OIL PIPELINE LOGISTICS - Carnegie Mellon University

OIL PIPELINE LOGISTICS

Jaime Cerdá

Instituto de Desarrollo Tecnológico para la Industria Química Universidad Nacional de Litoral - CONICET Güemes 3450 - 3000 Santa Fe - Argentina Pan American Study Institute on Emerging Trends in Process Systems Engineering August 11-21 , Mar del Plata, Argentina 1

OUTLINE

Motivation

The multiproduct pipeline planning problem

Available pipeline planning approaches

Presentation of a continuous planning approach

Critical operational decisions & major problem constraints

An illustrative example

Static vs dynamic planning problem

The detailed weekly pipeline schedule

Conclusions

2

ACKNOWLEGMENT

The material included in this presentation have been extracted from DIEGO C. CAFARO’s Doctoral Thesis

currently in preparation

3

LIQUID PIPELINE OVERVIEW Most reliable, safest and cheapest way of delivering large volumes of a wide range of refined products from refineries to distant depots. Batches of different grades and products are pumped back-to-back in the line without any separating device. Batches move forward in the line and products are transferred to terminals whenever a new batch is injected at the head terminal. Distribution Terminals D1

D2

D3

D4

D5

Head Terminal Segment P2

Segment

P1

P3

P4

Refinery Interfaces 4

PIPELINE MAJOR FEATURES Usually buried and invisible to the public

With several intermediate entry and exit points With segments of varying diameter

Trans Alaska Pipeline System

Large diameter pipelines due to high construction costs With crude oil and refined products moving in separate lines Always remaining full of liquid and pumping in only one direction.

REFINERIES

Colonial Pipeline

COLONIAL

5

PIPELINE OWNERSHIP – REMOTE OPERATION Owned by a large number of companies, almost all are common carriers An increasing number are owned by non-oil companies Operations are fully automated and remotely performed From centrally located control rooms, operators direct the product flow From there, they start & stop pumps, open & close valves, fill & empty tanks Supervisory control & data acquisition systems, known as SCADA, are used SCADA continuously monitors:

pump pressures batch locations

flow rates tank levels

6

PIPELINE ADVANTAGES Operate around the clock all seasons and under all weather conditions No container moves with the cargo. Products only move. No backhauls Employment is only 1% of that of the trucking industry THE CHEAPEST MODE OF TRANSPORTATION

BUT THE SLOWEST MODE (3 TO 8 MPH)

Very low transport damage to products and especially to the environment. Lines coated with corrosion-resistant chemicals to prevent corrosion Chance of leaks reduced by an extensive maintenance program “Smart pigs” sent through the line - detect dents and imperfections - measure wall thickness

THE SAFEST MODE

“Scraper pigs” clean the inside of a line by removing residual material clinging to the walls

Scraper PIGS

7

INTERMODAL PRODUCT MOVEMENTS Pipelines dominate the oil industry transportation

80

PIPELINES

70 60

Participate in intermodal product movements with other modes of transportation

50 40

VESSELS

30

Pipelines (%) 20

Vessels (%)

- tankers & pipeline combination for crude oil

10 0 1980

- pipeline/truck combination for refined products

Trucks (%) Trains (%) 1985

1995

2000

2005

CRUDE OIL DOMESTIC TRANSPORT MARKET IN USA

A batch in the line arriving at a terminal: - can be placed in a tank - can be rerouted into another pipeline Lines provide tanks to buffer the flow rates between two connecting pipelines or line segments of different diameters

1990

REFINERIES

PIPELINES

CLIENTS CRUDE OIL IMPORTS OIL FIELDS

DISTRIBUTION TERMINALS

8

MONITORING BATCH STATUS The specific gravity of the flow is continuously monitored at every terminal

When it changes, the operator knows that: - one product batch is ending - another product batch is beginning to arrive The operator can visually observe the transition

Refined products are often “color-coded” with dye

Distribution Centers D1

D2

D3

D4

D5

Head Terminal P2

P1

P3

P4

Refinery Interfaces 9

POWER CONSUMPTION Liquid products are propelled by centrifugal pumps sited at the pumping stations one at the origin and the others distributed along the line. The capacity of a pipeline can be increased by installing additional pumping stations along the line to rise pressure. The power consumption is the largest pipeline operating cost.

