Escolar Documentos
Profissional Documentos
Cultura Documentos
Outline
I.
Supply Chain
Traditional View:
Logistics in the Economy
Freight transportation
Inventory expense
Administrative expense
1990
$352
$221
$ 27
1996
$455 billion
$311 billion
$ 31 billion
11%
10.5% GNP
Traditional View:
Logistics in the Manufacturing Firm
Profit: 4%
Profit
Logistics
Cost
Marketing
Cost
Manufacturing
Cost
The grocery industry could save $30 billion (10% of operating cost by
using effective logistics and supply chain strategies
Laura Ashley turns its inventory 10 times a year, five times faster than 3
years ago
Manufacturer
Distributor
Retailer
Customer
Supplier
Manufacturer
Distributor
Retailer
Customer
Supplier
Manufacturer
Distributor
Retailer
Customer
Retailer
Replenishment Cycle
Distributor
Manufacturing Cycle
Manufacturer
Procurement Cycle
Supplier
10
11
Replenishment cycle
Manufacturing cycle
13
Procurement cycle
14
PUSH PROCESSES
Customer
Order arrives
Customer
Order
Cycle
PULL PROCESSES
16
New
Marketing
Product
and
Development Sales
Distribution
Service
17
Implied
Demand
Uncertainty
Regular Demand
Uncertainty due to
customers demand and
Implied Demand
Uncertainty due to
uncertainty in
Supply Chain
18
High Fashion
Emergency steel
Customer Need
Low
Responsiveness
High
Low
Low
10%
1-2%
0%
High
High
40-100%
10-40%
10-25%
Frequent breakdowns
Unpredictable and/or low yields
Poor quality
Limited supplier capacity
Inflexible supply capacity
Evolving production processes
AUTOMOTIVES
Low
High
Low
Cost (efficient)
21
High Cost
of it
e
n ic F
o
Z eg
t
ra
t
S
Responsiveness
spectrum
Low Cost
Efficient
supply chain
Certain
demand
Implied
uncertainty
spectrum
Uncertain
demand
22
Responsive
Primary goal
respond quickly
Product design
strategy
Pricing strategy
maximize performance
at minimum cost
Lower margins
create modularity
Manufacturing
strategy
Inventory strategy
Supplier strategy
higher margins
maintain flexibility
maintain buffer inventory
aggressively reduce
select based on speed,
flexibility, reliability and
quality
23
of it
e
n ic F
o
Z eg
t
ra
t
S
Responsiveness
spectrum
Efficient
supply chain
Product
Maturity
Implied
uncertainty
spectrum
Product
Introduction
24
Strategic Scope
Suppliers Manufacturer Distributor
Retailer
Customer
Competitive
Strategy
Product Dev.
Strategy
Supply Chain
Strategy
Marketing
Strategy
25
Inventory
Transportation
Facilities
Information
Drivers
TRADE OFF FOR EACH DRIVER
26
Inventory
Transportation
28
Facilities
29
Information
Components
30
Efficiency
Responsiveness
Inventory
Cost of holding
Availability
Transportation
Consolidation
Speed
Facilities
Consolidation /
Proximity /
Dedicated
Flexibility
What information is best suited for
each objective
Information
31
Globalization
32
Lead Time
s
u
C
iz
m
to
High
n
o
i
at
Long
Short
Mass
Customization
Low
Low
Co
st
High
34
35
II.
Designing the supply chain network
36
FACILITY DECISIONS:
Network Design Decisions
Facility role
Facility location
Capacity allocation
Strategic
Technological
Macroeconomic
Competition
Logistics and facility costs
38
Local FG
Mix
Regional FG
Local WIP
Central FG
Cost
Central WIP
Central Raw Material and Custom production
Custom production with raw material at suppliers
High
Low
Response Time
High
39
Inventory costs
Transportation costs
40
Facilities
Costs
Number of Facilities
41
Costs
Transportation
Frequent inbound
Number of facilities
42
Total Costs
Percent Service
Level Within
Promised Time
Facilities
Inventory
Transportation
Labor
Number of Facilities
43
COMPETITIVE strategy
Choose location
Competitive STRATEGY
INTERNAL CONSTRAINTS
Capital, growth strategy,
existing network
PRODUCTION TECHNOLOGIES
Cost, Scale/Scope impact, support
required, flexibility
COMPETITIVE
ENVIRONMENT
PHASE I
Supply Chain
Strategy
PHASE II
Regional Facility
Configuration
REGIONAL DEMAND
Size, growth, homogeneity,
local specifications
POLITICAL, EXCHANGE
RATE AND DEMAND RISK
PHASE III
Desirable Sites
PRODUCTION METHODS
Skill needs, response time
FACTOR COSTS
Labor, materials, site specific
PHASE IV
Location Choices
AVAILABLE
INFRASTRUCTURE
LOGISTICS COSTS
Transport, inventory, coordination
45
Manufacturer
Retailer
Customers
46
Factories
Retailer
Customers
47
Factories
Warehouse storage by
distributor/retailer
Customers
48
Distributor/retailer
warehouse
Customers
49
Retailer
Cross Dock DC
Pickup sites
Customers
50
Local DC
Cross-Dock
Store 1
Customer 1
DC
Local DC
Cross-Dock
Customer 2
DC
Regional
Finished
Goods DC
Local DC
Cross-Dock
Store 1
Store 2
Store 2
Store 3
Store 3
51
Speculative Strategy
Hedging Strategy
Single sourcing
Match revenue and cost exposure
Flexible Strategy
Key Costs:
Fixed facility cost
Transportation cost
Production cost
Inventory cost
Coordination cost
yi
xij
Max Profit = f i y i + ci , j xi , j
i =1
Dj
= Dj
K i yi
i =1
m
j =1
i =1 j =1
i, j
i, j
j = 1,...m
Ki
yi {0,1}
i = 1,...n
i = 1,...n
53
d
x, y
xn, yn
dn
fn
Dn
( x x n) + ( y y n)
2
Warehouse Coordinates
Coordinates of delivery location n
Distance to delivery location n
Cost per ton mile to delivery location n
Quantity to be shipped
54
Min C = ci , j xi , j
xij
i =1 j =1
s.t.
