and effective mathematical formulation

for short-term scheduling of crude oil unloading, storage and processing with different
types of oil. They acknowledged the non-convex bilinear constraints generated when
calculating crude oil mixing and proposed a solution algorithm that iteratively solves
two MILP models and a MINLP. However, their algorithm entails solving the NLP at each
iteration.
Reddy et al. (2004) also developed an approach to crude oil scheduling introducing
facility such as single-buoy mooring (SBM). They presented a MINLP formulation and a
MILP–based solution approach. The motivation behind this was due to the nonlinear
nature of crude mix. They concluded that the constraint on the bilinear term of the
blending and accumulation of crudes in the storage tank is still missing. If there is a
mass accumulation in a unit like the charging tank, using individual component flow and
solving the model with MILP in some situation provides inconsistent result.
2.4.3.2 Production unit scheduling
In order to satisfy the demand of the customers, the decision on which mode of
operation to be used in each processing unit while minimizing the production cost and
37
considering intermediate storage capacities is carried out by the schedulers in the
production units. Magalhaes (2004) reported that Moro (2000) in his PhD work studied
some issues related to refinery operations. In his work he initially considers a planning
problem to develop a general formulation for this type of problems, time is dealt with
aggregately. The formulation developed is a general framework for modelling the
refinery process units and tanks and the interrelation among them. The developed
formulation was applied to two different refineries planning problems and were solved
using NLP. The formulation was then used as a basis to develop a general formulation
for scheduling problems, where time is a very important variable and solved as an MILP
model. The model was then applied to an LPG system.
2.4.3.3 Product distribution subsystem scheduling
In product distribution area of the refinery, Rejowski and Pinto (2003) presented a
model on the product distribution part of the petroleum refinery, a multi-product
pipeline and several depots connected to local consumers. Simple transport network
between the refinery has been developed, while some focused on the combined
blending and shipping, some focused more on blending (Guyonnet et al., 2009).
Pitty et al. (2008) presented a dynamic model of refinery supply chain operation, they
considered suppliers and customers, functional departments in the refinery, production
units and refinery economics. However, their model did not provide explicit details on
crude unloading operations and product distributions. They demonstrated that a
dynamic model of an integrated supply chain can serve as a valuable quantitative tool
that aids in such decision-making.
Alabi and Castro (2009) solved a large scale integrated refinery planning using Dantzig-
Wolfe and Block Coordinate-descent Decomposition method.
2.5 Integration of planning and scheduling
The integration of planning and scheduling has received increasing attention in recent
years. This is due to refinery interest in improving the overall competitiveness in the
global market place by reducing costs and inventories while meeting due dates. While
there has been progress towards integrating planning and scheduling, performing
simultaneously these tasks still remains elusive. This is due to the fact that
38
simultaneous planning and scheduling involves in principle solving the scheduling
problem for the entire planning horizon. This, however, results in a very large scale
optimization problem since the problem is defined over long time horizons (Muge and
Grossmann, 2007). Dan and Marianthi (2007) indicated that the boundaries between
planning and scheduling problems are not well established and there is an intrinsic
integration between these decision making stages, lots of work in the literature are
addressing the simultaneous consideration of planning and scheduling decisions.
Maravelias and Sung (2009) wrote in their review that planning and scheduling can be
integrated following the hierarchical approach where the planning problem is solved
first to determine production set target while the scheduling problem is solved to meet
these set targets. However, if the scheduling model is used to meet the production
target then it will be solved iteratively. If the interdependences of these two levels are
not considered at the planning stage, the resulting schedules generated by the
schedulers to meet the planning target will be infeasible.
Numerous trade-offs exist between decisions made at the various level of the different
subsystems, due to interconnections between the different levels. Therefore, to attain
global optimal solutions these interdependences between the different planning
functions should be taken into account and planning decisions made concurrently. That
is, planning problems should be integrated (Maravelias and Sung, 2009).One model
cannot be used for both planning and scheduling, if used will give insufficient
information for both levels (Hartmann, 1997).
2.6 Summary
In this chapter, a detailed survey on optimization techniques has been carried out and
the disadvantages and advantages of applying LP method were also discussed. Planning
in oil refineries and current practice in modelling process units for planning in oil
refinery has been extensively reviewed. The review also covered scheduling, its models
for the refinery, and current advances in refinery scheduling. Finally, the integration of
planning and scheduling were reviewed. From the review, it can be summarised that:
 Despite the disadvantages in using linear models for refinery planning, the
advantages supersede. These advantages make LP tools the choice for refinery
39
planning since it is the trend used for most of the simulation packages (Pinto et
al 2000). These tools are well developed (Pelham and Harris, 1996). Finally, the
required details of refinery processes are captured by linear model formulation
(Zhang and Zhu, 2006).
 Rigorous modelling is best applied to process units for planning purposes
because of its level of accuracy (Li, 2004).
 An integrated modeling approach in refinery subsystems planning would
provide a better link. Also inventory management will be achieved while
resolving the issues between the crude oil supply chains (Guyonnet et al., 2009).
 In refinery planning, generally only one crude oil is considered. There is no
report of using two or more crude oils in refinery planning. Therefore, it is a
knowledge gap to deal with refinery planning with two different crude oils pre-
mixed.
In this work, LP shall be the optimization tool to be used. For the process units in the
production and product blending subsystem, rigorous modelling shall be used for
modelling and finally the planning model derived shall then be used in an integrated
refinery subsystem modelling.
40
3 Rigorous Modelling for CDU and VDU
3.1 Introduction
This chapter focuses on developing and analysing models for CDU and VDU. The
CDU and VDU were modelled rigorously in Aspen HYSYS. The flow diagram in
Figure 3-1 illustrates the methodological sequence for developing and analysing
the CDU and VDU model in this chapter. The CDU model was validated by using
data from literature and the summary given in the end.
Figure 3-1 Overview of Chapter Methodology
41
3.2 Process description for the refinery production
Figure 3-2 Configuration of a Refinery (Arofonosky et al., 1978)
Figure 3-2 is a process flow diagram (PFD) used to describe the process being
proposed. The process is made up of five main units namely, CDU, VDU, FCC unit,
HDT unit and products blending header. The crude mix or blended crude oil which
is comprised of different crudes mixed at a defined volumetric ratios enters the
CDU, six different products fraction are obtained which include: Off Gas (OG),
Naphtha (N), Straight run Kerosene(Sr K), straight run Diesel, Atmospheric Gas oil
(AGO) and Atmospheric Residue (AR). The AR enters the vacuum tower as feed
and produces Vacuum Overhead (VO), Low vacuum gas oil (LVGO), High Vacuum
Gas oil (HVGO) and Vacuum residue (Vac Res). Naphtha from the CDU and the
gasoline from the FCC are further sent to the gasoline blending (GB) header. The
LVGO and HVGO from the vacuum tower and the AGO from the atmospheric tower
are sent to the HDT for sulphur removal before entering the FCC for further
processing and the following intermediates are produced: C2 + lighter, Liquefied
petroleum gas (LPG), gasoline, coke and total gas oil (TGO). Vacuum residue from
42
the vacuum tower and the TGO from the FCC are sent to the Fuel-oil (FO) blending
header for No. 6 Fuel oil production (Arofonosky et al., 1978).
3.3 Modelling of CDU and VDU in Aspen HYSYS®
In Aspen HYSYS® the modelling of the CDU and VDU was carried out in the
following sequence:
 Crude oil characterisation

 Blending of the crudes in different volumetric ratios
 Modelling of pre-Heating train
 CDU modelling
 VDU modelling
 Validation of CDU model
Aspen HYSYS is a modular mode simulation tool. Process equations which relate
the outlet stream and the performance variables of a unit to the inlet stream
variables are grouped within unit operation blocks. Each unit operation block is
then solved one at a time in a sequence with the output of one block becoming the
input to another. The direction of flow of information usually follows the material
flow (Aspen HYSYS, V7.2).
3.3.1 Crude oil characterization
In crude oil characterization, the feedstock for the CDU which is made up of two
types of crude oil were blended to form one crude mix before being processed into
different products as demanded. Crude characterisation is usually the first step
taken to facilitates other calculations and the minimum requirement are: (a) whole
crude True Boiling Point (TBP) curve, (b) whole crude American Petroleum
Institute (API) gravity and (c) whole crude light ends analysis (Watkins 1979).
Crude is characterised to provide a good representation of it during modelling. A
good start is acquiring a distribution curve of a crude assay such as the TBP. Other
useful data include property curves like density and viscosity as well as bulk
properties. Aspen HYSYS was used to carry out a rigorous steady state modelling
with the supplied assay data of the two different crudes to generate internal TBP,
43
sulphur and viscosity curves of the crude mix. The properties and types of the
crude used for the oil characterisation are in Table 3-1 and Table 3-2 while the
assay of the crudes used can be found in Appendix A.
Table 3-1 Crude oil types and properties
Properties Ratawi Brent

API 24.5 38.5
Sulphur wt. % 3.88 0.43
Viscosity (cp) 118.8 4.98
Molecular weight 320 209
Table 3-2 Sulphur concentration, API and Viscosity in the crude oil mixture
Volumetric mixing ratio Sulphur (wt. %) API Viscosity (cp)
0/100 0.43 38.50 4.98
2/98 0.46 34.96 4.07
4/96 0.50 34.81 4.15
25/75 1.02 33.45 5.59
40/60 1.79 29.85 15.78
50/50 2.11 30.98 23.48
60/40 2.47 27.36 32.26
75/25 2.97 25.54 54.92
100/0 3.88 24.50 118.80
From the crude assay data, the distillate production and product distribution from
the crude mix was then estimated. In Aspen HYSYS the minimum amount of
information required for crude characterization are distillation curve and at least
44
one of the following bulk properties: molecular weight, Density, or Watson K
factor.
The light ends of the whole crude that is the pure components with low boiling
points was not provided for the crude used, so Aspen HYSYS was used to
automatically calculate the light ends of the blended crude by interpolation. These
are components in the boiling range of C2 to n-C5.
3.3.2 Blending or mixing of crude oil
Petroleum mixtures used as feedstock in the atmospheric tower are made up of

several hundreds of components. Modelling all these components is really not
feasible. Therefore, petroleum fractions are used in modelling. These are
represented as pseudo-components, characterised by the average boiling point
extending over a range of 5-10o C and the fraction density (Huang, 2000).
Two different types of crude were blended at different volumetric ratio e.g. 50/50.
Blends are determined based on the design configuration of the CDU. The blended
crude can be characterised as sour crude due to high sulphur content (2.11 wt. %).
In the blend, Aspen HYSYS automatically cuts the oil into 43 pseudo-components
and blended the two assays into one set of pseudo-components. Table 3-3 lists the
properties of the 43 pseudo component.
3.3.3 Pre-Heat train
The heat train is comprised of heat exchangers and furnaces. For simplicity, the
detailed modelling of the heaters and furnace shall not be dealt with in this work
but would be modelled as a simple heater.
45
Table 3-3 Pseudo Components and Physical Properties of crude oil mix
Density Viscosity 2
Comp..Name NBP (oF) Mol. Wt. (API) Viscosity 1 (cP) (cP) Watson K
NBP_110 109.6572 82.29238 80.63242 3.22E-10 1.05E-10 12.42523
NBP_137 136.9998 86.38583 75.44255 1.60E-02 5.32E-03 12.31227
NBP_164 163.5259 97.80189 61.03144 1.93E-02 5.97E-03 11.62216
NBP_189 188.8577 102.7134 58.37977 2.03E-02 6.08E-03 11.61534
NBP_213 213.1707 109.449 56.12497 2.01E-02 6.11E-03 11.61908
NBP_241 240.8172 114.833 53.30542 2.70E-02 8.28E-03 11.59912
NBP_267 266.7704 121.5616 51.16321 2.79E-02 8.46E-03 11.60454
NBP_292 291.604 129.0114 49.00944 3.06E-02 8.60E-03 11.59693
NBP_318 318.061 135.1389 46.39778 4.56E-02 1.12E-02 11.56176
NBP_344 344.3656 139.2865 44.89812 5.67E-02 1.37E-02 11.59211
NBP_371 370.6562 148.1015 42.53774 7.82E-02 1.85E-02 11.56032
NBP_397 396.8318 155.8196 41.02245 9.83E-02 2.25E-02 11.57885
NBP_423 422.9229 166.7885 39.2412 0.131533 2.80E-02 11.57449
NBP_449 448.901 178.5111 37.64467 0.192253 3.43E-02 11.57768
NBP_475 474.765 192.1125 35.97294 0.281984 4.03E-02 11.57101
NBP_501 500.6102 205.5723 34.39887 0.440065 5.16E-02 11.56697
NBP_527 526.7237 218.6931 32.93148 0.700064 6.46E-02 11.56765
NBP_553 552.8504 233.3523 31.4009 1.085616 7.81E-02 11.56028
NBP_579 578.8869 249.7987 29.90041 1.491025 9.58E-02 11.55114
NBP_605 604.9589 265.6074 28.64219 1.921181 0.119293 11.55621
NBP_631 631.1607 283.6535 27.29981 2.414852 0.145585 11.55259
NBP_657 657.2655 300.6864 26.00811 3.026131 0.178205 11.5493
NBP_683 683.0257 315.8744 24.86838 3.467623 0.216761 11.55321
NBP_709 708.5046 333.221 23.71413 4.52638 0.272423 11.55254
NBP_735 734.6736 390.9735 22.49214 6.843832 0.709777 11.54654
NBP_761 761.024 471.4655 21.44972 9.389477 2.285399 11.55211
NBP_787 786.8905 547.3737 20.46646 11.63377 5.430985 11.55835
NBP_823 823.4059 618.4999 19.29659 13.56353 10.32791 11.58028
NBP_874 874.0338 721.0835 17.64109 16.85143 44.41249 11.60185
NBP_925 924.5961 864.6155 15.92057 22.07416 239.8425 11.61114
NBP_975 975.2554 1031.628 14.20575 28.03564 1300.467 11.6144
NBP_1025 1025.373 1197.455 12.6485 35.64969 7792.487 11.62251
NBP_1072 1072.085 1366.691 11.14675 46.80521 54790.5 11.62078
NBP_1119 1118.76 1535.381 9.611797 96.11141 503669.1 11.61133
NBP_1170 1169.634 1682.915 8.174267 2351.121 3473069 11.61522
NBP_1267 1267.396 1844.002 6.648318 4486162 47900123 11.71365
NBP_1379 1378.731 2068.365 4.204098 2E+11 2.39E+09 11.74852
NBP_1422 1422.462 2255.771 2.133449 3.01E+15 2.21E+10 11.66027
NBP_1559 1559.338 2302.5 0.283276 1.02E+19 1.27E+11 11.77109
NBP_1625 1625.063 2316.671 -1.34169 1.70E+21 4.76E+11 11.75075
46
NBP_1690 1689.971 2328.181 -2.54611 4.36E+22 1.13E+12 11.76161
NBP_1763 1763.472 2341.618 -3.89724 2.25E+24 2.76E+12 11.76954
NBP_1888 1887.707 2367.475 -6.4602 8.44E+25 7.09E+12 11.74409
3.3.4 CDU modelling
Aspen HYSYS was used to develop a generic model for the CDU, using a column
model, solved using HYSIM –Inside-out method which involves the principle of a
single equilibrium stage. Peng-Robinson (PR) equation of state is used as fluid
package. PR is used because over the years it has been updated for improved
calculation of vapour- liquid equilibrium (VLE) (Aspen HYSYS V7.2). The Soave-
Redlich-Kwong (SRK) and PR equations have been tested on hydrocarbon
mixtures, with both giving nearly similar results (Sim and Daubert, 1980).
The yield prediction from Aspen-HYSYS for the blended crude in the CDU is shown
in Table 3-5. The main specification for each intermediate or yield is the boiling
range of its components. The boiling range specification for the hydrocarbon
stream is expressed by either ASTM 95% points or the TBP endpoint. The ASTM
95% represents the temperature corresponding to 95% vaporisation on the ASTM
D86 distillation curve, while the TBP is the temperature at which the whole
fraction vaporises on the TBP curve. The TBP and ASTM D86 curve for the crude
mix in this work is shown in Figure 3-3.
47
Figure 3-3 TBP and ASTM D86 curve for the crude mix
The fractional distribution of each cut is analysed with the curve when it is plotted
against volumetric yield of the cuts. The TBP volumetric yield curve of the
individual crude oil entering the CDU was found in the crude assay otherwise it can
be provided by the refinery. Table 3-4 shows the volume percent of products
distilled at different temperature.
The values of the CDU fractions obtained in Aspen HYSYS are in volume fraction as
shown in Table 3-5 but the current practice provides CDU fractions in weight or
volume percent. This leaves us with the conversion from volume to weight. For this
conversion to be obtained, the was used. Table 3-6
shows the conversion to volume percent. In all the Tables 3-4, 3-5, and 3-6 below
the off gas component of the products from the CDU as shown in the Figure 3-2
was not part of the products listed, this is because of the insignificant quantity of
the off gas generated in the CDU with this crude mix. The off gas will subsequently
48
not appear in the list of products yield from the CDU. However, this may vary from
crude oil to crude oil.
Table 3-4 TBP and volume percent distilled
Vol. % Distilled Temp. (oF) Sulphur Content