10

INTERFACE MATERIAL Pipelines move different grades of a product or distinct products sequentially through the same line in “batches”. At the boundary of two consecutive batches some mixing occurs. Between batches of different grades

Interface

Mixed with the lower grade product PRODUCT DEGRADATION

Between batches of different products

Transmix

Separated and sent back to the refinery TRANSMIX REPROCESSING

11

INTERFACE COSTS Product degradation and transmix reprocessing costs both significantly contribute to the pipeline operating cost. Amount of Interface

Some products are prohibited to be consecutively injected to avoid a serious product degradation.

Number of batches

Arrangement of batches in the line

CRITICAL DECISIONS Batching Operations Batching Sequencing

Keep similar products from different refiners together Inject the lowest possible number of product batches Sequence batches by specific gravities

Batch sequencing is also important to meet product delivery due dates at terminals 12

PIPELINE OPERATING MODES More stringent environmental regulations on car fuels have resulted in a proliferation of refined products. Major refined product pipelines currently move 100-120 distinct products compared with 10-20 in the ‘60s. OPERATING MODES

BATCH MODE

FUNGIBLE MODE

the same volume accepted for shipment to a particular depot is the one delivered to that destination

standard refined products from different refiners are consolidated into a single batch

LARGER NUMBER OF BATCHES

SMALLER NUMBER OF BATCHES

HIGHER INTERFACE COSTS

LOWER INTERFACE COSTS 13

PIPELINE BATCHING OPERATIONS THREE PRODUCTS :

P1, P2 , P3

SELECTED BATCH SEQUENCE: P1 – P3 – P1 – P2 THE SAME AMOUNT OF PRODUCTS SHIPPED TO TERMINALS TIME HORIZON: 0

4 WEEKS weeks

1 P2

P1

P3

P2

P1

P1

2

P2

P1

P3

P2

T ra n sp orte

P1

P2

P1

P1

P3

P2

P1

Inventory Level N ive l d e In ve n ta rio d e P roof d u cto P P2 2 e n at e l Dthe e stinDepot o , se g ú n e l T ie m p o d e C iclo (T C ) Transportation T ie m p o de Time

P3

T ie m p o time [se m a n a s]

3 P1

P1

P3

P2

P1

P3

P3

4 P1

P1

P1

T ie m p o d e Number Period N o. de C iclo (T C ) B a tch e s Length of batches [d ía s]

7

7

116 6

141 4

88

282 8

4 4

4-PERIOD HORIZON 2-PERIOD HORIZON ONE-PERIOD HORIZON

days TC = 28

TC = 14 TC = 7 T ie m p o

time

SHORTER PERIOD LENGTH – SAME BATCH SEQUENCE IN EACH PERIOD LARGER NUMBER OF BATCHES AND INTERFACE COSTS SMALLER BATCHES AND LOWER TERMINAL TANK CAPACITIES 14

STRIPPING OPERATIONS (“CUTS”) Every new batch injection pushes some batches forward while others that arrive at their destinations are partially or completely sent out of the line (“stripping operations”) and loaded in the terminal tank. Therefore, both the size and the location of every batch in the line can change during the pumping of a new batch. Batch stripping takes place if the batch has arrived at the terminal and enough storage capacity to receive the material is available. Otherwise, the line should be temporarily stopped and deliveries are interrupted.

REF

B4 B5

D1

B5 B4 B3 B4

B3 B3 B2B2 B2

D2

B1 B1 B1 15

BATCH DUE DATES & DELIVERY LEAD-TIME A fungible batch may satisfy several product requirements at different terminals, i.e. multiple destinations. A fungible batch with multiple destinations will undergo several stripping operations (“cuts”) along the journey.