x
i =1
i, j
= Dj
j = 1,...m
x
j =1
i, j
Ki
xi , j 0
i = 1,...n
i = 1,...n; j = 1,...m
Cij
n
m
Dj
Ki
55
i =1
e =1
Min C = f i yi + f e ye + ch ,i xh ,i + ci ,e xi ,e + ce , j xe , j
Fixed costs
plants
warehouse
h =1 i =1
i =1 e =1
e =1 j =1
Shipping costs
Supply source to plant
Plant to warehouse
Warehouse to market
56
Supplier capacity
x
i =1
h ,i
Balance supply-plant
x
h =1
h ,i
Supplier capacity
x
e =1
i ,e
Balance plant-warehouse
x
i =1
i ,e
Warehouse capacity
x
j =1
e, j
Demand satisfaction
x
e =1
e, j
Sh
t
h = 1,...l
xi ,e 0
i = 1,...n
K i yi
i = 1,...n
e =1
xe , j 0
e = 1,...t
We ye
e = 1,...t
= Dj
j = 1,...m
j =1
57
III.
Supply chain management of flexible
process networks - Lagrangean-based
decomposition techniques
Peter Chen
Jose M. Pinto
Content
Motivation
Decomposition techniques considered
Problem Statement
CPFN Model
Structure
Objective and constraints
Observation
Decomposition Applied
Lagrangean relaxation
Lagrangean/surrogate relaxation
Strategies Proposed
Results
Conclusion
Future Work
Motivation
Technological barriers
Problem Statement
Processes I1~I4
I1 is dedicated
Chemicals J1~J6
Sites C1 & C2
Markets L1~L4
CFPN Structure
J1
J2
J3
J4
J5
J6
I1.K1
I2.K1
I4.K1
I2.K2
I4.K2
I3.K1
I3.K2
I3.K3
L1
L2
I3.K4
Site 1
I2.K1
I4.K1
I2.K2
I4.K2
L4
I3.K1
I3.K2
I3.K3
L3
Site 2
jJ lL cC tT
jlt Pjlct
jJ lL cC tT
iI jJM ik kK i cC tT
Wijkct
ikct
jcV jct
jJ cC tT
Changeover cost
iI kK i k 'Ki cC tT
jlt SF jlt
jJ cC tT
dlctYPdlct
d D lL cC tT
jct F jct
jJ lL tT
Transportation cost
List of Assumptions
Assumptions
Mass balance of raw materials and byproducts are proportional to the main
product of the process and respective production scheme.
The operating cost of a process is proportional to the amount of main
product produced.
Changeover only implies in cost and the overall time spent is negligible.
Only one delivery of chemicals from one market over c time interval is
allowed.
Demand is given by a range of values, having an upper and a lower bounds
i I j , j JOikc , j ' JM ik , k Ki , c C , t T
i I j , j JI ikc , j ' JM ik , k K i , c C , t T
i I j , j JM ik , k K i , c C , t T
i I j , k Ki , k ' Ki , c C, t T
ikct
kK i
=1
i I j , k Ki , c C, t T
lL
YP
d D
dlct
U
Pjlct
c 'C {c }
jc 't
j J , c C, t T
j J , l L, c C , t T
iI j kKi
lL
d D, l L, c C , t T
P
cC
jlct
a Ujlt
j J , l L, t T
jlct
d Ujlt
j J , l L, t T
cC
j J , l L, t T
cC
Bounds
U
V jct V jct
Z ikk 'ct 1
F jct , Pjlct , S jlct , SF jlt , V jct , Wijkct , Z ikk 'ct 0
Relaxation
Lagrangean relaxation
Easier to solve
Lagrangean/surrogate relaxation
Solving relaxed
problems
Terminating
criteria
No
Updating multipliers
Subgradient optimization
Accelerating convergence
Updating the
multipliers
Yes
Returning
Solutions
Lagrangean relaxation
Z (u ) = max cx + u (b ( Ax + By ))
Cx + Dy e
General MIP
Z = max cx + dy
Ax + By b
Cx + Dy e
x 0m
y {0,1}n
Lagrangean/surrogate relaxation
x 0m
Z u (t ) = max cx + t u (b ( Ax + By ))
y {0,1}n
Cx + Dy e
x 0m
y {0,1} .