(Wt. %)
10 236.51 0.0623
20 322.60 0.2285
30 436.60 0.6278
40 538.82 0.9332
50 646.42 1.0958
60 754.66 1.4837
70 885.78 2.0825
80 1040.63 3.2703
90 1282.84 5.1094
95 1520.49 6.8619
100 1859.34 11.6034
o
Table 3-5 The mixed crude Fractions TBP ( F) cut point Specification using Aspen HYSYS
prediction
Comp. Name IBP (oF) EBP(oF) Vol. Fraction Mass

Fraction
Naphtha 158 356 0.2291 0.195
Kerosene 356 554 0.0978 0.090
Diesel 554 644 0.1708 0.165
AGO 644 698 0.0520 0.052
Atm. Residue 698 1859 0.4503 0.497
IBP, Initial Boiling Point. EBP, End Boiling Point
49
Table 3-6 The mixed crude cut point Specification at ASTM D86using Aspen HYSYS prediction
Products ASTM95%(oF) Sulphur. (Wt. Flow TBP Vol.% Wt. %

%) rate (oF)
(m3/h)
Naphtha 318.7 0.0992 137.70 337.10 20.10 16.85
Kerosene(Light 446.5 0.4300 67.64 472.70 9.87 7.70

Distillate)
Diesel(Heavy 633.1 0.9008 117.20 676.80 17.10 14.23
Distillate)
AGO 733.9 1.2130 35.33 774.60 5.15 4.37
Atm. Residue 1534 3.5300 327.80 1712 47.80 56.90
3.3.5 VDU modelling
As the reduced crude leaves the CDU, it is re-heated in the vacuum furnace and fed
into the vacuum tower with vacuum steam. The primary purpose of the vacuum
steam is to reduce the hydrocarbon partial pressure in the flash zone of the
vacuum tower. When the hydrocarbon partial pressure in the flash zone is
lowered, vaporisation takes place and hence distillate production. The
specification for the vacuum gas oil is also in form of TBP. The approach used for
the CDU is also used for VDU. The yields at TBP and yields at ASTM D86 are
summarised in Table 3-7 and Table 3-8 respectively as predicted by Aspen HYSYS
for the blended crude. Figure 3-4 is the schematic of the generic CDU and VDU
model developed in Aspen HYSYS.
50
o
Table 3-7 Vacuum Tower Fractions TBP ( F) cut point Specification predicted by Aspen HYSYS
Comp. Name IBP (oF) EBP(oF) Vol. Fraction Mass Fraction
Vac. Ovhd 110 698 0.5497 0.502

LVGO 698 806 0.0934 0.0963
HVGO 806 1040 0.1564 0.1676
Vac Residue 1040 1859 0.0223 0.2335
IBP, Initial Boiling Point. EBP, End Boiling Point
Table 3-8 Cut point Specification at ASTM D86 for Vacuum Tower predicted by AspenHYSYS
Products ASTM 95% Sulphur (wt %) Flow rate TBP Wt.% Vol. %
(oF) (m3/h) (OF)
Vac. Ovhd 432.3 0.046 2.26 456.3 0.67 0.68

LVGO 681.4 1.057 32.33 718.3 8.82 9.76
HVGO 873.5 1.702 90.62 929.1 26.08 27.37
Vac 1612 4.617 205.80 1812.0 64.47 62.17
Residue
Since proper operation of the CDU, VDU, FCC, and HDT with regard to side
products cut points leads to meeting the economic objective of a refinery, this
work is focused on only these four units. For simplification, the other auxiliaries
like the pre-heat train, the desalter and the separator as shown in Figure 3-4, were
not modelled in detail.
51
Figure 3-4 Generic CDU and VDU model in Aspen HYSYS
3.3.6 Validation of CDU model
The product yields from CDU model in Aspen HYSYS were validated to show that
the yields can match with what is in the literature. A Venezuelan crude oil was
taken from Li et al. (2005) which is the same as in many other literatures like
Alattas et al. (2011), Watkins (1979) and simulated based on the TBP cut points
obtained from the crude assay at a throughput of 100,000 barrel per day (BPD).
The bulk properties of the crude oil is listed in Table 3-9 while its cut fraction
ranges from the distillation unit is shown in Table 3-10.
52
Table 3-9 Bulk properties of Tia Juana crude oil (Venezuela), (Watkins, 1979)
Vol. % distilled T(oF)
0 205.99
5
242.01
10 216.28
30 460.73
50 643.62
70 859.20
90 1076.60
95 1127.10
100 1174.70
Bulk Properties
API 31.6
SG 0.8686
Sulphur (wt %) 1.08
Viscosity: (cp)
Kinematic cst, @ 70oF 16.1
Kinematic cst, @ 100 oF 10.2
53
Table 3-10 TBP ranges of CDU fractions (Li et al., 2005)
CDU Fractions Boiling Range (oF)
GO(Gross Overhead) EBP 276.5-290.9

HN (High Naphtha) IBP 234.4
EBP 340.6-418.4
LD (Light Distillate) IBP 257.3-325.1
EBP 577.9-631.1
HD(Heavy Distillate) IBP 488.6-545.0
EBP 711.3
BR(Bottom Residue) IBP 611.8-630.6
The crude oil cut fractions or yields were estimated in CDU model in Aspen HYSYS
and the yield generated were compared with those obtained from literature. The
relative percentage errors were calculated. The validation for the Aspen HYSYS
model was carried out under the same conditions of EBP temperature of the
maximizing heavy naphtha mode (MN) as the main case of Li et al. (2005) and
Alattas et al. (2011). Gross overhead was also included for validation because it is
included in the data provided by Li et al. (2005). Li et al. (2005) made use of
empirical correlations but in this work, rigorous model is used and the results
obtained. However, the rigorous model needed to be compared with other method.
The CDU fractions obtained matched with Li et al. (2005). The result obtained is
shown in Table 3-11 and a chart is also plotted for better view of Aspen HYSYS
validation results as seen in Figure 3-5.
Table 3-11 Comparison of product yields from CDU model in Aspen HYSYS and Li et al. (2005)
Products Empirical-(Li et al. Rigorous-(This Relative Error

2005)(Vol. %) work) (Vol. %) (%)
GO 13.61 13.69 0.5
HN 10.60 11.73 9.6
LD 20.98 21.10 0.5
HD 6.91 6.97 0.9
BR 47.04 46.50 1.1
.
54
Figure 3-5 Comparison of product yields from CDU model in Aspen HYSYS and Li et al. (2005)
From Table 3-11 and Figure 3-5conclusion can be drawn that the CDU fractions
obtained from Aspen HYSYS is a true representation of the product cuts.
The cut fractions or yields from Aspen HYSYS model were also validated using the
ASTM D86 TBP from Aspen HYSYS and correlation from literature e.g. Li (2000)
polynomial regression method.
Table 3-12 shows the comparison between ASTM D86 to TBP. The ASTM D86
values were obtained from the Aspen HYSYS model. The table does not contain
products yields like gross overhead from the CDU because the polynomial
regression method used for the comparison in Li (2000) does not contain it either.
The names of the products yields as listed in Table 3-12 is the same as used in Li
(2000). This has been maintained for proper validation of the CDU.
55
Table 3-12 Conversion of ASTM D86 to TBP
S/N ASTM D86 TBP (OF)

Products (OF)
(Li 2000) Regression Aspen
Names
HYSYS
1 Naphtha 318.7 335.7 337.1
2 Kerosene 446.5 467.9 472.7
3 Diesel 633.1 664.7 676.8
4 AGO 733.9 773.9 774.6
5 Atm- Res 1534.0 1573.3 1712.0
The plot of the conversion approach was matched with Aspen HYSYS values as
shown in Figure 3-6.
Figure 3-6 Conversion of ASTM to TBP (using Li, 2000)
3.4 Process analysis for CDU and VDU under different crude
mix ratios
3.4.1 25/75 volumetric mixing ratio
The 25/75 crude mix was also used in Aspen HYSYS for CDU and VDU simulation
and the following results on the products yield were obtained and tabulated in
Table 3-13.From Table 3-13 for example, the off gas products were not included in
56
the analysis and the previous products from the CDU, this is because during the
simulation, the quantity of off gas generated was very small and is considered
negligible in the CDU as previously explained. However, in the VDU overhead gas
called vacuum overhead is considered due to the quantity generated.
Table 3-13 Product yields from CDU and VDU in HYSYS for 25/75 mix
Volumetric Blend 25/75

Products yield Vol. (m3/h) Vol. % Wt. %
Naphtha 180.8 25.77 22.50

Kerosene 75.11 10.71 9.98
Diesel 127.1 18.12 17.82
AGO 37.89 5.39 5.55
Atm- Res 281.7 40.09 44.14
Total 100 100
Vac Ovhd 2.68 0.95 0.93
LVGO 26.21 9.24 8.42
HVGO 41.93 14.38 14.07
Vac res 212.80 75.54 76.59
Total 100 100
The 75/25 crude mix was also used in HYSYS for CDU and VDU simulation and the
following results were obtained and tabulated in Table 3-14.
57
Table 3-14 Product yield from CDU and VDU from HYSYS for 75/25 crude mix

Naphtha 132.90 19.71 16.47
Kerosene 60.34 8.95 8.10
Diesel 108.60 16.11 15.47
AGO 33.34 4.94 4.93
Atm. Res 339.10 50.29 55.04
Total
100 100
Vac- Ovhd 2.71 0.79 0.74
LVGO 32.38 9.44 8.54
HVGO 54.62 15.92 15.01
Vac res 253.4 74.68 75.70
Total
100 100
The 50/50 crude mix was used in Aspen HYSYS for CDU and VDU simulation and
the following results were obtained and tabulated in Table 3-15.
Table 3-15 Products yield from CDU and VDU in HYSYS for 50/50 mix

Naphtha 156.60 22.84 19.52
Kerosene 67.65 9.87 9.12
Diesel 117.10 17.08 16.64
AGO 35.70 5.21 5.30
Atm Res 308.60 45.01 49.42
Total
100 100
Vac Ovhd 2.26 0.75 0.71
LVGO 29.29 9.40 8.50
HVGO 48.43 15.04 14.68
Vac res 231.50 74.82 76.11
Total
100 100
58
3.4.4 Analysis of the results on the different crude mix proportions
Appendix H is the summary of the quantities of various yields obtained from the
different crude mix proportion simulated in Aspen HYSYS for CDU and VDU.
The blend ratio was observed to affect the yield quantity. Ratawi crude is heavy
and sour while Brent crude is sweet. From Appendix H, it was noticed that as the
volume ratio of Ratawi crude increases, the quantity of heavy products yields like
AGO, Atm. Res, Vac Res and TGO increases. Likewise, if the volumetric ratio of
Brent increases, there will be more of the lighter products yields like naphtha,
kerosene and fuel gases. This variation is largely due to changes in the crude mix
density in the different proportions. The property of crude mix like sulphur,
density and viscosity affects the final product yields of interest. If the refiner wants
more of heavy products, he will consider the API density of the crude. While the
sulphur content of the crude determines the refining requirement to meet the
quality specification of the final product yields. The sulphur in the crude mix also
affects the cost of the crude and this relatively affects the profit of the final
products. From Appendix H conclusion can be drawn that the values obtained from
Aspen HYSYS are good representation of industrial and experimental data.
From the two crudes selected (Ratawi and Brent), the more Ratawi crude in the
crude mix ratio, the more the refining requirement due to its high sulphur content.
3.5 Summary
In this chapter, the following were carried out
 A rigorous CDU and VDU model was developed using Aspen HYSYS® with
two crude oils mixed together in different volumetric mixing ratios.
 The CDU model from Aspen HYSYS® was validated and the results obtained
showed that the products cuts from the rigorous model developed are a
good representation of CDU cut fractions.
 The rigorous model for CDU and VDU were then used to simulate more
volumetric ratios. e.g. 0/100, 2/98, 4/96, 8/92, 10/90… 100/0, etc. It was
noticed that as the volume ratio of Ratawi crude increases, the quantity of
59
heavy products yields such as AGO, Atm. Res, Vac Res and TGO increases.
When the volumetric ratio of Brent increases, there will be more of the
lighter products yields such as naphtha, kerosene and fuel gases.
60
4 Refinery Production and Product Blending
Subsystem Planning Based on Fixed Yield
Approach
4.1 Introduction
In this chapter, the information on products yield obtained from the rigorous and
empirical modelling on the various units e.g. CDU, VDU, HDT and FCC is used to
develop a simplified mathematical programming planning model for refinery
planning of the production and products blending subsystem based on fixed yield
approach.
The information on the crude assay and crude rate, and the cut points of the
product from the TBP, which are known priori and simulated in Aspen HYSYS in
chapter 3 of this paper, shall be used.
In fixed yield approach, to predict the intermediate products yields and qualities
obtained from the processing of the crude mix in the CDU and VDU, a linearization
of the approximated results from the rigorous models is used in the refinery
planning model. Tabulated values of yield and quality of intermediate products
produced for each mode are used as its linear model of the CDU and VDU (Brooks
and Walsem, 1999).
The planning model developed is then used to carry out planning under different
volumetric ratios (up to 53). A plot of profit against volumetric ratios is obtained.
61
4.2 Process description
Figure 4-1 Refinery Topology for the Process
Figure 4-1 is a process flow diagram representing a refinery process made up of

five main units namely, CDU, VDU, FCC unit, HDT unit and a products blending
header. Figure 4-1 is the same with Figure 3-2. However, in this Figure X1-X15 are
used to identify the variables for ease of modelling and calculation. The variables,
X1-X15 are also used to define the direction of stream flow.
The mixed or blended crude oil enters the CDU; six different product fractions are
obtained which includes: Off Gas (OG), Naphtha (N), Straight run Kerosene (Sr
Kerosene), straight run Diesel (Sr Diesel), Atmospheric Gas oil (AGO) and
Atmospheric Residue (AR). The AR enters the vacuum tower as feed and produces
Vacuum Overhead (vac-ovhd), Low vacuum gas oil (LVGO), High Vacuum Gas oil
(HVGO) and Vacuum residue (Vac Res). Naphtha from the CDU and the gasoline
from the FCC are sent to the gasoline blending header. The LVGO and HVGO from
the vacuum tower and the AGO from the atmospheric tower are sent to the HDT
62
for sulphur removal before entering the FCC for further processing and the
following intermediates are produced: C2 + lighter, Liquefied petroleum gas
(LPG), FCC-gasoline, coke and total gas oil (TGO). Vacuum residue from the
vacuum tower, the TGO from the FCC are sent to the gas oil (FO) blending header
for No. 6 Fuel oil production.
4.3 Mathematical programming model for CDU and VDU
The following assumptions were made:
 Perfect mixing is assumed