Every product delivery has its own due date. Multiple destinations Fungible batch Multiple due dates

A batch can travel to the farthest destination for 7-14 days (“delivery lead-time”). Delivery lead-time

Is a function of

Depot location Pumping rate Pipeline idle time

Most short-term product requirements are satisfied by batches currently in transit. 16

LOADING & UNLOADING OPERATIONS Terminals have few tanks just to facilitate stripping operations and quality control tasks. In fungible mode, a fewer number of larger storage tanks is usually needed. Tanks for long-term storage must be provided by the customer at entry & exit points. A common carrier pipeline terminal typically connects to the marketing terminals of its main shippers or to public storage terminals. Gasoline tank trucks are loaded from storage tanks at marketing terminals

Pipeline Terminal

Marketing Terminal Trunk Line

17

SHIPPER NOMINATIONS

US pipelines are mostly COMMON CARRIERS, i.e. services are provided to multiple oil refiners.

Customers contact the pipeline operator to place their shipment orders for the next month, called NOMINATIONS.

A NOMINATION specifies the product and the quantity to be shipped.

Customers should make the product timely available at the input terminal and provide enough storage capacity at its destinations.

The monthly planning horizon is composed by a number of periods, called CYCLES. Every nomination is divided into a number of equal-size batches, one for each cycle. A cyclic schedule is usually performed. 18

THE PIPELINE SCHEDULING TASK Planning pipeline operation in fungible mode implies to choose: - the set of batches of each product to be injected, and the batch sizing - the sequence of batch injections - batch injection rates and starting times

Operational decisions concerning to every batch to be injected include: - the assigned destinations (terminals) - the amount allocated to each destination (the cut sizing)

Operational decisions related to each batch pumping run include: - the set of “stripping operations” to be carried out in-transit batches to be stripped out - receiving depots - cut sizes - the location & size of every in-transit batch at the end of a batch injection 19

BATCH INJECTION & STRIPPING OPERATIONS

Depot D3

180

200

B4

B3

B2

B1

STRIPPING OPERATIONS

50

30

190

50

200

C5-L5 _ C5 150

Depot D2

150

150

160

130

180

B5

B4

B3

B2

B1

P1

REFINERY NEW BATCH B5

P2

P3

At time C4

CURRENT PIPELINE STATE

At time C5

NEW BATCH INJECTION

20

Depot D1

P4

STRIPPED BATCHES B4 – B3 – B2 – B1

20

PIPELINE SCHEDULING GOALS To minimize operating costs including: - the transmix reprocessing cost & the product degradation cost - the pumping cost - the inventory costs in refinery and depot storage tanks

To meet product delivery requests on time

To keep the pipeline running at nearly maximum capacity during off- peak hours

To enhance the information on the current status of batch movements

21

PROBLEM DATA

The sequence of “old” batches already inside the pipeline.

Their locations & volumes at the initial time of the planning horizon.

The scheduled production runs at the refinery.

The inventory levels in refinery and terminal tankage at the initial time.

The set of shipment requests, each one involving a refined product, the assigned terminals and the delivery due dates.

22

PIPELINE SCHEDULING APPROACHES Knowledge-based Search Techniques (Sasikumar et al., 1997) Metaheuristic Search Algorithms

Greedy algorithms (Hane & Rattliff, 1995) Genetic algorithms (Nguyen & Chan, 2006) Tabu search (García et al., 2008) Cyclic Scheduling Techniques (Used by pipeline schedulers)

Mixed-Integer Mathematical Programming Formulations - Discrete Formulations (Rejowski & Pinto, 2003) - Continuous Formulations (Cafaro & Cerdá, 2004 & 2008; Relvas et al., 2007) Discrete Event Simulation (Maruyama Mori et al., 2007) 23

MIP DISCRETE FORMULATIONS Discrete Formulations (Rejowski & Pinto, 2003)

Pack 1 Pack 2 Pack 3 Pack 4 t = T1

P1

P2

P1

P1

Very large MILP formulations for longer planning horizons

T2

P1

P1

P2

P1

T3

P1

P1

P1

P2

... The pipeline is divided into packs of uniform size at each segment

Each pack contains exactly one product

The time scale is divided into slots of fixed length (fixed pumping rate) Whenever a pack of product enters a segment, the content of the first pack in that segment is displaced to the next pack. 24

MILP CONTINUOUS APPROACH MAJOR FEATURES Continuous time & volume representation Pre-defined ordered sequence of empty batch slots of variable-size Multiperiod planning horizon Explicit treatment of interface volumes Delivery due dates at the end of every planning period A “cheap” generalization to pipelines with several intermediate input and exit points