n
Subgradient optimization
u k +1 = u k + t k g k
tk =
l k (w L(u k ))
gk
1 lk 2 2 ( 1 , 2 > 0 )
k +1
= u + tk d
k
1 L L( u k )
tk =
2
k
k d
d k = g k + k d k 1 ;
L = r L0 + (1 r ) Lc ;
r = 00.6933(r r )
e
3.26
d k 1 g k
k 1 2
k =
d
0
if r r2
otherwise
If d k 1 g k 0,
Otherwise
Observation
i I j , k Ki , k ' Ki , c C, t T
lL
lL
iI j kK i
{F}
jc 't
j J , c C, t T
c 'C c
d D, l L, c C , t T
j J , l L, t T
cC
Decomposition applied
C
V jctB V jct
C
and V jctB V jct
SF jltB = SF jltC
SF jltB SF jltC
A
B
Yikct
= Yikct
A
B
Yikct
Yikct
A
B
= YPdlct
YPdlct
A
B
YPdlct
YPdlct
and
A
B
YPdlct
YPdlct
B
C
YPdlct
= YPdlct
B
C
YPdlct
YPdlct
and
B
C
YPdlct
YPdlct
SF jltB SF jltC
and
and
A
B
Yikct
Yikct
Lagrangean relaxation
A
B
A
B
B
C
SF jltB SF jltC , Yikct
Yikct
, YPdlct
YPdlct
, Yp dlct
YPdlct
A
B
A
B
B
C
SF jltB SF jltC , Yikct
Yikct
, YPdlct
YPdlct
, Yp dlct
YPdlct
A
jcV jctB ikk 'c Z ikk 'ct jlt SF jltB dlctYPdlct
tT jJ cC
tT iI kKi k 'Ki cC
tT dD lL cC
tT jJ cC
Y
A
B
C
(Yikct
)
jct F jct + uVjct (V jct
V jctB ) + uikct
Yikct
tT jJ lL
tT jJ cC
tT iI kKi cC
YP 2
B
C
YP1
B
A
C
B
(
SF
SF
+ u SF
jlt
jlt
jlt ) + udlct (YPdlct YPdlct ) + udlct (YPdlct YPdlct )
tT jJ lL
tT dD lL cC
tT dD lL cC
jJ lL cC
iI jJM ik kKi cC
ikct
jJ cC
A
ikk 'c Z ikk 'ct jlt SF jltB dlctYPdlct
jct F jct
iI kKi k 'Ki cC
jJ cC
dD lL cC
jJ lL
C
Y
A
B
C
B
+ uVjct (V jct
V jctB ) + uikct
(Yikct
Yikct
) + u SF
jlt ( SF jlt SF jlt )
jJ cC
iI kKi cC
jJ lL
YP1
B
A
YP 2
B
C
+ udlct
(YPdlct
YPdlct
) + udlct
(YPdlct
YPdlct
)
dD lL cC
dD lL cC
The total profit is equal to the summation over the time periods
t
Z CFPN LR = Z CFPN
LR
tT
List of constraints
Wijkct = ijkcWij 'kct
i I j , j JI ikc , j ' JM ik , k K i , c C
i I j , j JOikc , j ' JM ik , k K i , c C
i I j , j JM ik , k K i , c C
B
Yikct
+ YikC'c ,t +1 1 Z ikk 'ct
kK i
B
ikct
=1
i I j , k Ki , k ' K i , c C
i I j , k Ki , c C
C
U
Pjlct YPdlct
Pjlct
lL
lL
j J , l L, c C
d D
C
B
A
YPdlc
,t 2 + YPdlc ,t 1 + YPdlct 1
d D , l L, c C
iI j kKi
c 'C {c }
jc 't
j J,c C
S
cC
jlct
a Ujlt
j J,l L
jlct
d Ujlt
j J,l L
j J,l L
cC
U
C
U
V jctB V jct
, V jct
V jct
,
Z ikk 'ct 1
C
F jct , Pjlct , S jlct , SF jltB , SF jltC ,V jctB ,V jct
, Wijkct , Z ikk 'ct 0
A
B
A
B
C
Yikct
, Yikct
, YPdlct
, YPdlct
, YPdlct
{0,1}
C
A
B
A
B
B
C
V jctB V jct
, SF jltB SF jltC , Yikct
Yikct
, YPdlct
YPdlct
, Yp dlct
YPdlct
Lagrangean/surrogate relaxation
jJ lL cC
iI jJM ik kKi cC
ikct
jJ cC
A
ikk 'c Z ikk 'ct jlt SF jltB dlctYPdlct
jct F jct
iI kKi k 'Ki cC
jJ cC
dD lL cC
jJ lL
C
Y
A
B
C
B
+ t uVjct (V jct
V jctB ) + uikct
Yikct
(Yikct
) + u SF
jlt ( SF jlt SF jlt )
iI kKi cC
jJ lL
jJ cC
YP1
B
A
YP 2
B
C
+ udlct
YPdlct
YPdlct
(YPdlct
) + udlct
(YPdlct
)
dD lL cC
dD lL cC
General Algorithm
|T| sub-problems
1
Starting next
iteration
Fix:
Yikc,1
Vjc,1
SFjl,1
YPdlc,1
3
Fix:
Yikc,2
Vjc,2
SFjl,2
YPdlc,2
Fix:
Yikc,3
Vjc,3
SFjl,3
YPdlc,3
Fix:
Yikc,t-1
Vjc,t-1
SFjl,t-1
YPdlc,t-1
t=2
t=3
Full Scale
Terminating Criteria
t=T
Yes
Return lower bound and
corresponding solution
Linking Constraints
Proposed strategy
Strategy A
Strategy B
Strategy C
Strategy D
Full-scale
Strategy A
Strategy C
600
400
200
0
0
20
40
time period
60
80
Full-scale
Strategy A
Strategy C
20
40
time period
60
80
Strategy A
Strategy C
0.000
0.000
14
0.554
0.536
21
1.195
1.723
28
0.829
0.826
35
0.833
0.824
42
1.317
1.313
Conclusion
Calculation time
Solution value
Alternative approaches
Decentralized approach
Future work
IV.