 There is no accumulation in the units
 The mass balance on yields of individual units is used to determine the
product flow rates. These are also function of the feed.
 Bound constraints are satisfied by individual process units.
 The properties of the inlet feed stream is not varying.
 Parameters such as temperature, pressure, heat transfer and energy
balance which were used for the process units are not considered in this
model.
4.3.1 CDU material balance constraints
The material balance constraints are in the form of inequalities such that output is
equal or less than input. This is because equality makes LP models constraining
and often prevents finding feasible solutions (Arofonosky et al. , 1978).
Based on fixed yield approach, the material balance for each product stream is
obtained by multiplying the feed into the refinery process unit (e.g. CDU) by its
coefficient. Therefore, the sum of the entire product coefficients from each unit is
equal to unity, on this premix, the following set of equations (Eq. 4-1 to 4-10) is
developed for CDU and VDU. For consistency, the feed into any process unit is
assigned negative sign indicating it is been consumed in that unit, while product
from such unit is positive and all balance constraint are on the right hand side, and
equated to zero.
63
(4-1)
(4-2)
(4-3)
(4-4)
(4-5)
(4-6)
4.3.2 VDU material balance constraint
(4-7)
(4-8)
(4-9)
(4-10)
4.3.3 Capacity constraint for CDU and VDU
The feed rates of crude oil to the units, averaged over a period of time can be any
value ranging from zero to the maximum plant capacity.
(4-11)
(4-12)
64
4.4 Empirical correlations used in mathematical programming
model for HDT and FCC units
4.4.1 HDT unit and material balance
Some of the products slates or cuts from the VDU serve as feed to the HDT due to
the high sulphur content before being sent to the FCC. The vacuum residue
produced from the VDU is sent to the FO header for blending to produce No 6 fuel
oil. Since there are very little publications on hydrotreater yields, an empirical
method described by Gary and Handwerk (2001) is adopted for this work. 98% of
products yield on feed are expected, while other feed conditions remains the same.
The yields obtained for 50/50 crude mix are listed in Table 4-1. The method used
to calculate the yield for the FCC unit can be found in appendix C. Figure 11.
Table 4-1 Products cuts from Hydro treating of the oils
Products Yields Vol. (m3/h)
Vol. (%) Wt. (%)
LVGO 26.22 24.93 29.29
HVGO 42.47 50.12 48.43
AGO 31.31 24.95 35.70
Implementing the correlation by Gary and Handwerk (2001) on HDT, the material
balance constraint is then:
(4-13)
4.4.2 FCC unit and material balance
Feed from the HDT enters the FCC unit. The FCC unit produces cyclic oil, and this
oil is blended in the FO header with the vacuum residue from the VDU for the
production of No 6 fuel oil. Empirical correlation for FCC by Gary and Handwerk
(2001) is adopted in this work. A fixed conversion rate is used.
65
The properties of the feed from the HDT after reducing the sulphur content by the
method proposed by Gary and Handwerk (2001) were obtained. A zeolite catalyst
was used and a Watson K factor of 11.8 for the gas oil feed into the FCC. The
information on the Watson K factor is obtained from Aspen HYSYS. This is because
the feed conditions into the HDT and subsequently into the FCC remained constant
from the VDU. The conversion level was obtained with respect to the gas oil feed
which is straight run gas oil or straight run plus coker gas oil. In this work the feed
is straight run gas oil from the CDU, tracing the K factor to the gas oil curve in
Figure 1 in appendix A. 68% conversion level was obtained. The rest of the yields
were obtained based on this conversion rate as in Appendix A from Figure 2 to
Figure 10.
The values obtained for the FCC product cuts in weight and volume percent for
50/50 crude mix are shown in Table 4-2.
Table 4-2 Yields fraction from FCC
Volumetric Blend 50/50
Products Vol. (m3/h) Vol.% Wt.%
Coke 5.8 3.96 5.20
C2+liighter 12.11 8.28 5.10
LPG 24.37 16.65 13.72
TGO 44.32 29.53 38.40
Gasoline 59.30 41.56 37.60
(4-14)
(4-15)
(4-16)
66
(4-17)
(4-18)
4.4.3 Capacity constraint for HDT and FCC units
4.5 CDU and VDU mathematical programming planning model

validation
The model developed was validated in GAMS with CPLEX solver. The yields
obtained in the 50/50 volumetric ratio in Aspen HYSYS are used to validate the
model. This is to determine the efficacy of the model. The CDU capacity of
100,000bbl/day and VDU capacity of 46,500bbl/day were used.
67
Table 4-3 The Coefficients generated for CDU and VDU in Aspen HYSYS at 50/50 crude mix
Products Coefficients Values from Aspen HYSYS Vol. %
Naphtha 22.84
Kerosene 9.87
Diesel 17.08
AGO 5.21
ATM RES. 45.01
LVGO 9.40
HVGO 15.04
Vacuum Ovhd 0.75
Vacuum Residue 74.82
Generated products from CDU and VDU in volume percent from GAMS are then
compared with the values obtained from Aspen HYSYS and relative difference
calculated. The material balance in the individual products streams is obtained by
multiplying the feed into the units by its coefficients.
From the validation results in Table 4-4, the relative error ranged from 0.48 to
3.98 which is acceptable range. Table 4-5 and 4-6 shows the split for the different
process units.
68
Table 4-4 Validation results for CDU and VDU
GAMS Aspen HYSYS Relative Difference

PRODUCTS Vol.(kbbl/day) Vol. (m3/h) Vol. (m3/h) %
(50/50)
Naphtha 22.84 155.75 156.60 0.50
Kerosene 9.87 67.30 67.65 0.52
Diesel 17.08 116.47 117.10 0.54
AGO 5.21 35.53 35.70 0.48
ATM RES. 45.00 300.04 308.60 2.77
LVGO 4.14 28.22 29.29 3.65
HVGO 6.84 46.66 48.43 3.65
Vacuum Ovhd 0.32 2.17 2.26 3.98
Vac. Residue 32.70 222.98 231.50 3.68
Table 4-5 Validation of CDU model with yield from GAMS
Products GAMS Aspen Relative

HYSYS Difference
Vol. Vol. (m3/h) Vol. (m3/h)
(kbbl/day)
Naphtha 22.84 155.75 156.60 0.0050
Kerosene 9.87 67.30 67.65 0.0052
Diesel 17.08 116.47 117.10 0.0054
AGO 5.21 35.53 35.70 0.0048
Atm-Res 45.00 300.04 308.60 0.0259
Total 100.00 675.09 685.65 1.54
69
Table 4-6 Validation of VDU model with yield from GAMS
Products GAMS Aspen HYSYS Relative

error
Vol.(kbbl/day) Vol. (m3/h) Vol. (m3/h)
Vac-Ovhd 0.32 2.18 2.26 0.0372
LVGO 4.14 28.22 29.29 0.0365
HVGO 6.84 46.66 48.43 0.0365
Vac-Res 32.70 222.99 231.50 0.0367
Total 44 300.05 311.48 0.0367
It can be observed from the validation Tables 4-5 and 4-6that the CDU and VDU
model when implemented in GAMS presented yield with difference ranging from
0.5% to 2.6% for CDU and 3.65% to 3.72% for VDU. Considering the error ranges
which are in percentage, the simplified linear Programming models developed are
good representation of the CDU and VDU with respect to the cut fractions obtained
from Aspen HYSYS. Similarly, Figures 4-2 and 4-3 are plots comparing the volume
of product obtained from GAMS and Aspen HYSYS which shows a good match for
all the products for the different process units.
Figure 4-2 CDU yield from Aspen HYSYS model validation with GAMS
70
Figure 4-3 CDU yield from Aspen HYSYS model validation with GAMS
Further validation was carried out using the 25/75 volumetric mixing ratios for
the CDU and VDU.
Table 4-7 Product yield of 25/75% crude oil mix from HYSYS
Product Yield (25/75 mix) Vol.(m3/h) Vol. %
Naphtha 180.8 25.73
Kerosene 75.11 10.69
Diesel 127.1 18.09
AGO 37.89 5.39
AtmRes 281.7 40.09
Vac ovhd 2.681 0.945
LVGO 26.21 9.245
HVGO 41.93 14.78

Vac res 212.8 75.03
71
Table 4-8 Validation of yield results for 25/75 volumetric mixing ratio from Aspen HYSYS
Products
(25/75) GAMS Aspen HYSYS Absolute Relative
error error %
Vol. Vol. (m3/h) Vol. (m3/h)
(kbbl/day)
Naphtha 25.73 175.50 180.80 5.30 2.93
Kerosene 10.69 72.90 75.11 2.21 2.94
Diesel 18.09 123.36 127.10 3.74 2.94
AGO 5.39 36.76 37.89 1.13 2.98
Atm-Res 40.09 273.37 281.70 8.33 2.96
Total 100 681.89 702.6 20.71 2.95
Vac ovhd 0.38 2.58 2.68 0.1 3.72
LVGO 3.71 25.27 26.21 0.94 3.58
HVGO 5.93 40.40 41.93 1.53 3.65
Vac res 30.08 205.12 212.80 7.68 3.61
Total 40.10 273.37 283.62 10.25 3.61
From the validation carried out as tabulated in Tables 4-7 and 4-8 that the CDU
and VDU model when implemented in GAMS presented yield with difference
ranging from 2.93% to 2.98% for CDU and 3.58% to 3.72% for VDU. Considering
the error ranges which are in percentage, the simplified linear Programming
models developed are good representation of the CDU and VDU with respect to the
cut fractions obtained from Aspen HYSYS.
72
4.6 Model for refinery production and product blending
subsystem planning in GAMS
The planning model in this section for refinery production and products blending
subsystem considered two types of balances (Al-Qatahni and Elkamel, 2010);
these are
 Fixed plant yield

 Unrestricted balances.
4.6.1 Mathematical model formulation
The material balance for the CDU yield:
Equation (4-19 to 4-22), indicates that the minimum flow into a process unit must
be satisfied so as to prevent a no flow condition.
Material balance for the VDU yield:
Material balance for the FCC yield:
The unrestricted balance occurred also in the HDT and therefore, the sum of the
gas oils are the feed.
73
4.6.2 Capacity constraint
4.6.3 Blending Constraint
The intermediates used for blending to produce final products are obtained by the
summation of the required or desired constituents, which is represented by the
following set of equation.
4.6.4 Market requirement Constraint
Constraints that defines the product requirements over the planning horizon. The
maximum production requirement constraint in bbl/day is:
4.6.5 Crude oil composition constraint
The sum of the percentage of each crude oil in the crude mix must be equal to 100.
= crude oil percentage in the crude mix.
4.6.6 Objective function
The objective function is formulated to maximise profit. It is determined as:
74
Total Revenue generated from final product minus Total cost of operations
and total cost of crude .
Total income
This is the total revenue generated from sales of the final products and it is
represented in Equation (4-28).
The cost of crude oil mix and the cost of intermediate materials like butane are the
input material cost.
Cost of operation
The cost of all the refinery units multiply by the level of the process makes up the
cost of the refinery operations.
4.6.7 Crude limit constraint
The amount of crude oil that is allowed to feed the CDU,
75
4.7 Refinery planning under different volumetric mixing ratios
53model runs in GAMS is then carried out on 50 different volumetric mixing ratio
runs of rigorous models generated from Aspen HYSYS.
The 53GAMS model runs is used to determine the optimal volumetric ratio that
generates the highest profit and the sulphur content based on some specified
constraint. Other variables like the CDU feed rate and the flow of intermediates
were also determined.
4.7.1 Solution strategy
 Up to 53 different volumetric mixing ratios are evaluated in Aspen HYSYS

e.g. 0/100, 2/98, 4/96, 6/94, 8/92, 10/90,...., 100/0
 Using the refinery planning model developed in section 4.6,53
mathematical programming planning runs are carried out in GAMS on
individual volumetric ratio
 Plot of the profits obtained from the 53 different volumetric mixing ratios
versus the individual volumetric ratios is then carried out.
4.7.2 Problem statement and implementation
Two crude vessels are expected to arrive with two different sets of crude: Ratawi
and Brent. The refinery has one CDU which has been designed to process a
capacity of not more than 100,000 barrel per day as shown in Table 4-9. The
refinery is capable of handling any blend of the two crudes at different
proportions. It is assumed that the product price and processing cost as shown in
Table 4-10and Table 4-11 are the same throughout the 30-day period. The
assumption made for the processing cost is due to unavailability of processing cost
data at the different volumetric mixing ratio.
In this problem there are eighteen flow rates (i.e. decision or optimization
variables), crude input and products output whose optimal values are required in
order to obtain the maximum profit. In this case study, the objective is to use the
76
developed model to solve the proposed petroleum refinery problem. The model
will be implemented with GAMS using CPLEX solver.
The model since its profit maximization treats the crude oil input and products
output as values for the model to determine rather than given.
 Crude flow rate into the CDU

 Quantities of individual intermediates and final products
 Optimal profit
 Which volumetric ratio gave the highest profit
A planning horizon of 30 days is used for this case study and each day is a time
interval. The objective of the optimisation problem is to achieve maximum profit
over the entire planning horizon given the type of crude oil and facilities needed in
the refinery to produce various products. The cost of purchasing the crude oil and
the cost of processing the crude oil are subtracted from the revenue or income
generated from the finished products as stated in equation 4-27.The objective
function and all constraints are linear.
4.7.4 Decision variables
The following decision variables have been used in the optimization problem
formulation: Flow rates of materials (crude and intermediate) to and from the
refinery and the quantities of final products.
4.7.5 Parameters used
Capacity of processing units and the Prices of final products are in Table 4-9 and
Table 4-10.
77
Table 4-9 Capacity of processing units in the Refinery
Processing Unit Capacity (bbl/day)
CDU 100 000
VDU 46500
FCC 16800
HDT 17140
Table 4-10 Product price for Final Products (EIA 2011)
Final Products Product price($/bbl)
Gasoline Blend (GB) 146.25
Kerosene 114.12
Fuel-oil 111.78
Fuel-gas 50.07
Diesel 147.24
Data on operating cost of the units, maximum production limit and product price
were obtained from literature as shown in Tables 4-11 and 4-12, while the crude
oil mix price is obtained from the volumetric ratios of the crude mix as shown in
Table 4-13which are then optimized for 30 days planning horizon.
Table 4-11 Process Units Operating Cost (Ajose, 2010)
Process Units Operating Cost ($/bbl produced)

CDU
0.59
VDU 0.72
HDT 0.19
FCC 0.41
78
Table 4-12 Production limit (Ajose, 2010)
Products Maximum Production