25

MILP CONTINUOUS APPROACH MAJOR DECISION VARIABLES Allocation variables assigning products to “empty” batch slots Control variables indicating the arrival of a batch at the assigned terminal to start the stripping operation Assignment variables denoting the planning period at which a batch injection ends MAJOR CONTINUOUS VARIABLES Starting and completion times of new batch injections (the time events) Initial sizes of batches to inject in the pipeline Location and size of in-transit batches at the end of a new batch injection Stripping operations to take place during a batch injection (batch to be stripped, cut size, receiving terminal) Inventory levels at refineries and pipeline terminal tanks at every time event 26

MAJOR MODEL CONSTRAINTS A single product can at most be assigned to a batch slot

∑ yi , p ≤ 1

∀i ∈ I new

p∈P

The size of the interface between consecutive batches depends on the assigned products flow P1

P3 B2

B1

WIFi , p , p ' ≥ IFp , p ' * ( yi −1, p ' + yi , p − 1) ∀i ∈ I , i > 1 p, p´∈ P

WIFB2,P1,P3 = IFP1,P3

A new batch injection can be started after completing the previous one Ci − Li ≥ Ci −1 + τ p , p´ * ( yi −1, p ' + yi , p − 1) ∀i ∈ I new ; p, p´∈ P

Li ≤ Ci ≤ hmax

∀i ∈ I new

27

MAJOR MODEL CONSTRAINTS The length of a pumping run depends on the batch size & the pumping rate vbmin * Li ≤ Qi ≤ vbmax * Li

∀i ∈ I new

The size of a flowing batch changes during a batch injection due to the execution of stripping operations Depot D1

200

At time C4 200 DB4,D1

C5-L5 _ C5

100

(B5)

= 40

160 (B4)

100 (B5) 100

200 – 40 = 160

Wi ( i' ) = Wi ( i' −1 ) − ∑ Di , j ( i' )

At time C5 260

∀i ∈ I ,∀i' ∈ I new ,i' > i

j∈J

28

MAJOR MODEL CONSTRAINTS The overall amount of products delivered to terminals through stripping operations is equal to the size of the new batch injected in the line Depot D1

Depot D3

190

180

200

B4

B3

B2

B1

At time C4

20

50

30 50

200

C5-L5 _ C5 150

Depot D2

150

150

160

130

180

B5

B4

B3

B2

B1

At time C5

150 (in) = 50 + 30 + 50 + 20 (out)

A single time period will contain the completion time of a pumping run 0h

23 h B1

70 h

35 h B2

T1

93 h

65 h

48 h wB1,T1 = 1

72 h wB2,T2 = 1

125 h

BATCH INJECTIONS

B4 T3

T2

0h

102 h

B3 T4 96 h wB3,T3 = 1

144 h

TIME PERIODS

wB4,T4 = 1

29

MAJOR MODEL CONSTRAINTS Feasibility conditions for stripping operations - An upper bound on the cut size - The flowing batch has reached or will reach the depot during the pumping run Depot D1

Depot D2 350

160 100

300

100

B5

B4

B3

200 available for D2

10

100

10

B5 50

B6

50

60

B5

190

250 B6

CUT 1

290

100

B4

B3

CUT 2 240

100

B4

CONSECUTIVE STRIPPING OPERATIONS DURING INJECTION OF B6

B3 190

B6

100

CURRENT PIPELINE STATE

50 already gone 10

50 reserved for D1

At time C5

100 B5

CUT 3 50

100

B4

B3

At time C6

30

THE OBJECTIVE FUNCTION Delivery time constraints Batch injections completed up to period t are available to meet product requirements to be delivered to terminals before the end of period t ⎛t ⎞ ( l) DM ≥ ∑ p, j ⎜ ∑demp, j,k * (wi,t − wi+1,t ) ⎟ − Bp, j,t + Bp, j,(t−1) l=1 ⎠ ⎝ k =1 i

l∈I new

∀p ∈ P, j ∈ J p , t ∈T , i ∈ I new

OBJECTIVE FUNCTION

Minimize pumping cost, interface reprocessing cost, pipeline idle time and inventory carrying cost Min z =

