Scheduling, Logistics, Planning, and Supply Chain Management for Oil Refineries

Transcription

1 Scheduling, Logistics, Planning, and Supply Chain Management for Oil Refineries José M. Pinto Feb 7, 28

2 OUTLINE Introduction Scheduling models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

3 Economical Targets MOTIVATION Profitability Maximization (Pelham and Pharris, 1996; Ramage, 1998) Minimization of Operational Costs (Bodington and Shobrys, 1996) Maximization of integrated margins (Thijssen and Lasschuit, 23) Maximization of supply chain value (Chopra and Meindl, 23) Decision Support Tools

4 ADVANCES Availability of more powerful and less expensive computers; Mathematical Developments: Time representation; (Moro and Pinto, 1998) Combinatorics in MIP; (Raman and Grossmann, 1994) Non-convexities in MINLP; (Viswanathan and Grossmann, 199) Consequences for the Petroleum and Chemical Industry: (Ramage, 1998) Unit Level Optimization Plant-wide Optimization Supply Chain Optimization 198 s 199 s 2 s

5 ROADMAP OIL OPERATIONS Planning of oil exploration (Carvalho) Oil supply scheduling (Más) Refinery Scheduling Distillation (Smania) Fuel Oil / Asphalt area (Joly) LPG scheduling (Moro) Utility systems (Micheletto) Paraffins (Casas-Liza) Pipeline scheduling (Rejowski Jr., Hassimotto) Refinery planning (Moro) Pipeline network planning (Assis) Supply chain management (Neiro, Chen)

6 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

7 SHORT TERM CRUDE OIL SCHEDULING Crude Oil System

8 OBJECTIVES Maximize operating profit revenue provided by oil processing cost of operating the tanks Generate a schedule for crude oil operations receiving oil from pipeline waiting for brine settling feeding the distillation units

9 TIME SLOT REPRESENTATION

10 MILP OPTIMIZATION MODEL Max subject to: total operating profit Timing constraints Pipeline material balance equations Pipeline operating rules Pipeline always connected to a tank Material balance equations for the tanks Volumetric equations Component volumetric balance Tank operating rules Minimum settling time Rules for feeding the distillation unit

11 DECISION VARIABLES slot k Yp j,k Yd f,j,k fraction f

12 REAL-WORLD EXAMPLE Oil parcel Volume Start time End Time Composition (m 3 ) (h) (h) 1 6, 8 2 1% Bonito 2 5, % Marlin 3 1, % Marlin 4 6, % RGN Tank initial conditions Distillation target flowrate = 15 m 3 /h

13 RESULTS

14 MODEL SOLUTION GAMS / OSL CPU time 2.8 hrs (Pentium II 266 MHz 128 MB RAM) Variable size time slot model 912 discrete variables 3237 continuous variables 5599 equations Fixed size time slot model 2154 discrete variables!

15 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

16 FUEL OIL/ASPHALT PRODUCTION SCHEDULING PROBLEM The plant produces 8% of Brazilian fuel oil; The plant has significant storage limitations; Complex distribution operations;

17 Product Base Diluent used FO1 FO2 FO3 FO4 UVO1 UVO2 CAP7 CAP2 RASF RASF RASF RASF RASF RASF RASF RASF 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

18 MATHEMATICAL MODELS Uniform Discretization of Scheduling Horizon; Objective Function: Minimize Operational Cost. First Approach: non-convex MINLP (viscosity constraints); Linear Transformation Second Approach: MILP;

19 MINLP MODEL Minimize: Subject to: COST = Raw-Material Costs + Inventory Costs + + Pumping Costs + Transition Costs Material Balance Constraints Demand Supply of Plant Products Plant Operating Rules at each t, the plant production must be stored in one single tank simultaneous tank loading and unloading is not allowed (exception: HG storage tank) while asphalt is produced, the OCC stream from UFCC must be directed to storage in TK-4228 UVO / Asphalt may be sent to truck terminals only between 6: a.m. and 6: p.m. while asphalt is produced, the RASF diluent must be HG Material Flow Constraints flowrates to oil-pipelines must obey pump limitations flowrates to truck terminals must obey pump limitations Viscosity Specification Constraints