Planning, Scheduling and Supply
Chain Management for Operations
in Oil Refineries
JOS M. PINTO
OUTLINE
Introduction
Planning Models
refinery diesel production
Scheduling models
crude oil scheduling
fuel oil / asphalt area
Logistics
oil supply model
MOTIVATION
Profitability Maximization
Refinery Targets
1950s
CPI in general:
(Reklaitis, 1991; Kudva and Pekny; 1993)
(Linear Programming)
(Dantzig, 1963)
1970s
ADVANCES
Availability of more powerful and less expensive computers;
Mathematical Developments:
Time representation;
Combinatorics in MIP;
(Raman and Grossmann, 1994)
Non-convexities in MINLP;
(Viswanathan and Grossmann, 1990)
1980s
1990s
(Symonds, 1955)
(Cutler, DMCC,1983)
Planning models
(Bodington, 1993)
OUTLINE
Introduction
Planning Models
refinery diesel production
Scheduling models
crude oil scheduling
fuel oil / asphalt area
Logistics
oil supply model
Conclusions
UNIT EQUATIONS
- Feed flowrate:
Qu ,F =
Qu' ,s ,u
u U u s S u ,u
- Feed Properties:
Pu,F,j = fj ( Qu,s,u , Pu,s,j )
jJF, sSU
j Js, sSU
Qu ,s =
Qu ,s ,u
u U s,u
s SU
HT MODEL
Feed flowrate:
QHT,F=QCD1,HD,HT+QCD2,HD,HT+QFCC,LCO,HT+QCK,CGO,HD
Feed properties:
PHT,F,j = fj (Qu,s,HT , Pu,s,j )
u UHT , s Su,HT, j
JHT
PHT , F , FP
10006.1
415
= 0 .55
ln( ) + 14 .0922
Qu ,HD ,HT I u , HD , FP
U HT
Qu , HD , HT
U HT
u UHT
REAL-WORLD APPLICATION
Planning of diesel production
Petrobras RPBC refinery in Cubato (SP, Brazil).
Three types of diesel oil:
Metropolitan Diesel.
Regular Diesel.
Maritime Diesel.
DIESEL SPECIFICATIONS
Property
DENSITY
min / max
FLASH POINT
min (C)
ASTM 50%
min / max (C)
ASTM 85%
max (C)
CETANE NUMBER min
SULFUR CONTENT max
(% WEIGHT)
REGULAR
0.82/
0.88
DIESEL
METROPOLITAN MARITIME
0.82/
0.82/
0.88
0.88
60.0
245.0/
310.0
245.0/
310.0
245.0/
310.0
370.0
360.0
370.0
40.0
0.5
42.0
0.2
40.0
1.0
MAIN RESULTS
Potential Improvement
OUTLINE
Introduction
Planning Models
refinery diesel production
Scheduling models
crude oil scheduling
fuel oil / asphalt area
Logistics
oil supply model
Conclusions
LPG Scheduling
Crude
Scheduling
FRACTIONATION
OBJECTIVES
subject to:
Timing constraints
Pipeline material balance equations
Pipeline operating rules
Pipeline always connected to a tank
DECISION VARIABLES
slot k
Ydf,j,k
Ypj,k
fraction f
REAL-WORLD EXAMPLE
Volume
Start time
End Time
(m3)
(h)
(h)
60,000
20
100% Bonito
50,000
48
58
100% Marlin
1,000
58
58.2
100% Marlin
60,000
100
112
100% RGN
Oil parcel
Composition
RESULTS
MODEL SOLUTION
GAMS / OSL
CPU time
2.80 hrs (Pentium II 266 MHz 128 MB RAM)
OUTLINE
Introduction
Planning Models
refinery diesel production
Scheduling models
crude oil scheduling
fuel oil / asphalt area
Logistics
oil supply model
Conclusions
Product Base
FO1
FO2
FO3
FO4
UVO1
UVO2
CAP07
CAP20
RASF
RASF
RASF
RASF
RASF
RASF
RASF
RASF
Diluent used
OCC+LCO or OCC or LCO
OCC+LCO or OCC or LCO
OCC+LCO or OCC or LCO
OCC+LCO or OCC or LCO
pure LCO
pure LCO
pure HG
pure HG
major specification:
viscosity
MATHEMATICAL MODELS
Uniform Discretization of Scheduling Horizon;
Objective Function: Minimize the Operational Cost.