Requirement (kbbl)
Gasoline
61.144
Kerosene 20.79
Diesel 36.04
Fuel oil 55.37
Fuel Gas 6.31
Table 4-13 Crude oil and Butane data (EIA 2011)
Crude Oil cost ($/bbl)
RATAWI 88.58
BRENT 115.93
79
4.8 Results and analysis
Table 4-14 Summary of the profit generated from different volumetric mixing ratios
scenarios Profit $/30 scenarios Profit $/30 scenarios Profit $/30

Days Days Days
0_100 1 28483 36_64 20 59337 72_28 39 77816
2_98 2 35039 38_62 21 60820 74_26 40 78712
4_96 3 36682 40_60 22 61801 76_24 41 79495
6_94 4 37796 42_58 23 63707 78_22 42 80296
8_92 5 39434 44_56 24 64500 80_20 43 80081
10_90 6 40900 46_54 25 66232 82_18 44 80993
12_88 7 42019 48_52 26 67517 84_16 45 81757
14_86 8 43634 50_50 27 70908 86_14 46 81591
16_84 9 44954 52_48 28 70187 88_12 47 82768
18_82 10 46062 54_46 29 71707 90_10 48 83379
20_80 11 47650 56_44 30 72972 92_8 49 83531
22_78 12 48996 58_42 31 73624 94_6 50 83755
24_76 13 50805 60_40 32 75269 96_4 51 83983
25_75 14 50624 62_38 33 74756 98_2 52 84614
26_74 15 51674 64_36 34 75623 100_0 53 84168
28_72 16 52884 65_35 35 75986
30_70 17 54372 66_34 36 75777
32_68 18 55556 68_32 37 77022
34_66 19 57153 70_30 38 76751
80
Figure 4-4 Plot of Volumetric ratios versus Profits
Figure 4-4 is a plot of different profit obtained against volumetric ratios or

scenarios. From Figure 4-4, when only the Brent crude oil is used, the profit is as
low as $28483 for the 30 days planning horizon as shown in Table 4-14. But as
soon as the Ratawi crude is introduced into the Brent crude, the profit jumped to
$35039, the increase in profit continues not in a straight line but undulating as
shown in the same Figure. The plot shows increase in the profit as the volumetric
ratio with more sulphur content increases. This is due to the assumption that the
operating cost for all the scenarios or volumetric mixing ratios are the same. The
cost of the volumetric ratio with more sulphur is less than the cost of the
volumetric ratio with lower sulphur content.
In real refinery case data, the cost of processing a more sulphur rich crude may not
be the same with the cost of processing a less sulphur rich crude. The variation in
processing cost affects the refinery profit. Another aspect of cost that may affect
the profit is the blending cost as stated by Robertson et al. (2011). In their paper,
they analysed how blending cost that has been ignored by several authors affects
profit. The cost for processing has been assumed in this work the same because of
lack of real data.
81
From the results obtained, the highest profit of $84614 was obtained at volumetric
ratio of 98% of Ratawi and 2% of Brent crude oil and the sulphur content at this
volumetric ratio is 2.655 wt percent. The flow into the CDU is about
84,000bbl/day. The volumetric mixing ratio that gave the highest profit is used as
a base case for analysis.
Table 4-15 Summary of results obtained from the refinery planning model using Fixed yield
method
Products fractions Fixed Yield Method (98/2)

( Kbbl/day)
Naphtha 14.36
Kerosene 6.96
Diesel 12.99
AGO 4.23
ATM RES. 46.50
LVGO 4.48
HVGO 7.81
Vacuum Ovhd 0.34
FCC-gas 6.36
TGO 5.01
C2+Lighter 1.35
LPG 2.83
FINAL PRODUCTS (kbbl/day)
GASOLINE BLEND (GB) 20.72
FUEL-OIL 39.36
KEROSENE 6.96
DIESEL 12.99
FUEL-GAS 4.511
PROFIT ($/30 DAYS) 84614
82
Table 4-15 is a summary of the results obtained from the volumetric mixing ratio
that gave the highest profit. The volume of intermediates and final produced are
listed in the table. At the end of the 30 days planning horizon, the profit came out
to be $84614 from sale of the final products after every cost has been removed.
Conclusion can be drawn on the use of mixed crude for refinery processing, that
the cheaper the crude oil the more the profit.
4.9 Summary
Fixed yield was successfully implemented in planning refinery production and

product blending subsystem with the models developed from the information
generated from the previous Chapter 3 of this thesis.
 The information from the rigorous and empirical model was used to
develop a mathematical programming model for CDU, VDU, HDT and FCC
based on fixed yield method.
 The developed model is implemented in GAMS and validated.
 The validated mathematical programming model was then used to
implement a case study in a refinery planning model for production and
product blending subsystem.
 The proposed two crudes used in chapter 3 of this thesis was mixed in up to
50 different scenarios (volumetric mixing ratios) in Aspen HYSYS model
and the volume percent of the different yields obtained. The 53runs of the
rigorous model in Aspen HYSYS were then run up to 53 times again in a
mathematical programming model for planning in GAMS.
 It was discovered that the highest profit can be achieved in 98/2 volumetric
mixing ratio.
From the procedures, the process of obtaining the refinery profit in the various
volumetric mixing ratios is cumbersome i.e. the yields for individual products
fractions and for individual volumetric ratios are achieved 53 times in Aspen
HYSYS and 53 times in GAMS.
83
5 Refinery Production and Product Blending
Subsystem Planning with Aggregate Model
5.1 Introduction
This chapter is aimed at developing an aggregate model for CDU and VDU yield
prediction used for refinery planning of the production and product blending
subsystem. This is achieved by using the information obtained from the rigorous
modelling for the CDU and VDU combined with models (obtained from empirical
correlations) for FCC and HDT detailed in chapter3 and 4 solved with CPLEX
solver. 53 different volumetric ratios were generated from rigorous model. The
generated volumetric ratios were then used to derive regression model with
sulphur (x) as the independent variable and the cut fractions or yields (y) as the
dependent variables. The derived regression models were then used in place of the
regular fixed yield in the refinery production and products blending subsystem
planning model. An overview of the procedure is presented in Figure 5-1.
Figure 5-1 Procedures for Developing Aggregate Model
The same case study as in chapter 4 was then used to demonstrate the applicability
of the proposed method and solution approach.
85
5.2 Derivation of regression model for CDU and VDU with
information from rigorous model
5.2.1 Determination of factor that is used as independent variable
Linear regression is used in this work to estimate the effect of changing the
sulphur content in crude mix (X) which is an independent variable over yields (Y)
the dependent variable or to find the line that best predicts Y from X. This idea is
inspired by Robertson et al. (2011).
Some of the 53 different volumetric mixing ratios that were generated from
rigorous model and the properties at the different volumetric mixing ratio such as
sulphur, API and viscosity are shown in Table 5-1. Sulphur is one of the key
components of crude oil that determines if a crude oil is expensive (sweet) or not
(sour).
Table 5-1 Different volumetric mixing ratios and their properties
Volumetric mixing ratio Sulphur (wt. %) API Viscosity (cp)
100/0 3.88 24.50 118.80
75/25 2.96 25.54 54.92
60/40 2.44 29.67 32.49
50/50 2.11 28.52 23.39
40/60 1.78 29.85 15.78
25/75 1.00 33.45 5.59
0/100 0.43 38.50 4.98
Sulphur was used as the independent variable and key property of the crude
because it has a linear function to the yields generated in the different volumetric
ratios Robertson et al. (2011). From the regression plots, the deviated points are
not far from the straight line and the reliability for each yield fraction is around
0.9. However, the API and Viscosity from Table 5-1 are highly nonlinear i.e. the
86
lines are not straight, they are rather polynomial. This made the reliability R2far
less than 0.9 details are Appendix E.
5.2.2 The derived regression models for CDU and VDU
The results obtained from the 53runs in Aspen HYSYS on the yield fractions from
the different 53 scenarios or volumetric mixing ratios are shown in Figure 5-2 for
CDU and VDU.
87
Figure 5-2 Plots of predicted yields from CDU and VDU
88
The regression model derived for the various CDU and VDU are listed in Table 5-
2withthe R2 which is the confidence level or reliability of the function. The more
the R2is close to 1, the lesser the error, the more linear the relationship of Y from X.
These functions are then used as the coefficients for yield prediction in refinery
planning for production and products blending subsystem as shown in Table 5-3.
Table 5-2 Regression model derived for CDU and VDU yields
CDU and VDU product Yield Regression models R2

Naphtha ( -4.91 X+ 30.81) 0.9552
Kerosene ( -1.39 X + 12.10) 0.9521
Diesel (-1.54 X + 19.68) 0.9552
Atmospheric gas oil (AGO) (-0.25 X + 5.59) 0.7486
Atmospheric Residue (8.64 X+ 31.49) 0.9603
Low vacuum gas oil (LVGO) (0.17 X + 9.13) 0.8201
High vacuum gas oil (HVGO) (1.14 X + 13.66) 0.9477
Vacuum Residue (-0.89 X + 76.49) 0.9055
Vacuum Overhead (-0.14 X+ 1.08) 0.9483
Table 5-3 Regression model derived for CDU and VDU units yield
CDU and VDU Yields Regression model Coefficients for Regression

Naphtha ( -4.91 X+ 30.81)
Kerosene ( -1.39 X + 12.10)
Diesel (-1.54 X + 19.68)
AGO (-0.25 X + 5.59)
Atm Res. (8.64 X+ 31.49)
LVGO (0.17 X + 9.13)
HVGO (1.14 X + 13.66)
Vac Res ( -0.89 X + 76.49)
Vac Ovhd (-0.14 X+ 1.08)
89
5.2.3 Validation with GAMS model for CDU and VDU
Some selected volumetric mixing ratios of the crude mix obtained in Aspen HYSYS
were run in GAMS using CPLEX solver with the aim of validating the results. The
summary of the results were listed in Tables 5-4, 5-5 and 5-6.
Table 5-4 GAMS results for comparison with Aspen HYSYS model
Products 50/50 25/75 75/25 60/40 40/60

(Kbbl/day) (Kbbl/day) (Kbbl/day) (Kbbl/day) (Kbbl/
day)
Naphtha 22.84 25.73 19.71 21.95 24.07
Kerosene 9.87 10.69 8.95 9.43 10.20
Diesel 17.08 18.09 16.11 16.77 17.50
AGO 5.210 5.39 4.94 5.11 5.34
ATM RES. 44.00 40.09 46.50 46.50 42.88
LVGO 4.138 3.71 4.39 4.67 3.98
HVGO 6.842 5.92 7.40 7.36 7.82
Vac. Ovhd 0.319 0.38 0.37 0.39 0.38
Vac. Res. 32.70 30.08 34.34 33.70 29.85
Table 5-5 Comparison of Yields obtained from GAMS Model and results from AspenHYSYS in
50/50 crude mix.
PRODUCTS 50/50 GAMS Aspen HYSYS Difference

Vol.(Kbbl/day) Vol. (m3/h) Vol. (m3/h) %
Naphtha 22.84 155.749 156.60 0.5
Kerosene 9.87 67.3 67.65 0.4
Diesel 17.08 116.47 117.10 0.6
AGO 5.210 35.527 35.70 0.5
ATM RES. 44.00 300.04 308.60 2.5
LVGO 4.138 28.217 29.29 3.4
HVGO 6.842 46.656 48.43 3.7
Vacuum Ovhd 0.319 2.17 2.259 1.3
Vac. Residue 32.70 222.98 231.5 3.6
90
Table 5-6 Comparison of Yields obtained from GAMS Model and results from AspenHYSYS in
75/525crude mix.
Yields 7 5/25 GAMS AspenHYSYS Error
Vol. Vol. (m3/h) Vol. (m3/h) %

(Kbbl/day)
Naphtha 19.71 134.4 132.90 0.37
Kerosene 8.95 61.03 60.34 1.1
Diesel 16.11 109.85 108.60 1.1
AGO 4.94 33.68 33.34 1.0
ATM RES. 46.50 317.08 339.10 6.4
LVGO 4.39 30.00 32.38 7.3
HVGO 7.40 50.46 54.62 7.5
Vacuum Ovhd 0.37 2.523 2.711 7.4
Vac. Residue 34.34 234.16 253.4 7.5
From the results shown in Table 5-5 and 5-6, conclusion can the drawn that the
values obtained in Aspen HYSYS used to derive regression models in Table 5-1 are
good representation of product fractions obtained in CDU and VDU.
5.3 Derivation of regression model for FCC and HDT with

information from empirical correlation
During the rigorous modelling of the CDU and VDU in Aspen HYSYS, some of the
feed from the CDU and some from the VDU were sent to the FCC. From this
information, the inflow into the FCC is known. The feed outflow from the FCC is
then calculated using empirical correlations.
The method used for generating the regression model is shown in Appendix C.
Table 5-7 is the regression model generated for the FCC unit yield.
91
Table 5-7 Regression model for FCC
Yields Regression model

FCC gas (-0.85X + 41.77)
TGO ( 0.55 X +29.45)
Coke (0.0062 X+3.96)
C2+lighter (0.009 X+ 8.28)
LPG (0.21 X+16.59)
Before the feed enters the FCC, it went through the HDT. For the HDT, the 98%
method from Gary and Handwerk (2001) applies.
5.4 Aggregate model development
The detailed regression model derived from the four individual process units are
then used to form the aggregate model. The models will not be repeated in this
section.
5.5 Refinery production and product blending planning using

aggregate model
The regression models derived were used to replace the coefficients in a typical
fixed yield in a refinery planning model as obtained in Section 4-1 of 50/50
volumetric ratio. And the results compared.
Case study was performed to verify the efficacy of the model. The same case in
Chapter 4 is used in the chapter. The difference is that in the previous chapter, the
yield coefficients generated from Aspen HYSYS were used directly as fixed but in
this chapter, the regression models derived from the different volumetric ratios
and sulphur content in Table 5-3 were used instead.
The refinery planning model outlined in equation 4-19 to 4-33 is used for this case.
92
capacity of not more than 100,000 barrel per day. The refinery is capable of
handling any blend of the two crudes at different proportions. The model since its
profit maximization treats the crude oil input and products output as values for the
model to determine rather than given. The following are to be determined at the
end of the planning horizon:
 Crude input into the CDU

 Quantities of individual products
 Optimal profit
 Which volumetric ratio gave the highest profit
 The sulphur content at that volumetric ratio
interval. The objective of the optimisation problem is to achieve maximum profit
over the entire planning horizon given the type of crude oil and facilities needed in
the refinery to produce various products. The cost of purchasing the crude oil and
the cost of processing the crude oil is subtracted from the revenue or income
generated from the finished products. As stated in equation 4-27.The objective
function and all constraints are linear.
refinery and the quantities of final products.
The parameters used in this Chapter such as capacity of processing units and
product prices are the same as that used in Chapter 4.
93
Table 5-8 Summary of results obtained from the refinery planning model on Aggregate Model
Product Yields Aggregate Method (Vol. Kbbl)