20 TK-4331 REAL-WORLD EXAMPLE instance evaluated Scheduling horizon: 3 days Time span: 2 hours Nominal production: 2, m 3 /month START PRODUCTION SCHEDULE AND STORAGE INFORMATION 1 END TK TK-4337 TK-4331 TK-4331 TK-4337 TK-4331 TK-4332 TK-4418 TK-4332 TK-4418 TK-4331 TK-4418 TK-4331 TK-4418 TK-4332 TK-4337 TK-4331 TK TK-4331 TK-4331 TK-4337 TK-4337 TK-4331 TK-4418 TK-4418 TK-4337 TK-4331 TK-4331 TK-4331 TK-4333 TK-4332 TK-4337 TK-4331 TK-4331 TK-4331 UVO1 UVO2 CAP7 CAP2 FO1 FO2 FO3 FO4

21 Each interval = 2 m 3 /h diluent from TK (HG) diluent from TK-4228 (OCC+LCO) pure OCC from UFCC pure LCO from UFCC Schedule of diluents in the mixer FO1 (p=1) TK-4331 (i=1) TK-4332 (i=2) UVO1 (v=1) TK (q=1) TK (q=2) FO2 (p=2) TK-4333 (i=3) TK-4334 (i=4) 1 36 UVO2 (v=2) TK (q=3) TK (q=4) FO3 (p=3) TK-4335 (i=5) TK-4336 (i=6) 1 36 CAP-7 (v=3) TK-4418 (q=5) 1 36 Volume (x 1-3 m 3 ) in product storage tanks FO4 (p=4) TK-4337(i=7) 1 36 CAP-2 (v=4) TK-4411 (q=6) TK (q=7) TK (q=8) 1 36

22 (m 3 /h) LOCAL OIL-PIPELINE FO1 4 (WITH TRANSITION MODELING) FO2 3 FO3 2 FO4 number of -1 variables number of constraints number of continuous variables (m 3 /h) OIL-PIPELINE TO SÃO PAULO FO1 4 (WITH TRANSITION MODELING) FO2 3 FO3 2 FO MINLP model number of -1 variables number of constraints number of continuous variables MILP model MILP models WITHOUT TRANS. CONSTRAINTS WITH TRANS. CONSTRAINTS case MIP model nodes iterations CPU time (s) objective MILP A MINLP MILP B MINLP MILP C MINLP MILP D MINLP

23 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

24 CRUDE OIL SUPPLY PROBLEM crude oil terminals oil refineries

25 MOTIVATION Increasing utilization of the system Larger demand for crude oil in refineries Outsourcing of transportation Potential economic impact No systematic scheduling Operations involve high costs and aggregated values Petrobras distribution complex 4 integrated refineries

26 PETROBRAS DISTRIBUTION COMPLEX

27 PROBLEM SPECIFICATION idetermined by the petroleum origin iapproximately 42 types of crude oil may be processed Types of crude oil

28 PROBLEM SPECIFICATION isets of crude oil types with similar properties inecessary due to limited amount of tanks Classes of crude oil Types of crude oil i7 classes

29 PROBLEM SPECIFICATION itransport types of crude oil Tankers ioverstay incurs in additional costs - US$ 1 k to US$ 2 k per day Classes of crude oil Types of crude oil

30 PROBLEM SPECIFICATION Piers idifferent capacities Tankers Classes of crude oil Types of crude oil

31 PROBLEM SPECIFICATION Piers Tanks istore classes of crude oil Tankers iminimum storage levels Classes of crude oil Types of crude oil isettling time between loading and unloading operations

32 PROBLEM SPECIFICATION Piers Tanks Tankers Pipelines iflow rate at each pipeline limited by the density of the heaviest crude oil class Classes of crude oil Types of crude oil ipossible to connect to at most one tank at every time

