Quantitative Methods for Strategic and Investment …egon.cheme.cmu.edu/ewocp/docs/Petrobras_Brenno_Menezes_IG.pdfDelayed Coker . Terminal/Pipeline Atmospheric . Distillation . EWO
Post on 04-Apr-2018
215 Views
Preview:
Transcript
Goal: develop high-level decision-making optimization to predict structural modifications in refining and logistics assets using more rigorous formulations
PETROBRAS Current Tool for Strategic Planning (PLANINV) – LP
No Process Design Synthesis Quantitative Methods
PLANINV Process Design Opt. (MILP)
1
Delayed Coker
Terminal/Pipeline
Atmospheric Distillation
EWO Meeting – March 2014
What, Where, When to Invest?
Only optimize streams transfers (oil and fuels import/export, market supply)
+ NLP Processing
Blending
Quantitative Methods for Strategic and Investment Planning in the Oil-Refining Industry
Brenno C. Menezes, Ignacio E. Grossmann, Lincoln F. Moro and Jeffrey D. Kelly
𝐲𝐲𝐲𝐲=expansion of an existent unit
𝐲𝐲𝐲𝐲𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 𝐐𝐐𝐐𝐐𝐮𝐮𝐋𝐋 ≤ 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 ≤ 𝐲𝐲𝐲𝐲𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 𝐐𝐐𝐐𝐐𝐮𝐮𝐔𝐔
𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 = 𝐐𝐐𝐄𝐄𝐐𝐐𝐄𝐄𝐄𝐄𝐫𝐫,𝐮𝐮,𝐧𝐧 + 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭−𝟏𝟏 + 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭−𝟏𝟏 𝐲𝐲𝐞𝐞𝐞𝐞𝐞𝐞𝐧𝐧𝐞𝐞𝐞𝐞𝐞𝐞𝐧𝐧: 𝐮𝐮,𝐧𝐧 𝐲𝐲𝐞𝐞𝐞𝐞
𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 ≤ 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 𝐮𝐮,𝐧𝐧 𝐲𝐲𝐞𝐞𝐞𝐞
QF= operational flow Q𝐐𝐐= expanded capacity QC= total capacity
Capital Investment Planning Formulation
(R,U,N,T) R=Refinery U=Unit type N=Number of an unit type T=Time
Crude dieting
ISW
Processing
Blending
ON
QC=QCt-1+QNEW MILP
QF≤QC NLP
INVREF
OPREF
𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 = 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭−𝟏𝟏 + 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭−𝟏𝟏 𝐞𝐞𝐧𝐧𝐞𝐞𝐭𝐭𝐞𝐞𝐢𝐢𝐢𝐢𝐞𝐞𝐭𝐭𝐞𝐞𝐞𝐞𝐧𝐧: 𝐮𝐮,𝐧𝐧 𝐞𝐞𝐧𝐧𝐞𝐞
𝐲𝐲𝐞𝐞𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 𝐐𝐐𝐐𝐐𝐮𝐮𝐋𝐋 ≤ 𝐐𝐐𝐐𝐐𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 ≤ 𝐲𝐲𝐞𝐞𝐫𝐫,𝐮𝐮,𝐧𝐧,𝐭𝐭 𝐐𝐐𝐐𝐐𝐮𝐮𝐔𝐔
Maximize: NPV = DemandSales - SupplyCosts - OperatingCosts - InvestmentCosts
Subject to:
Where:
Sahinidis et. al., CACE, 13, (1989) and Sahinidis & Grossmann, CACE, 15, (1991).
T1 T2
Take an investment decision (binary)
Count on the additional production
Project execution Improved Formulation: - Installations;
year
- Project Execution; - NLP Operational Layer.
Q𝐐𝐐= installed capacity
𝐲𝐲i= installation of a new unit
⋀ 𝐮𝐮,𝐧𝐧 𝐞𝐞𝐧𝐧𝐞𝐞
3 EWO Meeting – March 2014
Options to Formulate the Problem
(1) Menezes, et. al. (2014). Nonlinear Production Planning of Oil-Refinery Units for the Future Fuel Market in Brazil: Process Design Scenario-Based Model. Ind. Eng. Chem. Res. DOI: 10.1021/ie025775. Published online: Feb, 17, 2014.