Naphtha 15.18
Kerosene 7.16
Diesel 13.33
AGO 4.21
ATM RES. 46.5
LVGO 4.46
HVGO 7.76
Vacuum Ovhd 0.32
FCC-gas 6.36
TGO 4.97
C2+Lighter 1.34
LPG 2.78
GASOLINE BLEND (GB) 21.54
FUEL-OIL 39.44
KEROSENE 7.16
DIESEL 13.33
FUEL-GAS 4.44
PROFIT ($/30 DAYS) 87306
From the results obtained, the highest profit of $87306 was obtained at volumetric
ratio of 98% of Ratawi and 2% of Brent crude oil and the sulphur content at this
volumetric ratio is 2.655 wt percent. The flow rate of crude into the CDU is about
85,432bbl/day. Implementing the case study with aggregate model, the same
volumetric ratio that gave the highest profit in fixed yield method also gave the
highest profit in the aggregate model. However, the profit obtained using aggregate
model is $87306, and that obtained using fixed yield in Chapter 4 is $84614. The
94
difference in profit is as a result of the quantity of crude oil used in the CDU in both
methods. The more the crude used, the more the profit. The quantity of crude used
in the fixed yield method is 84,000bbl/day while that used in the aggregate
method is 85,432bbl/day.
Table 5-7 is a summary of the results obtained using aggregate model. The volume
of intermediates and final produced are listed in the table.
5.7 Comparison between fixed yield method and the

regression method
Direct comparison of fixed yield and regression method shows the following detail
in Table 5-9.
Table 5-9 Comparing Fixed yield and Aggregate method
Product Yields Fixed yield method Aggregate Method Relative Error

(Vol. Kbbl) (%)
(Vol. Kbbl)
Naphtha 14.36 15.18 5.4
Kerosene 6.96 7.16 2.7
Diesel 12.99 13.33 2.5
AGO 4.23 4.21 0.5
ATM RES. 46.50 46.5 0.0
LVGO 4.48 4.46 0.5
HVGO 7.81 7.76 0.6
Vacuum Ovhd 0.34 0.32 5.8
Vacuum Residue 34.35 34.46 0.3
FCC-gas 6.36 6.36 0.0
TGO 5.01 4.97 0.8
C2+Lighter 1.35 1.34 0.7
LPG 2.83 2.78 1.8
GASOLINE BLEND 20.72 21.54 3.8
(GB)
FUEL-OIL 39.36 39.44 0.2
KEROSENE 6.96 7.16 2.7
DIESEL 12.99 13.33 2.5
FUEL-GAS 4.51 4.44 1.6
PROFIT ($/30 84614 87306 3.0
DAYS)
For direct comparison of fixed yield method and aggregate method, the same case
on 98/2 volumetric ratio was used to represent the typical fixed yield and
95
improved method since both gave the highest profit. The results are summarized
in Table 5-9. The quantities of final products obtained in fixed yield are the same
all through the 30 day planning horizon also the quantities of final products
obtained in the aggregate method are the same throughout the 30 day planning
horizon also since both cases are deterministic.
The models developed in this chapter are evaluated by comparing the predictions
to fixed yield. Since it is evident that the profit $84614obtained for 30 day planning
horizon on fixed yield and $87306 for 30 day planning horizon on improved
method are close to each other also the flow of intermediates on both approaches
are very close and even in some cases like the Atm-res, vacuum residue e .t. care
the same so conclusion can then be drawn that he aggregate model can be used for
yield estimation or prediction in refinery planning.
The relative percent error between the predicted values from the proposed
method and the fixed yield values ranges from 0 to 5.8% as shown in Table 5-9.
5.8 Summary
In this chapter an aggregate model for refinery planning was proposed using
regression method. The model was successfully applied for yield prediction in a
refinery planning model.
It was discovered that blending of different crude oil changes the properties of the
crude. This is evident in Table 5-1. These properties affect the yields obtained from
the crude oil.
The procedures are:
 Using the information obtained in Chapter 3 by running 50 times in Aspen

HYSYS, regression models were derived for individual products fractions.
 The regression models derived were used in place of the regular fixed yield
approach in a mathematical programming model for planning of production
and products blending subsystem. The results obtained from the two
methods were compared.
96
The aggregate model provided a better approach for yield estimate; it is more
convenient than the regular fixed yield approach. The main differences between
the two approaches are:
 For the fixed yield method, the rigorous modelling was carried out 50 times
on different volumetric mixing ratios and the mathematical programming
was also carried out 50 times for individual volumetric mixing ratios before
being able to determine the volumetric ratio that gave the highest profit.
 For the aggregate model, the rigorous modelling was carried out 50 times,
the values were then used to generate linear functions which were then
used to carry out only one run of the mathematical programming to find the
highest profit.
 The predicted data by the aggregate model is close to that obtained by fixed
yield method. Conclusion can be drawn that regression model can be used
to predicts yields in refinery production and product blending subsystem
planning.
97
6 Planning for the Integrated Refinery
Subsystems under Deterministic Condition
6.1 Introduction
This chapter is aimed to deal with tactical planning for the integrated refinery
subsystems with MILP under deterministic condition. In this work, the three main
subsystems of a refinery are integrated under deterministic conditions. These
include crude unloading, production and product blending, and product
distribution subsystems. The profit is maximized considering the revenue from the
products, raw material costs, inventory costs, transportation costs, and operation
costs. A case study is used to demonstrate the applicability of the proposed model
and solution approach.
The refinery subsystems models for integration considered in this work include
the following:
 The modified scheduling model for crude unloading subsystem by Lee et al.
(1996)
 The aggregate model for production and product blending subsystem
developed in Chapter 5 of this Thesis.
 A planning model for product distribution by Alabi and Castro (2009)
The following assumptions are considered:
 Quantity of crude oil remaining in the pipe is neglected

 The mass balance on yields of individual process (i.e. CDU, FCC e.t.c.) units
is used to determine the product flow rates, which are also function of the
feed.
99
6.2 Planning model for the crude unloading subsystem
In this subsystem,
 The short term scheduling model presented in Lee et al. (1996) was
modified for planning purposes.
 The planning horizon is 30 days
 The bilinear term for component balance is not considered
Given the following parameters:
 Waiting time of each vessel in the sea

 Unloading duration time for each vessel
 Crude unloading rate from vessels to storage tanks
 Crude oil transfer rate from the mixing tank to CDU
 Inventory levels of storage tank ( Lee et al. 1996)
Figure 6-1 Flow of crude oil for Crude unloading subsystem
The following operating rules must be obeyed in this problem:
 Each vessel should turn up and depart from the docking station for crude oil
discharge within the planning horizon.
 A vessel can only unload its crude oil if it arrives at the dock station.
 If a vessel leaves the docking station, it cannot unload the crude oil.
100
 The vessel should leave the docking station after its arrival.
 The vessel cannot arrive at the docking station if the preceding vessel does
not leave.
 Crude oil cannot be fed into the mixing tank that is feeding the CDU.
 CDU receives one mixed crude at time interval.
6.2.1 Mathematical model formulation for crude unloading

subsystem
Vessel operations rule
A vessel arrival at the ducking station once within the horizon
Each vessel is allowed to leave once all through the horizon
Each unloading activity begins
Each unloading activity ends
Each vessels unloads as soon as it arrives as set by the planners
101
Unloading duration is thus,
The initial vessel must complete unloading its contents one time interval before
the ensuing vessel comes for unloading.
The possible time periods for unloading to occur is between time and
Flow in and out of storage tank cannot occur simultaneously
6.2.2 Material balance equations for the vessel:
Crude oil in vessel v at time t is equal to initial crude oil in the vessel v minus crude
oil transferred from vessel v to storage tank i up to time t.
102
Operating constraints on crude oil transfer rate from vessel v to storage tank i, at
time t.
In an event where there are more vessels than the available storage tanks, the
storage tanks are still used for the mixing operation.
The volume of crude oil transferred from vessels v to storage tank during the time
horizon equals the initial crude oil volume of vessel, v.
6.2.3 Material balance equations for the storage tanks:
The crude oil in storage tank i at time t equal to initial crude oil in storage tank i
plus crude oil transferred from vessels to storage tank i, up to time t minus crude
oil transferred from storage tank i, to charging tanks j up to time t:
Operating constraints on crude oil transfer rate from storage tank i to charging
tank j at time t.
Volume capacity limitations for storage tank i:
103
6.2.4 Material balance equations for charging tank:
Crude oil mix in charging tank j at time t equal to initial mixed oil in charging tank j
plus crude oil transferred from storage tanks to charging tanks j up to time t minus
crude oil mix j charged into CDU l up to time t:
Operating Constraints on mixed oil transfer rate from charging tank j to CDU l at
time t:
Volume capacity limitations for charging tank j at time t:
The flow rate (PC) of mixed crude entering the CDU l at time t, is the sum of the
individual crude oil entering the charging tank ( .
At this point, the yields obtained in the CDU are a function of the type of crude mix,
the property (sulphur in the crude mix) and flow rate of the crude mix.
104
6.2.5 Rules that must be followed for charging of crude oil during
operation
The charging tank j is allowed to feed at most one CDU l at any time interval.
CDU l can only receive charged crude oil from one charging tank j at any time
interval.
Changeover penalty: If CDU l is fed by crude oil mix j, at time t and later changed
from crude tank j to j’ changeover is incurred. This is defined by Lee et al. (1996).
Set up penalty: When vessels are allowed to stand for brine settling, in other to
start further unloading and transfers, cost is incurred. For this reason a set up
penalty has been included in the model. This penalty is for any period any tank is
allowed to stand.
This is included in the objective function.
105
Information in Table 6-1 is used for the crude unloading subsystem
Table 6-1 Data used for the 30 day horizon (Lee et al., 1996)
Planning Horizon (# of unit Time) 30days Period
Number of Vessels Arrivals 2
Arrival Time Amount of Crude Oil Sulphur Content
Vessel 1 Period 1 1,500,000/ bbl 0.0043
Vessel 2 Period 16 1,500,000 /bbl 0.0388
Number of Storage Tank 2
Storage Tanks Max Capacity Initial oil Amount
Tank 1 1,500,000/bbl 250,000 bbl
Tank 2 1,500,000/bbl 750,000 bbl
Number of Charging Tanks 2
Charging Tanks Max Capacity Initial oil Amount Sulphur content
Tank 1 1,500,000/ bbl 100,000/ bbl (0.0043)
Tank 2 1,500,000/ bbl 100,000/ bbl (0.0388)
Number of CDU 1
Maximum flow from vessel to storage 100,000bbl/period
Maximum flow from one storage tank to charging tanks 100,000bbl/ period
Cost involved in Vessel operation Unloading: US$8,000/ period
Sea waiting US$5,000/ period
Tank Inventory Cost Storage Tank: US$ 0.005/( period x bbl)
Charging Tank: US$ 0. 008/( period x bbl)
Crude Oil Demand 100,000bbl/ period
106
6.3 Planning model for the production and product blending
subsystem
The planning model for this subsystem is the planning model for mixed crude
developed in Chapter 5 of this thesis. The refinery topology used in that same
chapter applies to this also. The CDU, VDU, FCC and HDT process model in equation
4-19 to 4-31 is used. This relates to the properties of the feed stream as well as
operating variables. Normally, mass balances and yield expressions are used to
determine products flow rates.
6.4 Planning model for the products distribution subsystem
The model presented in Alabi and Castro (2009) is adopted. Final products from
the blending headers are stored in corresponding product tanks from where the
products are transported to different depots. In this subsystem, the following
constraints are applicable.
Mathematical model formulation for product distribution subsystem
Inventory cost of the product tank:
The total inventory cost of the product tanks at time t is equal to the sum of all the
volume of final product in the product tank multiply by the unit cost of inventory
of the product tank.
Product tank capacity limitations:
The volume of final products in the product tanks is less than maximum capacity of
the product tank and greater than the minimum capacity of the products tanks.
107
Flow Limitations
Flow of final product equal to the flow of final product to the depot at time t:
Flow of final product to the depot satisfies the demand at the depot
Material balance of final product in the product tanks at time t:
The volume of final product fp in the product tank pt at time t is equal to the initial
volume of product in the product tank pt plus flow of product to the depot minus
the demand of final product by the depot.
Transport cost of final product to the depot at time period t
The total transport cost of final product to the depot at time t is equal to the sum of
flow of final product to the depot multiply by the unit cost of transportation of the
final products to the depot.
108
Depot mass balance for final product at time period t
The quantity of final products in the depot is equal to the initial volume of final
product in the depot multiply by the sum of flow of final products into the depot
6.5 Planning model for the integrated refinery subsystems
In this section, the planning model developed for the individual subsystems are
being integrated and used in a refinery planning.
For the model to be feasible the production and product blending subsystem was
first solved and the requirement of the CDU was then passed to the crude oil
unloading subsystem.
The bilinear terms is avoided in this thesis.
Figure 6-2 Schematics for the Integrated Refinery Subsystems
109
Figure 6-3 Schematics for the Integrated Refinery Subsystems
6.6 Refinery planning for integrated subsystems
Because it is deterministic, the cost of crude oil and products demand do not vary
all through the planning horizon.
Material balance for crude mix
Capacity constraint
The capacity of the process units is at all times greater than the sum of all the feed
stream flows into it.
Material balance for intermediates
The coefficients of the intermediate fractions multiply by the level of the process
and the flow of intermediate materials that are purchased from the third party
110
(Butane) are greater than the total intermediate material that are blended into
final products.
Final products constraint
The flow rate of individual final products is equal to intermediate final products
that reach blending
Quality constraint
Quality attribute (i.e. specifications) for the products must be less than maximum
and greater than minimum specifications. These are represented in Equation (6-
40) and (6-41).
111
Crude mix cost and cost of all purchased intermediates
The cost of crude oil mix and the cost of intermediate materials e.g. butane form
the input material cost.
Cost of operation
The cost of all the refinery units multiply by the level of the process makes up the
cost of the refinery operations.
Total income
This is the total revenue generated from sales of the final products and its
represented in Equation (6-44).
capacity of not more than 100,000 barrel per day. The refinery is capable of
handling any blend of the two crudes at different proportions. In this case, the
product price, the process capacities are the same as in chapter 5 of this thesis. The
model since its profit maximization treats the crude oil input and products output
as values for the model to determine rather than given.
112
interval.
The objective function for the integrated refinery subsystem is to maximize the
profit of the refinery, defined as the sales revenue (REV), minus the cost of refining
(COP), the cost of feed (IMPCST), minus total operating cost which is unloading
cost (CUNLOAD), storage tank inventory cost (CINVST), sea waiting cost (CSEA),
transport cost (CTR), product tank inventory cost (CINVPT), changeover cost
(CSETUP) and penalty cost (PMD).
refinery, what is the volumetric mixing ratio that yielded the profit and the
quantities of final products.
Same as Chapter 5 for the production and product blending subsystem, and
Information for crude unloading is in Table 6-1.
113
Table 6-2 Final products quality specification (AMD Refinery data)
Octane VP (mmHg) Density Sulphur