First Approach:
non-convex MINLP (5 bilinear products in the viscosity constraints);
Linear Transformation
Second Approach:
MILP;
MINLP MODEL
Minimize:
Raw-Material Costs + Inventory Costs +
+ Pumping Costs + Transition Costs
COST =
COST =
t =1
s =1
i =1
q =1
d =1
o = 1 p =1 n =1
Subject to:
Material Balance Constraints:
volume in i at t = initial volume in i
i =1
q =1
(XICi,t ) + (XQCq,t ) = 1
FO area
t = 1,...,T
UVO/asphalt area
simultaneous tank loading and unloading is not allowed (exception: HG storage tank)
O
XICi,t + (XIDi,o,t ) 1
o =1
i = 1,...,I; t = 1,...,T
FO area
XQDq,t HTb
b = 1,..., [ T /( 12 / DT )];
( 12 / DT ) ( b 1 ) + 1 t ( 12 / DT ) b;
q = 1,...,Q
(XQCq,t ) XDRASF1,t = 0
t = 1,...,T
q =5
while asphalt is produced, the OCC stream from UFCC must be directed to storage in TK-42208
XDRASF1,t + ( 1 XZ t ) 1
t = 1,...,T
q = 1,...,Q; t = 1,...,T
Viscosity Constraints:
at each t, the viscosity adjustment must be done regarding the kind of product
Q
VISCt =
q =1 vVq
t = 1,...,T
i =1 pPi
{[
s2
d =1 sS d
= VISCt
d =1
t = 1,...,T
or
D
{[
s2
d =1 sS d
t = 1,...,T
d =1
5 bilinear products
non-convex MINLP
Structure:
MILP Model = MINLP Model +
Linearized Constraints
t'
t =1
o =1
(46)
(47)
t =1
FRASFU t MIRASF +
s2
d =1 sS d
I
i =1
q =1
Flow
U = cons tan t
VI i ,t MI i = VIK i ,t
Viscosity
Material Balance
U = cons tan t
t = 1,...,T
i = 1,...,I; t = 1,...,T
q = 1,...,Q; t = 1,...,T
i = 1,...,I; t = 1,...,T
q = 1,...,Q; t = 1,...,T
(48)
(49)
(50)
(51)
(52)
FIDi ,o ,t MI i = FIDK i ,o ,t
(53)
q = 1,...,Q; t = 1,...,T
(54)
TK-4
41 13
TK-4
33 07
TK-4
33 01
TK-4
33 01
TK-4
33 07
TK-4
33 0
TK-4 1
33 02
TK-4
41 08
TK-4
33 02
TK-4
41 08
TK-4
33 01
TK-4
41 08
TK-4
33 01
TK-4
41 08
TK-4
33 02
TK-4
33 07
TK-4
33 01
TK-4
41 1
TK-4 3
33 01
TK-4
33 01
TK-4
33 07
TK-4
33 07
TK-4
33 01
TK-4
41 08
TK-4
41 08
TK-4
33 07
TK-4
33 01
TK-4
33 01
TK-4
33 01
TK-4
33 0
TK-4 3
33 02
TK-4
33 07
TK-4
33 01
TK-4
33 01
TK-4
33 01
TK-4
33 01
REAL-WORLD EXAMPLE
Scheduling horizon: 3 days
instance
evaluated
UVO1
START
PRODUCTION SCHEDULE AND STORAGE INFORMATION
UVO2
CAP07
CAP20
FO1
FO2
END
FO3
FO4
Each interval = 20 m /h
180
160
140
120
100
80
60
40
20
0
11
13
15
17
19
21
23
25
27
29
31
33
35
TK-43302 (i=2)
TK-43303 (i=3)
TK-43304 (i=4)
1
TK-44113 (q=1)
TK-44114 (q=2)
36
TK-43305 (i=5)
TK-43306 (i=6)
1
36
FO4 (p=4)
TK-43307(i=7)
36
CAP-07 (v=3)
TK-44108 (q=5)
TK-44111 (q=3)
TK-44112 (q=4)
1
10
8
6
4
2
0
FO3 (p=3)
UVO2 (v=2)
36
10
8
6
4
2
0
FO2 (p=2)
7 10 13 16 19 22 25 28 31 34 37
UVO1 (v=1)
2
0
12
10
8
6
4
2
0
FO1 (p=1)
TK-43301 (i=1)
10
8
6
4
2
0
36
CAP-20 (v=4)
36
TK-44110 (q=6)
TK-44115 (q=7)
TK-44116 (q=8)
36
st
day)
(m3 /h)
FO2
100
80
60
40
20
0
FO3
300
FO4
200
100
0
1
(m3/h)
400
nd
day)
FO1
(m3 /h)
FO2
100
80
60
40
20
0
FO3
300
FO4
200
100
0
1
(m3/h)
400
nd
/3
rd
days)
FO1
(m3 /h)
FO2
100
80
60
40
20
0
FO3
300
FO4
200
100
0
1
(m3/h)
400
rd
th
days)
FO1
(m3 /h)
FO2
100
80
60
40
20
0
FO3
300
FO4
200
100
0
1
11 13 15 17 19 21 23 25 27 29 31 33 35
11 13 15 17 19 21 23 25 27 29 31 33 35
9 11 13 15 17 19 21 23 25 27 29 31 33 35
11 13 15 17 19 21 23 25 27 29 31 33 35
11 13 15 17 19 21 23 25 27 29 31 33 35
11 13 15 17 19 21 23 25 27 29 31 33 35
9 11 13 15 17 19 21 23 25 27 29 31 33 35
9 11 13 15 17 19 21 23 25 27 29 31 33 35
Refinery
Market
to
t1 > to
t2 > t1
Final Product A
Final Product B
t3 > t2
Final Product C
UNDESIRABLE MIX
TRANSITION COST
(m /h)
400
LOCAL OIL-PIPELINE
(WITH TRANSITION MODELING)
FO1
FO2
FO3
300
FO4
200
6890
4465
2629
1512
1512
100
MINLP model
0
1
(m /h)
400
MILP model
11 13 15 17 19 21 23 25 27 29 31 33 35
OIL-PIPELINE TO SO PAULO
FO1
FO2
FO3
300
FO4
200
1512
100
0
1
case
A
B
C
D
11 13 15 17 19 21 23 25 27 29 31 33 35
MIP model
MILP
MINLP
MILP
MINLP
MILP
MINLP
MILP
MINLP
nodes
937
1296
764
1197
-
iterations
15674
13815
16626
15508
13086
23792
23080
12845
WITHOUT TRANS.
CONSTRAINTS
WITH TRANS.