33 PROBLEM SPECIFICATION Piers Tanks Tankers Pipelines ibuffer operations between terminal and refineries Classes of crude oil Types of crude oil Substations istore difference in flow rate between inlet and outlet pipelines

34 Full scale model unsolvable PROPOSED STRATEGY Pipelines Terminal Decomposition of the problem in three formulations Substations Refineries

35 MATHEMATICAL FORMULATION MILP model formulation Time representation Continuous Based on events Inventory level (cont. variable) Amount generated (cont. variable) Decision to produce (disc. variable) V i V i+1 V i+2 V i+3 Q i Q i+1 Q i+2 Q i+3 X i X i+1 X i+2 X i+3 time Time events (cont. variable) T i T i+1 T i+2 T i+3

36 PROPOSED MODEL - VARIABLES Binary variables Decisions Assignment of ship n to pier p: Unloading of ship n to tank t: Unloading of tank t to oil pipeline o: A n, p LT n, t, e UT t, o, e Continuous variables Timing Inventory Flowrates Operating profit

37 PORT MODEL - CONSTRAINTS Decisions Assignment of tanker n: Operation of tank t: Operation of tanker n: Operation of oil pipeline o: p P n n N t T n A n t, p =1 LT + n UT LT n t T o, t, e t, o, e o O, t, e UT t 1, o, e t 1 1

38 PROPOSED MODEL - TIMING Ships, tanks and pipelines Timing variables in each time event Initial Final Matching of the timing variables Unloading from ship n to tank t s = n, e s t e TN TT, f = n, e Unloading from tank t to pipeline o s = t, e s o e TT TD, f t e TN TT, f = t, e f o e TT TD,

39 PROPOSED MODEL CONSTRAINTS Matching of timing variables Ships Tanks TN TN s s H.( 1 LTn, t, e) TTt, e TNn, e + H.(1 LTn, t, s n, e e f f H.( 1 LTn, t, e) TTt, e TNn, e + H.(1 LTn, t, f n, e e ) ) Tanks Pipelines TT TT s s H.( 1 UTt, o, e) TDo, e TTt, e + H.(1 UTt, o, s t, e e f f H.( 1 UTt, o, e) TDo, e TTt, e + H.(1 UTt, o, f t, e e ) )

40 PROPOSED MODEL OBJECTIVE FUNCTION max profit = E 1 class REVRcl, r. Qut r cl CLRr o O r t ( Tcl To) e' = 1 + cl REVP class cl. t T cl T ( V V ) t, E t t, o, e' (oil revenue to the refineries) (final - initial oil revenue in the port) c p.. n, c end start ( τ τ ) n, p (oil cost in the tanks) (pier utilization cost) COSTn. Tn (overstay cost of the oil tankers) n E 1 face COST. INT (interface cost) o COST COST cl CLO crude c pier p o se se n N cl CLO cl cl n N o p C c cl, cl n, p e= 1 cl, cl, o, e

41 REAL-WORLD PROBLEM Problem 4 Substation Model SEBAT RECAP OSBAT IV Capuava RPBC REPLAN OSVAT IV SEGUA Paulínia OSVAT II OSvAT III REVAP Problem 2 Guararema OSBAT III OSBAT II OSVAT I São José dos Campos Substation Model Problem 3 Cubatão Substation Model GEBAST P3 P1 P4 P2 São Sebastião Problem 1 Port Model

42 COMPUTATIONAL RESULTS Smaller optimality gaps for the Port Model Large variation on computational times Problem 1 Problem 2 Problem 3 Problem 4 Number of continuous variables Number of binary variables Number of constraints Relaxed LP solution 21, Best Integer Objective 2, Optimality gap 7.78% 82.61% 9.91% 31.74% Nodes Iterations CPU time (Pentium III 45MHz) 1, s 62.7 s s s Port Model Substation Models