⎛

∑ ∑ ⎜⎝ cp p ∈ P j∈ J

+

∑ ∑ cf

p '∈ P i∈ I p ' ≠ p i >1

p, p'

p, j

*∑

∑ DP

i∈ I i '∈ Inew

WIF i , p , p ' +

( i ') p ,i , j

⎞ ⎟+ ρ H ⎠

∑ ∑ ∑ cb p ∈ P j ∈ J t ∈T

⎛ ⎞ + cu ⎜ h max − PH max − ∑ L i ⎟ i∈ Inew ⎝ ⎠ ⎡ 1 ⎛ + ⎢ ∑ cid p , j * ⎜ ∑ ID ∑ new card ( I ) p ∈ P ⎣ j∈ Jp ⎝ i '∈ Inew

(t ) p, j

( i ') p, j

* B p , j ,t

⎞ ⎛ ⎟ + cir p * ⎜ ∑ IRS ⎠ ⎝ i '∈ Inew

( i ') p

⎞⎤ ⎟⎥ ⎠⎦ 31

A REAL-WORLD PIPELINE PLANNING EXAMPLE PROBLEM DATA A pipeline system with a single entry point and multiple exit points (5 terminals)

Four different products (gasoline, diesel, LPG, jet fuel) are sent to terminals Time horizon length: 4 weekly periods (672 h) Unidirectional flow Pipeline Length: 955 km Variable Segment Diameter: 12 – 20 in Pump rate range: 800 – 1200 m3 per hour

32

OPTIMAL STATIC PLANNING

ASSUMING A FIXED PLANNING HORIZON FIVE BATCH INJECTIONS D1

D2

D5

D4

D3

R

1220

55.00_198.33

50 135

INITIAL STATE INJECTING P4

247.5 140 70 70

310

400

10 80

425

200 135 200 STRIPPING OPERATIONS 190 70 550 10 152.5 410 120

5.00_52.00 425

700 60

400 90

0

190

Run Time Interval [h]

INJECTING P2

415

1720 962.5

202.33_309.21

672.5

400.37

358.28_412.07

679.63

0

200

400

INJECTING P3

555

1180

524.50_672.00

INJECTING P1

320.37

600

800

1000

P2

P3

P4

1200

1400

134.63

INJECTING P1 A VERY LARGE BATCH

1600 2

IDLE TIME P1

Volume [10 m3]

THE HORIZON-TIME EFFECT

33

DYNAMIC PIPELINE PLANNING TASK As time goes on, new transport requests are received and others are cancelled The current pipeline schedule should be periodically updated at the start of a new period A sufficiently long rolling time horizon should be considered Periodical planning update permits to eliminate the horizon-end time effect and, more important, the pipeline idle time The horizon-time effect arises because later batch injections have the only purpose of pushing batches to their destinations As the planning horizon rolls, such later batches will be injected because of new real shipment requests 34

DYNAMIC PIPE PLANNING ALGORITHM

Initialization Stage

-

Set Set Set Set Set Set

h (time period length) [hours] N (number of time periods to be considered) sf = ⎜TSF ⎜ (soft-frozen time periods) hf = ⎜THF ⎜ (hard-frozen time periods) k = 1. ddk-1 = 0. clock = 0 [h]. Run clock

Trigger Stage

INPUTS

n

clock = ddk-1 ?

y - Capture pipeline batch scenario (products, volumes and locations) (Ioldp, Woi, Foi) - Capture product inventories at refinery and depot tank farms (IRop, IDop,j) Data Updating Stage

- Import updated refinery production schedule and product output rates for periods k to k+N-1 (time horizon [ddk-1 ; dd(k+N-1)])

SCADA Remote Pipeline Controlling System

Refinery Production Schedule

Demand Updating Process Update Product Demand periods k to k+N-1

Data

for

OUTPUTS Updating the Pipeline Schedule

Rescheduling Stage

Dispatching Stage

Multiproduct Pipeline Run the Scheduling Optimization System (MPSOS) for the planning horizon including periods k to k+N-1