CONSTRAINTS
objective
969.61
966.99
965.72
961.14
954.99
956.99
950.65
959.49
OUTLINE
Introduction
Planning Models
refinery diesel production
Scheduling models
crude oil scheduling
fuel oil / asphalt area
Logistics
oil supply model
Conclusions
Solution of oil supply problems among crude oil terminals and refineries
MOTIVATION
Increasing utilization of the system
Larger crude oil demand for crude oil in refineries
Outsource of transportation
PROBLEM SPECIFICATION
iApproximately 42 types of
crude oil may be processed
Types of
crude oil
PROBLEM SPECIFICATION
iSets of crude oil types with
similar properties
i7 classes
Types of
crude oil
PROBLEM SPECIFICATION
Classes of
crude oil
Types of
crude oil
PROBLEM SPECIFICATION
Piers
iDifferent capacities
Tankers
Classes of
crude oil
Types of
crude oil
PROBLEM SPECIFICATION
Piers
Tanks
Tankers
Classes of
crude oil
Types of
crude oil
PROBLEM SPECIFICATION
Piers
Tanks
Pipelines
Tankers
Classes of
crude oil
Types of
crude oil
iPossible to connect to at
most one tank at every
time
PROBLEM SPECIFICATION
Piers
Tanks
Pipelines
Tankers
Substations
Classes of
crude oil
Types of
crude oil
iBuffer operations
between terminal and
refineries
OBJECTIVES
Tankers
Operating
constraints
Initial
inventory
Costs
Possible
allocations
iSchedules
Tanks
Crude types
Substations
Pipelines
Refineries
Piers
Mathematical formulation
Input
parameters
System
operation
- Allocation of
crude oil types
to classes
- Assignment of
tankers to piers
- Loading
- Unloading
- Settling
PROPOSED STRATEGY
Pipelines
Terminal
Impossible to solve
complete problem
Substations
Refineries
PROPOSED STRATEGY
iDecomposition of the problem in three
formulations
- Port Model
- Substation Model
- Algorithm to adjust timing of pipelines
PROPOSED STRATEGY
Port Model
Port
Results
Allocation of tankers to
piers
Loading and unloading
profiles of tanks
Loading of pipelines
Timing of interfaces in
pipelines
PROPOSED STRATEGY
- Algorithm to adjust timing of
pipelines
Pipelines
Port
PROPOSED STRATEGY
Substation Model
Pipelines
Port
Substations
PROPOSED STRATEGY
- Algorithm to adjust timing of
pipelines
Pipelines
Terminal
MATHEMATICAL FORMULATION
MILP model formulation
Time representation
Continuous
Based on events
Inventory level (cont. variable)
Vi
Vi+1
Vi+2
Vi+3
Qi
Qi+1
Qi+2
Qi+3
Xi
Xi+1
Xi+2
Xi+3
time
Ti
Ti+1
Ti+2
Ti+3
Continuous variables
Timing
Inventory
Flowrates
Operating profit
An , p
LTn ,t ,e
UTt ,o ,e
TN
s
n ,e
= TT
f
n ,e
= TT
s
t ,e
f
t ,e
TN
s
o ,e
TT = TD
TT = TD
s
t ,e
f
t ,e
f
o ,e
pPn
LT
Operation of tank t:
n ,t , e
nN t
Operation of tanker n:
=1
n, p
LT
tTn
n ,t , e
oOt
UT
tTo
+ UTt ,o ,e 1
t ,o ,e
Operational constraints
Ships and tanks:
flowrate bounds
Timing
Ships
TN ns,e
Tanks
pPn
start
n, p
+ An, p . nent, p
TN ns,e TN nf,e1
TN nf,e
TT H
f
t ,e
oOt
end
n, p
pPn
LT
nN t
TTt s,e
n , t , e
TTt ,fe 1
1)
TTt ,se H .(1 UTt ,o,e ) TDos,e TTt ,se + H .(1 UTt ,o ,e )
TTt ,fe H .(1 UTt ,o,e ) TDof,e TTt ,fe + H .(1 UTt ,o ,e )
r clCLRr
cl
E 1
REVRclclass
, r .
Cn,c
c
nN c
pier
end
start
COST p . n, p n, p (pier utilization cost)
p
nN p
COSTccrude .
COSTnse .Tnse
E 1
face
COSTcl ,cl . INTcl ,cl ,o,e
o clCLOo cl CLOo
e =1
cl cl
(interface cost)
- Subject to:
- Assumptions of the substation model
- Operating constraints
8 Tank loading/unloading
8 Pipeline operation
- Timing constraints
8 Inlet pipelines
8 Tanks
8 Outlet pipelines
REAL-WORLD PROBLEM
Paulnia
REPLAN
III
T
vA
S
O
VA
RECAP
V
TI
Problem 4
SEGUA
VA
Substation
Model
T
II
SEBAT
Guararema
Problem 2
OS
Substation V
I
T
Model
A
OS
OS
Capuava
So Jos dos
Campos
REVAP
RPBC
OS
VA
Substation
Model
II
TI
Problem 3
BA
OS
Cubato
GEBAST
Problem 1
T II
OSBA
Port Model
P3
P4
P1
P2
So Sebastio
COMPUTATIONAL RESULTS
Smaller optimality gaps for the Port Model
Large variation on computational times
Problem 1
Problem 2
Problem 3
Problem 4
1996
1039
7203
21,768.32
20,073.96
7.78%
1118
62313
1,457.51 s
4954
759
10337
23.00
42.00
82.61%