43 PROBLEM 1 TANKERS AND TANKS PEDREIRAS FRONT BREA MURIAÉ REBOUÇAS VERGINA II NORTH STAR PRESIDENTE CANTAGALO a a 1 A b b TBN1 TBN2 RAVEN TBN3 TBN TQ328 TQ B Vmax 28 Vmax 1 C D B E Vmin 7 Vmin TQ3214 Vmax Vmin TQ3215 Vmax Vmin TQ3217 Vmax Vmin TQ3218 Vmax Vmin TQ3219 Vmax Vmin TQ3233 Vmax Vmin TQ3234 Vmax Vmin TQ3235 Vmax Vmin TQ3237 Vmax Vmin TQ3238 Vmax Vmin TQ3239 Vmax Vmin TQ324 Vmax Vmin TQ3241 Vmax Vmin TQ3242 Vmax Vmin TQ3243 Vmax Vmin TQ Vmax 4 2 Vmin

44 PROBLEM 1 GANTT CHART OSBATII/S SEGUA/E OSBAT OSVAT TQ3244 TQ3243 TQ3242 TQ3241 TQ324 TQ3239 TQ3238 TQ3237 TQ3235 TQ3234 TQ3233 TQ (P4) (P2) 5.3Hrs (P2) 26. Hrs (P2) Hrs (P2) 24 Hrs 6. (P4) 24 Hrs (P2) (P2) (P1) 46.4 Hrs (P4) 55.7 (P1) 29.4 (P4) 24 Hrs (P1) 24 Hrs (P2) 28.9 (P3) 3.7 (P4) 51.1 (P3) 11.9 (P4).6 (P4) 6. (P4) 3L (P-4) (P4) Hrs (P4) TQ (P4) TQ3217 TQ3215 TQ3214 TQ TQ328 P-4 CANTAGALO P-3 P-2 P-1 24 Hrs FRONT_BREA (P4) 37.8 (P2) REBOUCAS PEDREIRAS 43.2 Hrs MURIAE 12.2 (P2) TBN TBN2 VERGINA II 5.3 (P3) 24.7 (P3) RAVEN PRESIDENTE NORTH_STAR TBN (P4) TBN Tempo (horas)

45 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

46 SCHEDULING OF A MULTIPRODUCT PIPELINE SYSTEM Pipelines transport large amounts of products in the fastest and safest mode. Development of a systematic approach for such operation. amounts of products sent to all depots and operational sequencing at the refinery, pipeline and at depots; inventory levels for each product at all locations. Oil company that operates with a refinery and depots in several locations.

47 DISTRIBUTION SYSTEM Uberlândia Uberaba Brasília Consumer Market... 1 p P-1 P Consumer Market... 1 p P-1 P Consumer Market... 1 p P-1 P Consumer Market... Goiânia Consumer Market... Rib. Preto 1 p P-1 P 1 p P-1 P Petrobras OSBRA complex.. 1 p P-1 P REPLAN

48 PROBLEM DESCRIPTION Distribution problem a refinery must distribute P products through a single pipeline; Set of D depots which are connected to consumer markets. Subject to the following constraints upper and lower bounds for inventory levels for each product at all locations; flow rate bounds for the entire system; arrival times for products sent by the refinery to their destination; demands for all products established by each local consumer market

49 PIPELINE MODELING Segment d Interface between two different products VOD p,d,k VOT p,d,k V p,d,l,k Pack 1 Pack 2 Pack l Pack L-1 Pack L VOT p,d+1,k Feed (VOT p,d,k ) Displacement of products; Product contained in L; Sent to depot d (VOD p,d,k ); Sent to segment d+1 (VOT p,d+1,k ) No Feed VOT p,d,k = Products remain in the same pack;