kg/m3 (wt .
fraction)
GB >86 0.002
GB <90 9.0 0.002
Kerosene 0.0015
Diesel 0.005
Fuel-gas 860 0.005
fuel-oil 1100 0.04
The case study was implemented in GAMS and optimized with CPLEX solver. The
optimal profit of $75, 954 was obtained in 1.25 second after 1364 iterations. This
contains 6,367 variables, 9,625 equations and 397 discrete variables as shown in
Table 6-3. Figure 6-3 shows the summary of the amount of final products
produced. Vessel 1 arrived on the first day and finished unloading on day 15, while
vessel 2 arrived on day 16 and finished unloading on day 29.
Table 6-3 Summary of results for the integrated refinery subsystem
Planning horizon 30
Equations 9,625
Variables 6,397
Discrete variables 397
Objective Function$/30days 75954
The profit for the integrated subsystem $75,954came down compared to $87,306
obtained from the production and products blending subsystem, because the cost
items such as unloading, inventory etc. Cost in the crude unloading subsystem and
the cost in the product distribution subsystem were considered realistically. The
total quantity of crude purchased is 3,000,000bbl for the planning horizon. The
refinery made use of the cheaper crude more than the expensive crude.
114
1,720,000bbl of crude type 1 (Ratawi) was used compared to 1,280,000bbl of
crude type 2 (Brent) used during the entire planning horizon. The reason is that
Ratawi crude is cheaper crude than Brent. The volumetric mixing ratio of 57% of
Ratawi and 43% of Brent was used for the refinery operation to be profitable.
For the refinery operation to be profitable, more of the cheaper crude was used
than the more expensive one.
Though assumption was made that the processing cost of the crude no matter the
volumetric mix has the same cost of operation, this is not always the case in real
refinery process.
The CDU feed flow rate increased to100,000bbl/day compared to the feed rate of
non- integrated subsystem (85,432bbl/day), this is because the CDU is trying to
maximise the crude consumption for profit to be maximised. The CDU maximum
capacity is 100,000bbl/day.
Figure 6-3 Flow rate of Final Products
From Figure 6-3, the production of gasoline blend is more followed by the
production of fuel-oil. The production of Fuel-gas has the least product flow rate.
The quantity of gasoline blend increased due to the increased quantity of crude
type two compared to the quantity produced in the non-integrated subsystems.
115
The reason why the refinery made use of the stated volumetric ratio 57% and 43%
of the two crudes in this integrated subsystem compared with the 98% and 2%
obtained in the non-integrated subsystem are:
 The raw material cost and the products price in the two models are the
same, and there are more cost functions to be considered e.g. cost of
inventories and unloading, sea waiting costs and other constraints, so the
system has to optimally operate to maximise profit.
 From the flow of final products for the non-integrated subsystem in Table
5-8, the quantity of fuel-oil produced is more than the quantity of gasoline
blend produced because there are more of the cheaper crude than the
expensive crude and less constraints. But in the integrated subsystem, the
reverse is the case, this is because, the system observed that there are more
constraints and also from the market requirement, there is a chance of
producing more gasoline blend which has a higher market value than fuel-
oil and this will boost the refinery profit. To produce the gasoline blend, the
refinery requires appreciable quantity of the expensive crude; this made the
use of crude type two or the expensive crude to increase from 2% to 43%.
Also the quantity of other products e.g. kerosene, diesel increased in the
integrated subsystem.
The use of this volumetric mixing ratio affected the capacities of the process
units e.g. CDU, VDU and FCC. The CDU was maximally utilised while the VDU
because of the higher quantity of crude type two, it brought down the capacity
utilization from 46.5 to 31.67kbbl/day.
116
6.7.1 Variation in Crude oil Price
Figure 6-4 Refinery Profit and Variation in crude Price
Figure 6-4 shows how variation in crude price could affect the profit of a refinery.
Ratawi crude was used as an example in this case. The Figure shows that if there is
an increase or decrease in crude oil price, it will reflect in the profit of the refinery.
This also explains the effect of crude price on refinery margin as stated in the
motivation of this thesis. It is confirmed that crude price is up to 80% of the
refinery profit margin is true. This is in agreement with Arofonosky et al (1978).
The variation in price of ratawi crude only is stated in this thesis because both
crude types showed the same trend.
117
6.7.2 Variation in Final Products Price
Figure 6-5 Variation in Final product Price
Figure 6-5 shows a variation in the final products price by 10%. 10% increase in
products prices increased the final products flow rate and this also increases the
profit margin of the refinery from $75,954to $114,223. Increase in products prices
happens when there is scarcity of a particular product and also it could happen
when there is increase in raw material cost.
6.8 Summary
In this chapter, the three model for the three basic refinery subsystems were
joined together as one to form an integrated model and its then used for refinery
planning. In the integrated refinery case study result, it was discovered that the
system made use of more crude, but gave a lesser profit compared with the non-
integrated systems. This is because there are more cost function from the crude
unloading and products distribution subsystems taking from the production and
products blending subsystem. Variation in crude oil price and variation in final
product price were also carried out and the results obtained.
118
7 Conclusions and Recommendations for Future
Work
7.1 Conclusions
This work studies the tactical planning for the integrated refinery subsystems
using MILP as the optimization technique.
Recent advances in the modelling of process units for refinery planning and
scheduling were reviewed, revealing the current gaps in knowledge. In particular it
was observed that the trend has been the use of mixed crude for refinery planning
and that crude mix has not yet being reported using the proposed method. There is
no existing model on planning of the integrated subsystems of the refinery.
7.1.1 Rigorous modelling of CDU and VDU
The CDU and VDU were rigorously modelled in Aspen HYSYS and the yields
obtained from the CDU were validated with data from literature. The CDU alone
was validated because there is no case in the literature to validate the VDU yield
results with. Process analyses were further carried out on some of the scenarios or
volumetric mixing ratios. It was observed that the yields from the CDU match with
what is obtained in the literature.
7.1.2 Fixed yield method for refinery production and product

blending subsystem planning
The information from the rigorous modelling on CDU and VDU were then
transferred into simplified CDU and VDU model for planning. The FCC and HDT
units that were derived by correlations were also transferred into simplified FCC
and HDT model for planning.
About 53 runs for different scenarios or volumetric mixing ratios were carried out
in the rigorous model developed. The simplified model for the process units were
then used to run 53 mathematical model for the individual volumetric mixing
119
ratios using fixed yield method with the aim of determining the volumetric mixing
ratio that has the highest profit and the sulphur content at that volumetric mixing
ratio. At the end, the volumetric mixing ratio that has the highest profit was
determined to be blending ratio of 98% of Ratawi crude and 2% of Brent crude
and the profit obtained was $84614 at 2.655 wt. percent of sulphur.
7.1.3 Aggregate model for refinery production and product blending

subsystem planning
An aggregate model for the production and products blending subsystem of a

refinery was developed using regression method and used on the same case study
as in the fixed approach. For the aggregate model to be derived, information was
transferred from the rigorous model to mathematical programming model and
used as fixed yield approach and later regressed linearly and used. However, in
using the aggregate model developed in determining the volumetric mixing ratio
that yields the highest profit, a single run of the mathematical programming model
was carried out compared to the 53 runs in regular fixed yield counterpart. The
model developed serves as an alternative approach to calculating the yield in the
process units for refinery planning. The profit obtained in aggregate method is
$87306 and at the same volumetric ratio and sulphur content as the fixed yield
method.
7.1.4 Planning for the integrated refinery subsystems
Three models were integrated together to develop a refinery planning model for
the three refinery subsystems which are the crude unloading subsystem, the
production and products blending subsystem and the products distribution
subsystem.
The aggregate model developed for the production and products blending
subsystem, the modified Lee et al.(1996) model for scheduling the crude unloading
subsystem and the product distribution model developed by Alabi and Castro
(2009) were integrated into a refinery planning model and a case study
implemented. After solving the case study, it was discovered that the profit
120
obtained reduced to $75,954 and at a volumetric ratio of 57% of Ratawi and 43%
of Brent.
7.2 Recommendations for future work
7.2.1 Consideration of more than two pre-mixed crude oils
For the aggregate model developed, two crude oils were mixed together, there is
also the need to consider the mix of more than two different crude oils and their
properties like the API and the viscosity incorporated in the regression technique
and solves using nonlinear regression model.
7.2.2 Validation of CDU model
In this work the CDU model was validated alone due to lack of data for benchmark,
there is need for the VDU to be validated as well. Also more detailed modelling of
HDT and FCC should be carried out
7.2.3 Consideration of uncertainty
This model assumed certainty in the yield from the various processes, which may
not always the case. On this note uncertainty in the yield should be carried out.
Planning of the supply chain under uncertainty is very crucial due to the changing
market conditions and the existence of lead times i.e. difference between the time
an order is put and the delivery time in supply chain, the variables for production
need to be determined prior to the realisation of demand. There is need to
incorporate uncertainty in the integrated model.
7.2.4 Real life case study for Refinery Planning
Some of the data used for the case study were assumed; therefore the model
should be used on a complete and real refinery data to confirm the efficacy. In
Chapter 4 and 5 of this Thesis, assumption was made that the operating cost for all
the volumetric mixing ratios are the same which in real life situation, more sulphur
121
in the crude oil, may increases the operating cost. Some real life data should be
used to implement the idea.
Most of the refinery operations has a range of sulphur specification that the CDU
has been designed to handle, the regression would have to fall within the
specification range. The recent CDU’s are designed to be robust and is able to
handle wider range of crude oil types. In this case, the regression model would be
wide and should fall within range.
In this work sulphur only was considered, it will be good if other crude properties
like viscosity API etc. are considered when developing the regression model.
122
References
1. Ajose, E.B., (2010), Refinery Planning with Linear Programming, MSc.

Thesis, School of Engineering, Cranfield University, UK.
2. Alabi, A. and Castro, J. (2009), Dantzig-Wolfe and Block Coordinate-descent
Decomposition in Large-Scale integrated Refinery Planning, Computers and
Operations Research, vol. 36, pp. 2472-2483.
3. Alonso-Ayuso, A., Escudero, L. F. and Ortuno, M. T. (2005), Modeling
Production Planning and Scheduling Under Uncertainty, in Stein, W. and
William, T. Z. (eds.) Application of stochastic programming, siam, University
city science center, Philadlephia. P.A, pp. 217-252.
4. Al-Qahtani, K.Y., and Elkamel, A. (2010), Planning and Integration of

Refinery and Petrochemical Operations, WILEY-VCH Verlag GmbH & Co.
KGaA, Weinheim.
5. Alattas, A.M., Grossmann, I.E., and Palou-Rivera, I. (2011), Integrating Non-
linear CDU Models in Refinery Planning Optimization, Ind. Eng. Chem. Res.,
vol. 50, pp. 6860-6870.
6. Aronofsky, S., Dutton,J. M., and Tayyabkhan, M. T. (1978), Managerial
Planning with Linear Programming in Process Industry Operations, Wiley
and Sons, New York.
7. ASPEN HYSYS, ASPEN HYSYS, V7.2. Aspen Technology, Inc.: Reading.
8. Basak, K ., Abhilash K. S., Saibal, G., and Saraf D. N. (2002),On-Line