3784
74410
3,602.07 s
712
66
1158
11.00
21.00
90.91%
3921
19321
134.69 s
703
123
1682
11.39
15.00
31.74%
422
5244
28.28 s
Port Model
Substation Models
50
40
30
20
FRONT BREA
120
08
90
27
10
0
60
05
60
40
30
20
0
30
35
40
TBN01
60
20
40
15
15
20 25
TBN02
10
20
62
64
66
68
86
87
88
TQ3214
80
60
10
30
89
85
30
20
40
90
95
100
0
105 139
0
141
143
TQ3215
Vmin
Vmax
24
48
72
96
TQ3219
80
Vmin
0
24
48
72
96
Vmax
40
20
Vmin
0
0
24
48
72
96
TQ3237
80
Vmax
60
40
Vmin
20
0
0
24
48
72
96
TQ3241
80
48
72
96
Vmax
Vmin
20
0
Vmin
0
24
48
72
96
TQ3242
Vmax
60
Vmin
24
48
72
96
24
48
72
96
110
96
120
TQ3210
35
28
21
14
7
0
Vmax
Vmin
0
24
48
72
96
TQ3218
80
Vmax
60
40
20
Vmin
Vmin
0
0
24
48
72
96
TQ3234
80
24
48
72
96
TQ3235
80
Vmax
60
Vmax
60
40
Vmin
20
20
Vmin
0
0
24
48
72
96
TQ3239
80
24
48
72
96
TQ3240
80
Vmax
60
Vmax
60
40
Vmin
20
0
Vmin
20
0
24
48
72
96
TQ3243
80
Vmax
60
24
48
72
96
TQ3244
80
Vmax
60
40
40
20
0
72
40
40
40
48
Vmax
20
TQ3238
80
Vmax
60
24
35
28
21
14
7
0
24
0
0
40
20
Vmin
100
Vmin
TQ3217
60
TQ3233
60
40
168
5
0
90
Vmax
80
Vmax
60
161
29b
TQ3208
35
28
21
14
7
0
10
29a
0
100 105 110 115
40
20
145 154
80
40
20
15
40
20
95
20
10
90
100
20
60
26
0
90
95
TBN04
CANTAGALO
80
50
85
50
40
60
40
45
TBN03
60
D
100
0
40
PRESIDENTE
NORTH STAR
150
38
90
20
60
80
Vmax
60
65
75
RAVEN
VERGINA II
120
0
55
70 85
38
30
01b
45
0
60
01a
20
30
REBOUAS
40
0
10
45
MURIA
80
Vmin
20
0
Vmin
20
0
24
48
72
96
24
48
72
96
34.8
53.3
35
OSBAT
58.2
TQ3244
28.0
27.8
2.2
28.0
2.8
27.8
48.8
19.6
25.4
17.4
OSVAT
2.2
34.4
16.9
SEGUA/E
41
65
2.8
48.8
42.4
63.6
67.7
6.1 (P4)
26.0 Hrs
63.6 (P2)
TQ3242
58.2
2.2
24.3 Hrs
TQ3241
42.4 (P2)
TQ3240
42.4
62.4
16.9
60
60
61.8
62.4
61.8
50.3Hrs
26.0 (P2)
59.1
60.4
16.9
66.0 (P2)
62.4 (P2)
24 Hrs
62.4 (P1)
62.4
6.2 (P2)
60.0
24 Hrs
60.0 (P4)
38.0 (P4)
46.4 Hrs
14.1
55.7 (P1)
29.4 (P4)
67.7
24 Hrs
27.8
16.9 (P1)
16.9
24 Hrs
28.9 (P3)
1.4 (P2)
TQ3237
51.1 (P3)
3.7 (P4)
TQ3235
26.5
58.7
TQ3239
TQ3238
62.4
2.6
11.5
63.6
48.8
70.2
65.1
19.6
28.0
TQ3243
19.6
33.9
11.9 (P4)
TQ3234
60.0 (P4)
0.6 (P4)
TQ3233
25.4
30L (P-4)
4.5 (P4)
58.7
24 Hrs
2.8
4.7 (P4)
TQ3219
TQ3218
15 (P4)
TQ3217
19.6
24 Hrs
11.5 (P4)
11.5
43.2 Hrs
55.5 (P4)
5.3 (P3)
TQ3215
24.7 (P3)
TQ3214
12.2 (P2)
17.4
TQ3208
P-4
60.4
37.8 (P2)
TQ3210
19.6
REBOUCAS
CANTAGALO
TBN01
PEDREIRAS
P-3
TBN02
RAVEN
TBN03
TBN04
PRESIDENTE
P-2
MURIAE
FRONT_BREA
P-1
0
24
48
VERGINA II
72
Tempo (horas)
NORTH_STAR
96
120
144
168
1600
Max
1200
REPLAN+REVAP (detailed)
1000
RECAP+RPBC
600
950
400
900
200
350
Max
RECAP+RPBC (detailed)
340
800
330
400
Min
850
0
24
48
72
CL-5
GEBAST
320
Min
0
0
24 48
72
24
48
72
310
96 120 144 168
0
GEBAST
24
48
72
Flux de Oil
CL-6
CL-4
CL-5
CL-5
CL-5
Station
de
Rio
Pardo
CL-6
CL-5
CL-4
Flux de Oil
CL-5
CL-7
CL-7
CL-6
CL-6
CL-6
CL-7
0
24
SEGUA CL-3
CL-3
48
72
96
Tempo (hours)
120
144
CL-4
168
0
RPBC
24
48
72
96
Tempo (hours)
120
5
4
3
2
1
0
144
168
Station
de
Guaratuba
0
0
24
48
72
96
120
144
168
24
48
72
96
120
144
168
10.3
2.3
8.4
11.2
2.3
OSBATIV/E
8.5
12.6
11.2
8.4
4.6
TQ42
8.3
1.4
2.6
11.2
4.6
TQ41
1
8.5
7.5
15.2
TQ40
8.4
4.3
12.6
2.6
4.6
OSBATIII/S
1
0
7.5
24
8.3
1.4
48
72
Time (hours)
4.3
96
15.2
120
4.6
144
168
TQ40
25
25
20
Vmax
20
15
15
10
10
Vmin
Vmax
Vmin
0
0
Max
TQ42
25
20
15
10
5
0
Vmax
Vmin
0
80
70
40
Min
0
0
SEBAT
Flux de Oil
CL-7
CL-7
CL-7 CL-6
CL-6
0
0
RECAP
24
48
72
96
120
144
168
24
48
72
OUTLINE
Introduction
Planning Models
refinery diesel production
Scheduling models
crude oil scheduling
fuel oil / asphalt area
Logistics
oil supply model
Conclusions
CONCLUSIONS
Problems can be modeled as large scale MILPs / non-convex MINLP;
The LP based Branch and Bound Method (solver OSL):
is satisfactory to generate good feasible solutions;
no guarantee of global optimum solutions for all instances;
The OA/ER/AP Method (solver DICOPT++):
is efficient to circumvent the non-convexity problem;
is satisfactory to generate feasible solutions;
has computational performance similar to MILP model.