- Execute the Pipeline Schedule for the time horizon going from ddk -1to dd(k+hf-1) (periods k to k+hf-1)

Definite Pipe Schedule for periods k to k+hf-1 Definite Pipe Sequence for periods k+hf to k+hf+sf-1 Planned Pipe Schedule for periods k+hf to k+N-1

- Set k = k+tRS

35

OPTIMAL DYNAMIC PIPELINE PLANNING ASSUMING A 4-WEEK ROLLING PLANNING HORIZON D2

D1

D3

D4

D5

TEN BATCH INJECTIONS

R

425

550

INITIAL STATE

50 135

190

1220

70

415 120

55.00_168.00 1356

135

10 152.5 410 120

400 136

5.00_52.00 425

200

247.5 140 70 70

200

10 80

700 60

400

0

90

Run Time Interval [h]

SHORT IDLE TIME

0

107.5

65

129.62 350 120

247.72

213.66 80 6.34

254

49.04 130 9.62

1155.96

290.38

400

220

120

90

190

1446.34

200

327.5

120

247.72

402.28

259.62 3.62 400

504.00_630.25 1513.96 635.25_659.44 290.38

280 105.38

507.66

449.04

247.72

107.71

280

44.41 259.62

477.28

400

659.78

92.13

70.22 120

42.66 70 247.72 80 6.34

385.50_440.95 665.37

466.58_504.00 449.04

42.87

149.78

647.72 5.59 110

390

441.95_463.58 259.62

547.5

42.87

120 276.91

924.63

40

338.00_384.00 390

120

90 180 184.00_286.89 1234.63

295 247.5 160

1220 100 245

120

173.00_183.00 120

36.38

600

49.04 130 9.62

800

1000

P3

P4

1200

1400

1600 2

P1

P2

3

Volume [10 m ]

36

ADDITIONAL RESULTS Pipeline Usage

Qi

425

1356

120

1235

390

665 260

1963

290

[102 m3]

0

168

P1

P2

336

P3

P4

504

672

Idle time

Changeover

Time [h]

Refinery Inventory Profiles

3

Inventory Level [m ]

2500

2000

1500

1000

500

0 0

168

P1

336

P2

504

P3

P4

672

Time [h]

37

MULTIPLE-SOURCE TRUNK PIPELINES So far, we deal with single-source multiple-destination trunk pipelines Multiple-source pipelines include additional input terminals at non-origin points to collect oil product batches from downstream refineries INTERMEDIATE INPUT TERMINAL

s1

j1 B6

B4

j3

B2

B5 Al final de K1

j2

s2

B3

B1

xdB2,j2(K1) = xdB1,j2(K1) = 1 xdB1,j3(K1) = 1

B3

B6

B4 B5

B3 (K1) B3,s2

xs

B2

INJECTION OF BATCH B3

B1

=1 P1

P2

P3

P4

Need of choosing the input terminal where the next pumping run will occur At intermediate input terminals, a new batch can be injected or the size of a flowing batch can be increased 38

MULTIPLE-SOURCE TRUNK PIPELINES In multiple-source trunk pipelines, batches are not sequenced in the same order that they were injected in the line INTERMEDIATE INPUT TERMINAL

s1

j1 B6

B4

j3

B2

B5 Al final de K1

j2

s2

B3

B1

xdB2,j2(K1) = xdB1,j2(K1) = 1 xdB1,j3(K1) = 1

B3

B6

B4 B5

B3 (K1) B3,s2

xs

B2

B1

=1 P1

P2

P3

P4

A batch is not necessarily preceded by those previously pumped in the line Batch B4 is preceded by batch B3 even though B4 was inserted before Need of separately handling the pumping run sequence and the batch sequence 39

DETAILED PIPELINE SCHEDULE

Just the batch injections and stripping operations planned for the first period of the current time horizon are to be performed At the very operational level, a detailed pipeline schedule for the action period of the current horizon must be prepared A more detailed definition of the stripping operations to execute during a batch injection is required: sequence, timing and extent of stripping operations The basic information is provided by the monthly pipeline planning Additional systematic heuristic/algorithmic procedures providing a detailed description of the required stripping operations are to be applied

40

DETAILED PIPELINE SCHEDULE Nearest Active Depot First (NDF) rule:

PRIORITIZE DELIVERIES MDS

D2

D1

TO THE NEAREST DEPOT

D3

D4

D5

R

200

200

650

5.00_70.00 650 0 Mean Flow Rate = 10.00

400

200

400

650

135 150

700 300

400 200

Run Time Interval 0 [h]

200

50 135

AT THE PLANNING LEVEL

1500 1635

900

2

Volume [10 m3]

In which order the “stripping operations” should be executed during a batch R injection?