Optimization of a Crude Distillation Unit with Constraints on Product
Properties, Ind. Eng. Chem. Res., 41 (6), pp 1557–1568.
9. Bassett, M. N., Pekney, J. F., and Reklaitis, G. V. (1996), Decomposition
Techniques for the Solution of Large Scale Scheduling Problems. AICHE
journal, 42, 3373.
10. Briesen, H. and Marquardt, W. (2004), New Approach to Refinery Process
Simulation with Adaptive Composition Representation. Process Systems
Engineering, vol. 50, Nos. 3, pp.633-645.
123
11. Brooke, A., David, K., Alexander, M., and Ramesh, R. (1997), GAMS User
Guide. WASHINGTON DC: GAMS Development Corporation.
12. Brooks, R.W. and van Walsem, F.D (1999), Choosing Cut Points to Optimize
Product Yields. Hydrocarbon Processing.
13. Bruce, A., McCarl, Paul, V. D. E., Meeraus, A., Micheal, B., Steven, D. and Pete,
S. (2007), A McCarl Expanded General Algebraic Modeling Systems (GAMS),
User's Guide, GAMS Development Corporation, Washington.
14. Chufu, L., Xiaorong, H., Bingzhen, C., Qiang, X. and Chaowei, L. (2008), A
Hybrid Programming Model for Optimal Production Planning under
Uncertainty in Refinery, Chinese Journal of Chemical Engineering, vol. 16,
no. 2, pp. 241-246.
15. Dan, W. and Marianthi, I. (2007), Hierarchical Approach for Production
Planning and Scheduling Under Uncertainty., Chemical Engineering and
Processing, vol. 46, pp. 1129-1140.
16. Edgar, T F., Himmelblau, D.M. and Lasdon, L.S. (2001), Optimization of
Chemical Processes, second edition, McGraw-Hill Co. Inc., New York.
17. Energy Information Administration (EIA), Monthly Energy Review,
Washington: National Energy Information Centre, July-August, 2011
18. Energy Information Administration (EIA), Weekly Petroleum Status Report,
Washington: National Energy Information Centre, July –August,2011.
19. Enrique, C., Antonio, J. C., Pablo, P., Ricardo, G. and Natalia, A. (2002),
Building and solving mathematical programming models in Engineering
and science, Wiley and Sons, NY.
20. Floudas, C. A., and Lin, X. (2005), Mixed Integer Linear Programming in
Process Scheduling: Modeling, Algorithms, and Application, Annals of
Operations Research, vol. 139, pp. 131-162.
21. Gary, J. H. and Handwerk, G. E. (2001), Petroleum Refining Technology and
Economics, fourth ed., Marcel Dekker Inc, New York.
22. Gary, J. H. and Handwerk, G. E. (1975), Petroleum Refining Technology and
Economics, Marcel Dekker, New York.
124
23. Geddes, R.L.A., (1958), General Index for Fractional Distillation Power for
Hydrocarbon Mixtures, AIChE J., vol.4, pp.389-392.
24. Gilbert, R., Leather, J. and Ellis, J.,(1966), The Application of the Geddes
Fractionation Index to the Crude Distillation Units. AIChE Journal, vol.12,
pp:432-437.
25. Guyonnet, P., Grant H. F., and Bagajewicz J.M. (2009), Integrated Model for
Refinery Planning, Oil Procuring, and Product Distribution. Ind. Eng. Chem.
Res., 48, 463-482.
26. Harjunkoski I. and Grossmann I. E. (2001), A Decomposition Approach for
the Scheduling of a Steel Plant production, Comp. and Chem. Eng., 25,pp.
1647-1660.
27. Hartmann, J.C.M. (1997). Decision-making and modeling in petroleum
refining. Hydrocarbon processing,76(11),pp77-81.
28. Hess, F. E. (1977), Solution of Equations for Fractionating Columns (PhD
thesis), Texas A & M University college station, Texas A & M University
college station.
29. Holland, C. D., and Liapis,A. I. (1983),Computer Method for Solving Dynamic
Separation Problems, Mc Graw-Hill Book Company, New York, NY.
30. Huang, H., B. E. (2000), Simulation and Control of Complex Distillation
Processes (PhD thesis), Texas Tech University, Texas Tech University.
31. Katzer, J., Ramage, M., and Sapre, A. (2000), Petroleum Refining: Poised for
Profound Changes. Chem. Eng. Process, pp 41-51.
32. Jeongho, P., Sunwon, P., Choamun, Y. and Young, K. (2010), Integrated
Model for Financial Risk Management in Refinery Planning,Ind. Eng. Chem.
Res., vol. 49, pp. 374-380.
33. Jia, Z. and Marianthi, I. (2003), Mixed Integer Linear Programming Model
for Gasoline Blending and Distribution Scheduling, Ind. Eng. Chem. Res., vol.
42, pp. 825-835.
34. Jia, Z. and Marianthi, I. (2004), Efficient Short-Term Scheduling of Refinery
Operations based on a continuous time formulation, Comp. and Chem. Eng.,
vol. 28, pp. 1001-1019.
125
35. Joly, M., Moro, L. F. L. and Pinto, J. M. (2002), Planning and Scheduling For
Petroleum Refineries Using Mathematical Programming, Brazilian journal
of Chemical Engineering, vol. 19 (2), pp. 207-228.
36. Jones, D. S. J., Pujado, P.R., (2006), Handbook of Petroleum Processing,
Springer, Netherlands.
37. Kallrath, J. and Wilson, J. M., (1997), Business Optimization Using
Mathematical Programming, Macmillan Press Ltd, London.
38. Kondili, E., Pantelides, C. C. and Sargent, R.W.H. (1993), AGeneral Algorithm
for Short-Term Scheduling of Batch Operations-1. MILP formulation. Comp.
and Chem. Eng., 17, 211-227.
39. Kong, Ming-Teck (2002), Downstream Oil products Supply Chain
Optimization, (PhD Thesis, Dept. of Chem. Engineering. And Chem.
Technology. Imperial College London).
40. Khor, C. S., Elkamel, A. and Douglas, P. L. (2008), Stochastic Refinery
Planning with Risk Management, Petroleum Science and Technology, vol.
26, no. 14, pp. 1726-1740.
41. Lee, H., Pinto, J. M., Grossman, I. E. and Park, S. (1996), Mixed Integer Linear
Programming Model for Refinery Short-Term Scheduling of crude oil
Unloading with Inventory Management, Ind. Chem. Eng. Res., vol. 35, pp.
1630-1641.
42. Lee, W. and Tayyabkhan, M. T. (1970), Optimize Process Runs- Get more
profit, Hydrocarbon Process vol. 49, pp. 286-290.
43. Li, W., (2004), Modeling of oil Refinery For Production Planning, Scheduling
and Economic Analysis (PhD thesis), Hong Kong University of Science and
Technology.
44. Li, W., Chu Wai, H., Pu, L. and An Xue, L. (2004), Refinery Planning Under
Uncertainty, Ind. Eng. Chem. Res.,vol. 43, pp. 6742-6755.
45. Li, W., Hui, C. W., Hua, B. and Tong, Z. (2002), Scheduling Crude Oil
Unloading, Storage and Processing. Ind. Eng. Chem. Res,vol. 41, pp. 6723-
6734.
46. Li, W., Hui, C. and Li, A. (2005), Integrating CDU, FCC and Product blending
model into Refinery Planning,Comp. and Chem. Eng., vol. 29, pp. 2010-2028.
126
47. Li, X., (2000), Refinery-Wide Optimization,PhD thesis, Texas Tech
University.
48. Lin, T.D.V., (1993), FCC Advanced Control and Optimization, Hydrocarbon
Processing, vol.72 (4), 107-114.
49. Magalhaes, Marcus Vinicius de Oliveira. (2004), Crude oil scheduling model
as MINLP using both continuous and discrete time formulation. (PhD
Thesis, Dept. of Chem. Engr. Imperial College London).
50. Maravelias, C.T. and Grossmann, I.E., (2006), On The Relation of Continuous
Time and Discrete- time state task Network Formulation. AICHE Journal, 52,
843-849.
51. Maravelias, C.T. and Sung, C. (2009), Integration of Production Planning and
Scheduling: Overview, Challenges and Opportunities.Comp. and Chem.
Eng.,vol. 33, pp. 1919-1930.
52. Moro, L. F. L., Zannin, A. C. and Pinto, J. M. (1998), A Planning Model For
Refinery Diesel Production, Comp. and Chem. Eng., vol. 22, pp. 1039-1042.
53. Moro, L. F. L. and Pinto, J., M (2004), Mixed Integer Programming Approach
for Short Term Crude Oil Scheduling, Ind. Eng. Chem. Res., vol. 43, pp. 85-94.
54. Moro, L.F.L., (2000), Tecnicas de Otimizacao Mista Inteira para o
Planejamento e Proramacao de Prodducao em Refinarias de Petroleo. (PhD
Thesis. Universidade Sao Paulo).
55. Muge, E., and Grossmann, I. E., (2007), Simultaneous Planning and
Scheduling of Single-Stage Multi-product continuous plants with Parallel
lines, Comp. and Chem. Eng., vol. 32, pp.2664-2683.
56. Neiro, M. S. S., and Pinto, J. M. (2005), Multiperiod Optimization for
Production Planning of Petroleum Refineries,Chem. Eng. Communications,
vol. 192, pp. 62-88.
57. Neiro, M. S. S. and Pinto, J. M. (2004), A General Modeling Framework for
the Operational Planning of Petroleum Supply Chains, Comp. and Chem.
Eng., vol. 28, pp. 871-896.
58. Packie, J. W., (1941), Distillation Equipment in the oil Refining Industry,
AIChE Transaction, vol. 37, pp. 51-78.
127
59. Pelham, R. and Pharris, C., (1996), Refinery Operations and Control: A
Future Vision, Journal on Hydrocarbon Processing, vol. 75, pp. 89-94.
60. Pinto, J. M., Joly, M. and Moro, L. F. L. (2000), Planning and Scheduling
Models for Refinery Operations, Comp. and Chem. Eng., vol. 24, pp. 2259-
2276.
61. Pinto, J. M. and Moro, L. F. L. (2000), A Planning Model for Petroleum
Refineries, Brazilian Journal of Chemical Engineering, vol. 17 (4-7), pp. 575-
586.
62. Pongsakdi, A., Pramoch, R., Kitipat, S. and Bagajewicz, M. J. (2006), Financial
Risk Management in the Planning of Refinery Operations, International
Journal of Production Economics, vol. 103, pp. 64-86.
63. Pitty, S. S., Li, W., Adhitya, A., Srinivasan, R., Karimi, I. A., (2008), Decision
Support for Integrated Refinery Supply Chains –Part 1. Dynamic simulation.
Comp. and Chem. Eng., vol. 32: pp. 2767-2786.
64. Reddy, P. C. P., Karimi, I. A., Srinivasan, R.,(2004), Novel solution approach
for optimizing crude oil operations, AIChE journal. Vol.50 (6) pp. 1177-
1197.
65. Rejowsky, R. Jr., Pinto, J.M., (2003),Scheduling of a Multi-Product Pipe Line

System. Comp. Chem. Eng, vol. 27, pp. 1229-1246.
66. Robertson, G., Palazoglu, A., Romagnoli, J.A., (2011), A Multi-Level

Simulation Approach for the Crude oil Loading/Unloading Scheduling
Problem. Comp. And Chem Eng. Vol. 35, pp. 817-827.
67. Russell, R.A., (1983), A Flexible and Reliable Method solves Single-Tower
and Crude –Distillation –Column Problems. Chem. Eng. vol. 90(21): pp:53-
59.
68. Seider, D., Seider W. D. J., and Daniel, L. R., (1998), Process Design
Principles, John Wiley, New York.
69. Shah, N. K., Saharidis, K. D. G., Jia, Z. and Ierapetritou. M. G., (2009),
Centralized-decentralized Optimization for Refinery Scheduling, Comp. and
Chem. Eng., vol. 33, pp. 2091-2105.
128
70. Shah, N.K, Li, Z., and Ierapetritou, M.G. (2011), Petroleum Refining
operations: Key Issues, advance, and Opportunities. Ind. and Eng. Chem.
Res., 50(3), pp. 1161-1170.
71. Shah, N. (1996), Mathematical Programming Techniques for Crude Oil
Scheduling, Comp. and Chem. Eng., vol. 20, pp. 1227-1232.
72. Shaoping, L., Luoyong, D., Benxian, S., Lijuan, Z., Feng, T., Xinru, X.,
Jingyi, Y., and Beilei, Z., (2010), The Distillation Yield and Properties of
Ternary Crude Oils BlendingPetroleum Science and Technology Vol., 29(3),
pp. 271-281
73. Shapiro, A. and Homem-de-Mello, T. (1998), A Simulation-Based Approach
to Two -Stage Stochastic Programming with Recourse, Mathematical
Programming, vol. 81, pp. 301-325.
74. Sim, J.W and Daubert T.E., (1980), Prediction of Vapor –Liquid Equilibria of
Undefined Mixtures, Ind. Eng. Chem. Process Des. Dev. , vol. 19, pp 386-393.
75. Simon, J.D. and Azma, H.M., (1983), Exxon Experience With Large Scale
Linear and Nonlinear Programming Applications. Comp. and Chem. Eng.,
vol. 7(5), pp. 605-614.
76. Slaback, D. D., (2004), A Surrogate Model Approach to Refinery-Wide
Optimization (PhD thesis), Texas Tech University, Texas Tech University.
77. Stratiev, D., Dinkov, R., Petkov, K., Stanulov, K., (2010) , Evaluation of Crude
Oil Quality, Petroleum and Coal. Vol. 52, (1) pp. 35-43.
78. Tung-Hsiung, K. and Chuei-Tin, C. (2008), Application of a mathematical
programming model for integrated planning and scheduling of petroleum
networks,Ind. Eng. Chem. Res., vol. 47, pp. 1935-1954.
79. Trierwiler, D. and Tan, R. L. (2001), Advances in Crude Oil
Modeling,Hydrocarbon Asia, vol. 11, pp. 52-58.
80. Trzupek,Richard., (2002), Air Quality Compliance and Permitting Manual,

McGraw-Hill USA.
81. Wagner, H. M., (1975), Principles of operations research, Prentice Hall Inc.
129
82. Watkins, R. N., (1979), Petroleum Refinery Distillation, second ed., Gulf
Publishing Company, Houston.
83. Wu, N., Qian, Y., Zhou, M. and Chu, F., (2006), Issues on Short Term
Scheduling of Oil Refinery, Systems Management and Cybernetics, Taiwan,
IEEE, pp. 2920-2925.
84. Zhang, N. and Zhu, X. X., (2006), Novel Modeling and Decomposition
Strategy for total site Optimization, Comp. and Chem. Eng., vol. 30, pp. 765-
777.
85. Notes: Electronic source: http://wn.com/Standard_Oil_Company (Assessed
Dec. 2011)
86. Notes:Electronic
source:http://www.hydrocarbonstechnology.com/projects/corytonrefiner
y/coryton-refinery1.html (Assessed Dec. 2011)
source:http://www.buildinggreen.com/live/index.cfm/2010/12/21/Under
standing-Petroleum (Assessed Dec. 2011)
source:http://www.pacificenergypier400.com/index2.php?id=61
(Assessed Dec. 2011).
89. Notes: Electronic source:http://www.hydrocarbonprocessing.com

(Assessed Dec. 2011).
130
Appendices
Appendix A (Ratawi and Brent Crude Assay)
The crude assays were used to characterise the crude for rigorous modelling. The
crude assay was entered in the oil environment of Aspen HYSYS for the individual
crude. These assays will give the simulation model the estimated volume percent
of the liquid to be distilled at a given temperature. The rigorous modelling was
carried out in the Chapter 3 of this Thesis.
Complete crude oil assay of Ratawi crude oil
Page 131
Complete crude oil assay of Brent crude
Page 132
Appendix B (Crude oil cost)
The cost of the crudes were obtained from EIA
Table B- 1 Week Brent FOB in Dollars per Barrel (EIA, 2011)
Page 133
Table B- 2 Week Ratawi crude oil FOB in Dollars per Barrel (EIA, 2011)
Page 134
Appendix C (Charts used for Empirical correlations)
The charts in appendix C were used for empirical correlations for FCC yields. The
Figure 1 was used to determine the conversion rate of the feed from FCC. The other
charts were then used to calculate the composition of the intermediate products
from the FCC unit. Figure 11 is an example of how the products from the FCC unit
were calculated using 50/50 volumetric mixing ratio.
Figure 1Effect of feed composition on conversion at constant operating conditions
Page 135
Figure 2 Catalytic Cracking yields, Zeolite catalyst (coke)
Page 136
Figure 3 Catalytic Cracking yields, Zeolite catalyst (Fuel gas)
Figure 4 Catalytic Cracking yields, Zeolite catalyst (C3 and C4)
Page 137
Figure 5 Catalytic Cracking yields, Zeolite catalyst (C3 ratios)
Page 138
Figure 6 Catalytic Cracking yields, Zeolite catalyst (C4 ratios)
Page 139
Figure 7 Catalytic Cracking yields, Zeolite catalyst (C5 + gasoline)
Page 140
Figure 8 Catalytic Cracking yields, Zeolite catalyst (Heavy gas oil, feed K=11.8)
Page 141
Figure 9 Catalytic Cracking yields, Zeolite catalyst (Heavy gas oil, feed K=12.35)
Page 142
Figure 10 FCC product gravity Zeolite catalysts
Page 143
%vol. for 68% wt Vol.
50/50 Density conversion Wt % (kg/hr) (m3/hr)
density
to FCC 886.2 L.gas 413.3 - 5.1 5007.739 12.11647
volume 110.8 Lpg 553 0.22 13.72828 13479.93 24.376
Fcc
mass 98190.96 gas 622.604 - 37.60138 36921.15 59.30118
Coke 880 - 5.2 5105.93 5.802193
TGO 850.095 0.4 38.37035 37676.21 44.32
100 98190.96 145.9158
Figure 11Procedure for calculating the values for yields for FCC units
Page 144
Appendix D (Graphs of Yields versus API)
Page 145
Page 146
Page 147
Page 148
Page 149
Appendix E (Graphs of Yields versus Viscosity)
Page 150
Page 151
Appendix F (Process flow sheet)
Page 152
Appendix G Conversion from ASTM to TBP (Li 2000)
Polynomial Regression Method
A Ci Naphtha Kerosene Diesel AGO Atm Res