Issues:
time representation
blending/pooling
transitions
CONCLUSIONS - CHALLENGES
Large Scale Systems - Main theoretical difficulties:
Complex problems with high combinatorial features;
NP-Complete Problems
ACKNOWLEDGEMENTS
Researchers
Engineers
Financial Support
M. Joly
R. Ms
L. Moro
R. Rejowski
P. Smania
C.A. Gratti
M. F. Lehner
M.V. Magalhes
A.C. Zanin
Petrobras
CAPES
CNPq
FAPESP
Oil field
Infrastructure
International
Petroleum
Commerce
Operations
Research.
Crude Oil
Supply
Refinery Planning/
Scheduling
Ponnambalam (1992).
Bok et al. (1998).
Pinto and Moro (2000).
Distribution
Ross (2000).
Iakovou (2001).
Magato et al. (2002).
Stebel et al. (2002).
Rejowski and Pinto (2003).
Objective
QFu ,t =
Qu',s ,u ,t
( u',s )USu
Qu',s ,u ,t .PSu',s , p ,t
PFu , p ,t =
( u',s )USu
Qu',s ,u ,t
( u',s )USu
Feed mixing
QSu ,s ,t =
Qu ,s ,u',t
u' UOu ,s
VuL,v Vu ,v ,t VuU,v
Unit
model
Bounds
QFu ,t =
( u',s )USu
Qu',s ,u ,t
QSu ,s ,t = Qu',s ,u ,t
QSu ,s ,t = Qu ,s ,u',t
QFu ,t QFuU
uU SB t T
vVOu
uU f t T
uU pipe t T
PF ,PS ,V
y {0,1}
RPBC flowsheet
Intermediate connections
Modeling Example
,s SOCD1 , p POCD1,s ,t
QS
CD1, p ,t + QGainCD1,s
,V1 t.VCD1,V1 ,t ,t
QSCD1,s ,t ==QFCD1,t .PF
= QQ
,
QS
tan
k 1, petroleum
,CD1,t
CD1,s
,u',t
p
s
Qtan
500
20000,t
k 1, YC3C4
petroleum ,CD1,t
C3C4
LN
YLN
HN YHN
K
YK
Q
Qu ', p etro leu m ,C D 1 ,t , t
F C D 1 ,t =
u ' {ta n k 1 ,2 ,3 }
LD
YLD
n k 1 ,2 ,3 }
=
P
F
, t
C D 1 , p ,t
ATR
YATR
u ' {ta n k 1 ,2 ,3 }
p {Y C 3 C 4 ,Y L N ,Y H N ,Y K ,Y L D ,Y H D ,Y A T R }
Steep
Steep
c2
P2%
probc,t =1
cN
PN%
c C
Cbu yu ,t ,c
[ Cru + ( Cvu,v .Vu ,v ,t )] QFu,t
cC t T u U f
vVO u
cC tT uU \ {U f ,U p }
Cinvu ,t ,cVolu ,t ,c
t T u U p
Dual subproblems
Multipliers update
Strategy 1
Fixed assigment
Lagrangean
Subgradient
Strategy 2
Fixed inventory
Lagrangean
Subgradient
Strategy 3
Fixed inventory
Surrogate
Subgradient
Strategy 4
Fixed inventory
Lagrangean
Modified Subgradient
700
CPU seconds
600
500
400
300
200
100
0
0
10
20
30
40
50
Strategy 1
Strategy3
Strategy4
Strategy 2
60
General constraints:
Planning horizon: one / two time periods
Supply of 20 oil types
Generation of 32 products (6 transported with pipelines)
Case 2
Case 3
Larab
Marlin
Rgn
Cabiun
Albaco
Bicudo
Condoso
Bonit
Cabiuna
Larabe
Coso
600
600
660
0
0
60
10905
11404
7551
0
0
6551
16945
16489
11400
1389
1408
1389
15
15
15
8500
8500
8500
35500
35500
35500
90200
90200
80192
4400
4400
11660
3345
2889
10800
24540
25208
0
6403
6403
6403
6545
6089
11000
20018
20019
20018
17700
17700
17700
PFO1
A
60
224
666
48
0
58
0
53
0
PFO1
A
PFO4
A
Case 1
1
Case 2
1
Case 3
1
Constraints
2304 4607
2306 4611
2304 4607
Variables
2544 5087
2544 5087
2544 5087
Discrete variables
Solution time (CPU s)
Objective Value ($ x106)
195
390
116.8 656.2
20.4
41.3
195
390
195
390
152 915.6
157.8
2301
18.0
36.3
20.3
41.1
Conclusions
Mathematical programming
-General refinery topology
-General petroleum supply chain representation
-Representation of nonlinear properties and multiple periods
-Non-convex Large-Scale MINLP
Real-world applications
-General planning trends along multiple periods
-Analysis of scenarios (discrete probabilities)
-Intensive computational effort
Research needs
Modeling
9Upstream-Downstream Integration
9Multi country supply chains (royalties, tariffs)
9Modeling of uncertainties