50 SYSTEM WITH D SEGMENTS Consumer Market 1 Consumer Market d Consumer Market D Refinaria Refinery Depot 1 Depot d Depot D Segmento Segmento dd Segmento D Inventory Costs - Refinery Constraints at Depots VD, p, d, k = VDZERO P pk, d + VODp, d, k VOMP p, ddk K p,d, k=1 C = CERp, k VRp, k + CEDp, d, k VDp, d, k δ p, d, k VDp, d, k = + = VODp, d, k VOMp, d = k p 1k 1 p 1d= 1k= 1 P D K P P D K p, d, k VDp, d, k VDMAXp, d k p, d, k p, d, k p, d, k p, p' VD =, p, d, k=2,,k VDMIN, + CP VOD p= 1d= 1k= 1 p= 1p' = 1d= 1k= 1 + CONTACT Inventory Costs - Depots TY p, p', d, k Pumping Costs Interface Costs

51 REAL-WORLD DISTRIBUTION SYSTEM Uberlândia Uberaba Brasília Consumer Market... 1 p P-1 P Consumer Market... 1 p P-1 P Consumer Market... 1 p P-1 P Consumer Market... Goiânia Consumer Market... Rib. Preto 1 p P-1 P 1 p P-1 P.. 1 p P-1 REPLAN P

52 COMPUTATIONAL RESULTS OSBRA Example Relaxed Solution [$ x 1-2 ] 27, Solution [$ x 1-2 ] 3, CPU Time [s ] 1,775 Nodes Visited 5,7 Continuous Variables 6,316 Binary Variables 42 Equations 9,393

53 PIPELINE k=15 k=14 k=13 k=12 k=11 k=1 k=9 k=8 k=7 k=6 k=5 k=4 k=3 k=2 k=1 k= Segment REPLAN-Rib.Preto [x 1-2 m 3 ] 4 Segment Rib.Preto-Uberaba [x 1-2 m 3 ] 25 Segment Uberaba-Uberlândia [x 1-2 m 3 ] 25 k=15 k=14 k=13 k=12 k=11 k=1 k=9 k=8 k=7 k=6 k=5 k=4 k=3 GASOLINE DIESEL LPG JET FUEL. k=2 k=1 k= Segment Uberlândia-Goiânia [x 1-2 m 3 ] 6 Segment Goiânia-DF [x 1-2 m 3 ] 135

54 7 STORAGE LEVELS Diesel Oil Inventory Level [m 3 ] Time [h] LPG Gasoline Time [h] Aviation Fuel Time [h] Inventory Level [m 3 ] Inventory Level [m 3 ] Time [h] Rib. Preto Uberaba Uberlândia Goiânia Brasília REF Inventory Level [m 3 ]

55 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

56 PLANNING MODEL FOR REFINERIES To develop a general representation for refinery units streams with multiple inputs and destinations nonlinear mixing and process equations bounds on unit variables To apply to the production planning of a real world refinery diesel production to satisfy multiple specifications extension to supply chain systems

57

58 TYPICAL PROCESS UNIT

59 - Feed flowrate: = UNIT EQUATIONS Q uf, Q u ', s, u u Uu s Su, u - Feed Properties: P u,f,j = f j ( Q u,s,u, Pu,s,j ) u U u, s S u,u, j J s - Total flowrate of each product stream: Q u,s = f ( Q u,f, P u,f,j, V u ) - Unit product stream properties: P u,s,j = f j ( P u,f,j, V u ) - Product streams flowrates (splitter): Q = Qusu us,,, u Us,u j J F, s S U j J s, s S U s S U

60 REAL-WORLD APPLICATION Planning of diesel production Petrobras RPBC refinery in Cubatão (SP, Brazil). Three types of diesel oil: Metropolitan Diesel. Low sulfur levels Metropolitan areas Regular Diesel. Higher sulfur levels Maritime Diesel. High flashing point.