50.00_55.00 50 55.00_70.00 150 0

200

200

135

600

200

200

135

400

250

450

200

200

135

AT THE OPERATIONAL LEVEL

450

200

450

200

200

135

200

500

400

200

200

135 150

30.00_50.00 200

700

400

5.00_15.00 100 100 15.00_30.00 150

D5

D4

100

400

D3

150

Mean Flow Rate

DPS D2

D1

50

=

Nominations Q dd[h] N4 150 12

200

Flow Rates

Run Time Interval 0 [h]

Nominations Q dd[h] N2 100 18 N3 200 72

Nominations Q dd[h] N1 200 48

650 400

200 650

400 900

200

50 135

41

1500 1635 2

Volume [10 m3]

DETAILED PIPELINE SCHEDULE MILP Formulation: D2

D1

D3

D5

D4

R

Run Time Interval 0 Flow Rates [h]

11.42

30.50_48.00 200

9.09

48.00_70.00 200 0

200

135

700

50 135

600

200

50 135

600

200

50 135

200

50 135

100

200

400

250

450

200 200

18.00_30.50 100

200

150 400

5.00_18.00 150 150

8.00

700

200

11.53

400

200

650 400

650

400 900

1500 1635 2

Volume [10 m3]

Comparative Results:

Rule Valve operations NDF NEAREST DEPOT 5 FDF FARTHEST DEPOT 4 EDD EARLIEST DUE DATE 4 MILP 4

Earliness [h] 39 22 2 2

Tardiness [h] 43 4 2 0 42

A MULTIPLE-SOURCE PIPELINE SCHEDULE Depot 1

Depot 2

Demand

30

30

Depot 3

30

50

Batch B3 will be pumped in Refinery 2

B3 B5

30

70

Refinery 1

B4

Supply

B2

B1

40 Refinery 2 Horizon Length: 120 hs.

43

A MULTIPLE-SOURCE PIPELINE SCHEDULE Deliver Product P1 (Batch B5) to Depot 1

B3 B5

B4

B2

B1

Inject Product P1 (Batch B5) in Refinery 1

44


B3 B6

B4

B2

B1


45


B3 B6

B4

B2

B1


46


Refinery 2 is ready to inject Product P3 in Batch B3

B6

B4

B3

B1


47


B6

B4

B3

B1


48


B6

B4

B1


49

A MULTIPLE-SOURCE PIPELINE SCHEDULE Deliver Product P2 (Batch B1) to Depot 3 Note that Batch B7 has been preserved to be injected in Refinery 2

B7 B8

B6

B4


50

A MULTIPLE-SOURCE PIPELINE SCHEDULE Deliver Product P2 (Batches B4 & B6) to Depot 3

B7 B9

B8

B6


51


Refinery 2 is ready to inject Product P3 in Batch B7

B9

B8

B7

B6


52

CONCLUSIONS

Multiproduct pipeline planning is a very complex industrial problem A continuous pipeline planning approach has been presented Pipeline planning over a multiperiod rolling horizon with delivery due dates at period ends is performed The approach still remains competitive for a monthly time horizon The approach can even be applied to multi-source multiproduct pipelines Tools for generating a weekly detailed pipeline schedule have also been briefly described

53

OIL PIPELINE LOGISTICS Jaime Cerdá Instituto de Desarrollo Tecnológico para la Industria Química Universidad Nacional de Litoral - CONICET Güemes 3450 - 3000 Santa Fe - Argentina

Thanks for your attention! Questions?

Contact: [email protected]

54

OIL PIPELINE LOGISTICS - Carnegie Mellon University

Recommend Documents