0 - - - -89.99134827 - -89.991348
89.99134827 89.99134827 89.99134827 89.99134827
1 2.74798583 875.783084 1227.52527 1739.749829 2016.746801 4215.41026
2 -0.01068849 - - -4284.113639 - -25151.684
1085.626616 2132.790807 5756.919155
3 3.17355E-05 1027.285315 2828.738145 8053.076598 12544.56345 114556.825
4 -4.87231E- - - -7827.526108 - -269796.75
08 502.6474974 1939.988166 14134.57718
5 3.72814E-11 122.5752578 663.0896673 3791.875467 7937.375688 316679.023
6 -1.11569E- - - -718.4189184 - -145377.11
14 11.69058543 88.64200972 1743.272203
SUM 335.6876096 467.9407512 664.6518797 773.9260539 -4964.2767
Appendix H (Details of Rigorous runs in AspenHYSYS)
Page 153
RAT 88.58
BRENT 115.93
%RAT 0 2 4 6 8 10 12 14 16 18 20 22 24 25 26
% BRENT 100 98 96 94 92 90 88 86 84 82 80 78 76 75 74
SCENARIOS 0_100 2_98 4_96 6_94 8_92 10_90 12_88 14_86 16_84 18_82 20_80 22_78 24_76 25_75 26_74
NAPHTHA 25.29 28.42 28.24 27.95 27.77 27.58 27.28 27.1 26.91 26.62 26.33 26.06 25.86 25.77 25.59
KERO 10.28 11.45 11.36 11.26 11.27 11.17 11.08 11.08 10.98 10.89 10.79 10.79 10.68 10.71 10.61
DIESEL 17.04 18.92 18.93 18.74 18.75 18.75 18.57 18.57 18.48 18.39 18.22 18.23 18.09 18.12 18.13
AGO 4.87 5.47 5.48 5.48 5.38 5.37 5.38 5.39 5.39 5.39 5.38 5.41 5.68 5.39 5.3
ATM-RES 42.28 35.64 35.94 36.52 36.84 37.16 37.73 37.94 38.31 38.81 39.45 39.74 40.11 40.09 40.56
OFF GAS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
VAC-OVHD 0.95 1.04 1.03 1.02 1.01 1 0.99 0.98 0.97 0.96 0.95 0.94 0.932 0.95 0.92
LVGO 8.57 9.16 9.16 9.17 9.27 9.27 9.28 9.28 9.29 9.29 9.49 9.28 9.3 9.24 9.31
HVGO 13.78 13.92 14.03 14.23 14.24 14.34 14.45 14.45 14.56 14.66 14.98 14.65 14.78 14.88 14.89
VAC-RES 77.26 76.34 76.28 76.12 76 75.9 75.8 75.84 75.77 75.67 75.27 75.63 75.5 75.54 75.53
L.GAS 8.224387 8.28202 8.282187 8.282774 8.282858 8.282942 8.283446 8.283446 8.283867 8.307918 8.285894 8.286148 8.287166 8.286487 8.287166
LPG 15.2755 16.69007 16.69418 16.70859 16.71065 16.71271 16.72508 16.72508 16.73541 16.80494 16.78515 16.79139 16.81639 16.79971 16.81639
FCC-GAS 45.48242 41.47497 41.46333 41.42251 41.41668 41.41084 41.37578 41.37578 41.34653 40.3543 41.20561 41.18794 41.11711 41.16435 41.11711
COKE 3.938397 3.965995 3.966075 3.966356 3.966396 3.966437 3.966678 3.966678 3.96688 3.978397 3.96785 3.967972 3.96846 3.968134 3.96846
TGO 27.0793 29.58694 29.59423 29.61977 29.62342 29.62707 29.64901 29.64901 29.66731 30.55444 29.75549 29.76655 29.81087 29.78131 29.81087
SULFUR 0.43 0.455 0.502 0.553 0.602 0.652 0.7 0.749 0.799 0.848 0.897 0.946 0.982 1.021 1.046
API 38.5 34.96 34.81 34.64 34.48 34.32 34.16 34 33.9 33.86 33.77 33.66 33.61 33.45 33.38
VISCOUSITY 4.976 4.073 4.152 4.239 4.322 4.409 4.492 4.634 4.802 4.924 5.077 5.267 5.397 5.59 5.693
CRUDE COST 115.93 115.383 114.836 114.289 113.742 113.195 112.648 112.101 111.554 111.007 110.46 109.913 109.366 109.0925 108.819
PROFIT 28483 35039 36682 37796 39434 40900 42019 43634 44954 46062 47650 48996 50805 50624 51674
Page 154
%RAT 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56
% BRENT 72 70 68 66 64 62 60 58 56 54 52 50 48 46 44
SCENARIOS 28_72 30_70 32_68 34_66 36_64 38_62 40_60 42_58 44_56 46_54 48_52 50_50 52_48 54_46 56_44
NAPHTHA 25.39 25.21 24.92 24.74 24.53 24.34 24.19 23.96 23.78 23.5 23.21 22.84 22.74 22.55 22.36
KERO 10.61 10.52 10.52 10.43 10.36 10.26 10.27 10.17 10.08 9.98 9.98 9.87 9.79 9.69 9.7
DIESEL 17.93 17.88 17.94 17.75 17.81 17.72 17.03 17.53 17.44 17.35 17.35 17.08 17.17 17.18 16.99
AGO 5.3 5.29 5.31 5.31 5.23 5.23 5.13 5.24 5.1 5.25 5.25 5.21 5.25 5.16 5.16
ATM-RES 40.97 41.29 41.54 42.03 42.61 42.97 43.87 43.7 44.07 44.57 44.87 45.01 45.73 46.11 46.5
OFF GAS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
VAC-OVHD 0.92 0.91 0.93 0.89 0.89 0.88 0.89 0.87 0.86 0.85 0.85 0.75 0.83 0.83 0.82
LVGO 9.31 9.26 9.33 9.33 9.36 9.36 9.36 9.38 9.38 9.39 9.49 9.40 9.39 9.41 9.41
HVGO 14.8 14.78 14.91 15.02 15.16 15.27 15.38 15.49 15.6 15.61 15.72 15.08 15.84 15.75 15.85
VAC-RES 75.61 75.78 75.36 75.49 75.34 75.3 75.24 75.08 74.9 74.99 74.77 74.82 74.83 74.93 74.84
L.GAS 8.287677 8.287166 8.288188 8.288871 8.31499 8.292823 8.27695 8.292996 8.293515 8.317858 8.319301 8.280766 8.320841 8.321477 8.322114
LPG 16.82892 16.81639 16.84147 16.85822 16.97331 16.95524 16.56563 16.95948 16.97222 17.0416 17.07595 16.6593 17.1126 17.12774 17.14291
FCC-GAS 41.08162 41.11711 41.04607 40.9986 39.86937 40.72376 41.8275 40.71174 40.67564 39.67266 39.57371 41.56215 39.46814 39.42454 39.38086
COKE 3.968704 3.96846 3.968949 3.969276 3.981783 3.971168 3.963567 3.971251 3.9715 3.983157 3.983848 3.965395 3.984585 3.98489 3.985195
TGO 29.83308 29.81087 29.85532 29.88503 30.86056 30.05701 29.36635 30.06453 30.08712 30.98473 31.04719 29.53239 31.11383 31.14135 31.16892
SULFUR 1.093 1.142 1.189 1.237 1.284 1.331 1.787 1.425 1.472 1.518 1.52 2.112 1.614 1.661 1.707
API 33.25 33.12 32.99 32.86 32.77 32.66 29.85 32.43 32.3 32.21 32.32 30.98 32.12 32.02 31.93
VISCOUSITY 5.921 6.377 6.821 7.265 7.861 8.178 15.78 9.002 9.443 9.876 9.941 23.48 10.68 11.03 11.33
CRUDE COST 108.272 107.725 107.178 106.631 106.084 105.537 104.99 104.443 103.896 103.349 102.802 102.255 101.708 101.161 100.614
PROFIT 52884 54372 55556 57153 59337 60820 61801 63707 64500 66232 67517 70908 70187 71707 72972
Page 155
%RAT 58 60 62 64 65 66 68 70 72 74 76 78 80 82 84
% BRENT 42 40 38 36 35 34 32 30 28 26 24 22 20 18 16
SCENARIOS 58_42 60_40 62_38 64_36 65_35 66_34 68_32 70_30 72_28 74_26 76_24 78_22 80_20 82_18 84_16
NAPHTHA 22.06 21.84 21.57 21.29 21.18 21.07 20.89 20.46 20.32 20.13 19.83 19.66 19.23 19.14 18.87
KERO 9.6 9.51 9.51 9.41 9.41 9.3 9.32 9.22 9.12 9.12 9.03 8.93 8.83 8.83 8.76
DIESEL 16.99 16.9 16.8 16.83 16.62 16.61 16.62 16.44 16.33 16.35 16.26 16.21 16.06 16.07 16.01
AGO 5.16 5.16 5.06 5.07 5.07 5.06 5.07 4.99 5.07 5.08 5.08 5.08 5.09 4.98 5
ATM-RES 46.91 47.42 47.8 48.17 48.48 48.67 48.88 49.74 49.99 50.14 50.65 50.96 51.62 51.81 52.39
OFF GAS 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
VAC-OVHD 0.81 0.83 0.8 0.79 0.79 0.78 0.78 0.77 0.77 0.77 0.76 0.76 0.75 0.75 0.74
LVGO 9.41 9.41 9.42 9.42 9.43 9.43 9.44 9.55 9.54 9.55 9.55 9.55 9.56 9.57 9.49
HVGO 15.85 15.85 15.87 15.88 15.88 15.89 15.99 16.02 16.02 16.13 16.14 16.24 15.94 16.06 16.21
VAC-RES 74.83 74.91 74.81 74.84 74.87 74.86 74.74 74.48 74.66 74.56 74.54 74.48 74.74 74.65 74.6
L.GAS 8.322296 8.296381 8.323665 8.324123 8.290927 8.32504 8.325132 8.302906 8.302639 8.327158 8.29196 8.29196 8.293169 8.306303 8.30928
LPG 17.14725 16.5303 17.17985 17.19074 16.9087 17.21258 17.21476 17.20272 17.19617 17.263 16.93405 16.93405 16.96372 17.2861 17.35916
FCC-GAS 39.36837 41.14537 39.27446 39.24308 40.8556 39.1802 39.1739 40.02264 40.04119 39.03496 40.78378 40.78378 40.69972 39.78644 39.57946
COKE 3.985282 3.972872 3.985938 3.986157 3.97026 3.986596 3.98664 3.975997 3.975869 3.98761 3.970755 3.970755 3.971334 3.977624 3.979049
TGO 31.17681 30.05508 31.23609 31.2559 29.97451 31.29559 31.29957 30.49573 30.48413 31.38727 30.01945 30.01945 30.07206 30.64353 30.77305
SULFUR 1.754 2.471 1.85 1.898 1.921 1.944 1.99 2.04 2.116 2.126 2.171 2.216 2.26 2.304 2.344
API 31.8 27.36 31.58 31.42 31.36 31.31 31.2 31.08 30.77 30.89 30.8 30.7 30.61 30.52 30.53
VISCOUSITY 11.64 32.26 12.31 12.75 12.99 13.23 13.69 14.16 15.03 15.06 15.43 15.77 16.08 16.35 16.46
CRUDE COST 100.067 99.52 98.973 98.426 98.1525 97.879 97.332 96.785 96.238 95.691 95.144 94.597 94.05 93.503 92.956
PROFIT 73624 75269 74756 75623 75986 75777 77022 76751 77816 78712 79495 80296 80081 80993 81757
Page 156
%RAT 86 88 90 92 94 96 98 100
% BRENT 14 12 10 8 6 4 2 0
SCENARIOS 86_14 88_12 90_10 92_8 94_6 96_4 98_2 100_0

NAPHTHA 18.62 18.43 18.13 17.93 17.64 17.33 17.05 15.04
KERO 8.66 8.56 8.46 8.47 8.37 8.27 8.26 10.82
DIESEL 15.91 15.82 15.82 15.63 15.53 15.43 15.43 13.69
AGO 4.59 5.01 4.9 4.91 5.01 5.02 5.02 4.27
ATM-RES 53.29 53.27 53.77 54.16 54.56 55.04 55.22 57.79
OFF GAS 0 0 0 0 0 0 0 0
VAC-OVHD 0.73 0.73 0.722 0.72 0.73 0.73 0.73 0.73
LVGO 9.5 9.5 9.51 9.61 9.62 9.62 9.63 9.12
HVGO 16.33 16.45 16.55 16.65 16.67 16.67 16.79 16.83
VAC-RES 74.47 74.37 74.24 74.06 73.91 73.94 73.88 75.34
L.GAS 8.310732 8.310823 8.294295 8.312099 8.312464 8.31338 8.313563 8.221178
LPG 17.3948 17.39704 16.99137 17.42836 17.43733 17.45979 17.46429 15.19671
FCC-GAS 39.47847 39.47215 40.62139 39.38342 39.35801 39.29437 39.28163 45.70562
COKE 3.979744 3.979788 3.971873 3.980399 3.980574 3.981012 3.9811 3.93686
TGO 30.83624 30.8402 30.12107 30.89572 30.91162 30.95144 30.95942 26.93963
SULFUR 2.388 2.438 2.481 2.525 2.568 2.611 2.655 3.88
API 30.44 30.36 30.27 30.18 30.09 29.99 29.89 24.5
Page 157
VISCOUSITY 16.67 16.82 17.04 17.33 17.55 17.73 17.89 118.8
CRUDE COST 92.409 91.862 91.315 90.768 90.221 89.674 89.127 88.58
PROFIT 81591 82768 83379 83531 83755 83983 84614 84168
Page 158
Appendix I Publications
Journal Paper
 Edith Ejikeme-Ugwu, Songsong Liu, Lazaros Papageorgiouand Meihong Wang(2012),

Development of an Aggregate Model for Refinery Production Planning.(In Preparation)
Peer reviewed conferences
 Ejikeme-Ugwu, E. And Wang, M. (2012) Aggregate Model for Refinery Production Planning In:
Ian David Lockhart Bogle and Michael Fairweather (Editors). (eds.) 22nd European
Symposium on Computer Aided Process Engineering (ESCAPE-22). Computer Aided Chemical
Engineering, 17 -20 June 2012, London (Accepted).
 Ejikeme-Ugwu, E., Liu, S., and Wang, M. (2011) Integrated refinery planning under demand
uncertainty. In: Pistikopoulos, E.N., Georgiadis, M.C., and Kokossis, A.C. (eds.) 21st European
Symposium on Computer Aided Process Engineering (ESCAPE-21). Computer Aided Chemical
Engineering, vol. 29. Amsterdam, Elsevier. pp. 950–954.
Presentations
 IChemE CAPE Subject Group PhD Poster Day, University of Leeds, UK, 12 May 2010.
 Ejikeme-Ugwu, E. And Wang, M. (2012) Planning for the Integrated Refinery Subsystems,
9thInternational Conference on Computational Management Science (CMS), 18-20 April
2012Imperial College London (Accepted).
Page 159

Planning For The Integrated Refinery Subsystems PDF

Uploaded by

Copyright:

Available Formats

Planning For The Integrated Refinery Subsystems PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Planning For The Integrated Refinery Subsystems PDF

Uploaded by

Copyright:

Available Formats

Cranfield University

PLANNING FOR THE INTEGRATED REFINERY SUBSYSTEMS

PLANNING FOR THE INTEGRATED REFINERY SUBSYSTEMS

Supervisor: Dr. Meihong Wang

This thesis is submitted in partial fulfilment of the

© Cranfield University, 2012. All rights reserved. No part of

Keywords: Refinery planning, linear programming, optimization, linear regression,

I hereby extend my innermost gratitude to Dr. Meihong Wang my supervisor.

My sincere appreciation goes to Dr Lazaros Papageorgiou and Dr Songsong Liu at

I would like to extend my appreciation to Dr ‘Keluo Chukwuogo, the Managing and

My personal gratitude is extended to my colleagues, Chet, Mubarak, Alhaji Aminu,

1.1.3 Product distribution subsystem .......................................................................... 7

2.3.2 Modelling of process units for refinery planning ........................................24

2.3.3 Current methods for CDU modelling in refinery planning ......................27

2.4.2 Modelling for refinery scheduling .....................................................................35

2.5 Integration of planning and scheduling ...................................................................38

Table 1-1 Difference between planning and scheduling ................................................................ 9

Atmospheric Gas Oil

Crude Distillation Unit

Vacuum Distillation Unit

Fluid Catalytic Cracking

American Petroleum Institute

True Boiling Point

American Society for Testing and Materials

Mixed integer linear programming

Mixed integer non-linear programming

End Boiling Point

Initial Boiling Point

Energy Information Administration

General Algebraic Modelling Systems

High Vacuum Gas Oil

Low Vacuum Gas Oil

Liquefied Petroleum Gas

Normal Boiling Point

Process Industry Modelling Systems

Refinery and Petrochemical Modelling Systems

: Volume of crude oil in storage tank i at time t

: Storage tank i maximum capacity.

Volume of crude oil in charging tank j at time t

: Storage tank minimum capacity.

: Charging tank maximum capacity.

: Charging tank minimum capacity

: Unloading cost for vessel v

: Charging tank changeover cost

: Minimum concentration of component k in the blended crude oil of

: Maximum concentration of component k in storage tank i.

Maximum crude oil rate from vessel v to one storage tank i.

Volumetric flow rate from charging tank j to CDU l at time t

Maximum crude oil rate from charging tank j to CDU l

Minimum crude oil rate from charging tank j to CDU l

Volume of crude oil in vessel v at time t

Binary variable to denote if storage tank i is receiving crude from

Time for vessel v to finish unloading

Vessel v arrival time

Vessel v departure time

Binary variable to denote storage tank I holding at time t

Penalty for holding tank i

Total optimal operational cost

Parameter for demand violation for mixed crude j