61 DIESEL SPECIFICATIONS Property DENSITY min / max FLASH POINT min ( C) ASTM 5% min / max ( C) ASTM 85% max ( C) REGULAR.82/.88 DIESEL METROPOLITAN.82/.88 MARITIME.82/ / / / CETANE NUMBER min SULFUR CONTENT max (% WEIGHT)

62 MAIN RESULTS Potential Improvement US$ 23, / day or US$ 8,, / yr Implemented with on-line data acquisition

63 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

64 GENERAL PETROLEUM SUPPLY CHAIN

65 OBJECTIVE Development of an optimization model that is able to represent a petroleum supply chain to support the decision making planning process of supply, production and distribution

66 REFINERY - PROCESSING UNIT MODEL

67 SUPPLY, DISTRIBUTION STORAGE MODEL

68 SUPPLY CHAIN MODEL Max Profit = Revenues Crude Oil Costs Operating Costs Inventory Costs Transition Costs subject to: processing units tank pipeline units that compose refinery topology refineries that compose the supply chain petroleum and product tanks that compose refineries petroleum and product tanks that compose terminals refineries and terminals that compose the supply chain pipeline network for petroleum supply pipeline network for product distribution Large Scale MINLP

69 INTEGRATED SUPPLY CHAIN

70 SUPPLY CHAIN EXAMPLE GENERAL CONSTRAINTS: Planning horizon: one / two time periods Supply of 2 oil types Generation of 32 products (6 transported with pipelines) SCENARIOS: 1: Base Case model 2: Pre-selection of some crude oil supplies 3: Interruption of pipeline segment SG-RV

71 SUPPLY CHAIN EXAMPLE REVAP

72 SUPPLY CHAIN EXAMPLE RPBC

73 SUPPLY CHAIN EXAMPLE OIL SUPPLY

74 SUPPLY CHAIN EXAMPLE PRODUCT TRANSFER

75 COMPUTATIONAL RESULTS Case Case 1 Case 2 Case 3 Number of time periods Constraints Variables Discrete variables Solution time (CPU s) Objective Value ($ x1 6 )

76 DECOMPOSITION STRATEGIES Strategy Primal subproblem Dual subproblems Multipliers update 1 Fixed assigment Lagrangean Subgradient 2 Fixed inventory Lagrangean Subgradient 3 Fixed inventory Surrogate Subgradient 4 Fixed inventory Lagrangean Modified Subgradient Problem RMP Strategy 1 Strategy 2 Strategy3 Strategy4 CPU seconds Number of time periods.

77 OUTLINE Introduction Scheduling Models crude oil scheduling fuel oil / asphalt scheduling Logistics oil supply model pipeline distribution Planning Models refinery diesel production Supply Chain Management Models Conclusions

78 CONCLUSIONS Problems can be modeled as large scale MILPs / non-convex MINLP flexibility in representing general topologies complex logical decisions and operating rules can be modeled representation of realistic financial objectives Modeling Issues time representation blending/pooling transitions The LP based Branch and Bound Method is satisfactory to generate good feasible solutions no guarantee of global optimum solutions for all instances The OA/ER/AP Method is efficient to circumvent the non-convexity problem is satisfactory to generate feasible solutions has computational performance similar to MILP model

79 CHALLENGES Main theoretical difficulties Complex problems with high combinatorial features; NP-Complete Problems Exponential computational times Main practical difficulties The understanding of the problem itself can constitute the major difficulty Multiple systems and interfaces The cooperation between the modeler and the practitioner is essential and remains as a major challenge Continuous update necessary due to the dynamic nature of problems.

80 RESEARCH NEEDS Modeling Horizontal Integration (e.g. Upstream-Downstream-Final customer) Vertical Integration (e.g. planning and scheduling operations) Multi country supply chains (royalties, tariffs) Modeling of uncertainties Inventory design and management Demand planning models (including forecasting) Efficient solution methods Decomposition (spatial, temporal, functional) Techniques (Lagrangean Relaxation, Cross Decomposition, Metaheuristics, Hybrid Methods)

81 ACKNOWLEDGMENTS RESEARCHERS M. Joly R. Más L. Moro R. Rejowski P. Smania F o. S. Neiro M. C. A. Carvalho M. K. Hassimotto FINANCIAL SUPPORT CENPES ENGINEERS C.A. Gratti M. F. Lehner M.V. Magalhães A.C. Zanin E. Almeida Neto