1st- NLP Operational Problem Z=profit (m3/d) and QFu=unit throughputs to control capacity expansion
2nd- MINLP Investments Problem (NLP Operational Problem Embedded) Z=NPV ($) and QEu,t and QCu,t to control capacity expansion
3rd- MILP Investments Problem + NLP Operational Problem (Phenomenological Decomposition Heuristics) Z=NPV ($) and QEu,t, QIu,t and QCu,t to control capacity expansion and installation
Crude Diet
Processing
Blending - Crude - Cuts/Final Cuts - Final Products
NLP
Investment
Operational
MILP
QFu,t ≤ QCu,t link constraint
Full Space Problem MINLP
Aggregated Multi-Site Approach
Multi-Site Approach
4
Aggregated Multi-Site Approach
Refining Process Capacities (k m3/d) 2013 2016 2020Crude Distillation Unit CDU 310 372 536Vacuum Distillation Unit VDU 140 153 260Residue Fluid Catalytic Cracking RFCC 22 22 22Fluid Catalytic Cracking FCC 76 76 76Hydrocracking HCC 10 74Propane Deasphalting PDA 10 10 10Delayed Coker DC 42 74 124Light Cracked Naphtha Hydrotreater LCNHT 54 54 54Coker Light Naphtha Hydrotreater CLNHT 22 34 60Stabilizer ST 22 34 60Kerosene Hydrotreater KHT 15 15 15Diesel Hydrotreater (medium severity) D1HT 60 60 60Diesel Hydrotreater (high severity) D2HT 30 68 135Reformer REF 7 10 10EWO Meeting – March 2014
Conceptual Projects under reevaluation
5
Perform linear regression to convert the nonlinear Power Law to Fixed-Charge Relation: Investment Cost = a*Capacity+b*Setup
EWO Meeting – March 2014
Unit (u) αu βumi US$/1000m3 mi US$
CDU 11.7 227.1VDU 15.1 146.4FCC 57.5 241.2HCC 150.6 747.6RFCC 70.2 588.8DC 108.9 456.5KHT 21.8 115.9DHT 29.6 162.6LCNHT 14.0 69.3CLNHT 14.0 69.3ST 11.5 193.9REF 46.0 80.0
FK
FLD
ATR
CDU C1C2C3C4
SW2
VR
VDU
N
K
LD
HD
LCO
DO
HTD
HTK
FCC
D1HT
KHT
CLNCHN
CLGO
CHGO
CMGO
D2HT
DCREF
LCNHT
CLNHT
PQN
C1C2C3C4
HCNLCN
C1C2C3C4
FN
FHD
GLN(GLNC)
MSD
HSD
JET
LSD
HTCLN
HTLCN
FO
REFOR
C1C2 FG
LPGC3C4
LVGO
HVGO
00
ASPR
DAO
PDA
RFCC
SW3
SW1
C1C2C3C4
HCCO
Crude
HCCDHCCK
HCCN
HCC
USD
COKE
H2
COKE
LSDimp
GLNimp
(GLNA)
ETH
For RNEST
JETimp
LPGimp
ST
GOST
LNST
HNST
max NPV = �� 1− tr Ft0�1
(1 + ir)t0 ��(prp,t0−Refct0)Demp,tpt0t
−�prcr,t0υcr,tQFCDU,tcr
−� primp,t0QFimp,timp
−�ΥHT,tQFHT,tHT
�
−1
(1 + ir)tit�(αu,titQEu,t + βu,tityu,t)u t<Tend
�
12 discrete variables; 1127 continuous variables 1019 equations 4463 nonzero elements; 2552 nonlinear elements
Aggregated Multi-Site Approach
6
Aggregated Multi-Site Approach
Crude Diet
Processing
Blending - Crude - Cuts/Final Cuts - Final Products
NLP Operational
QFu,t ≤ M (=1000)
Crude Diet
Processing
Blending - Crude - Cuts/Final Cuts - Final Products
NLP
Investment
Operational
MILP
QFu,t ≤ QCu,t link constraint
Full Space Problem MINLP
Expansions, QE (k m3/d) 2016 Unit GLNC GLNCETH GLNC GLNCETH
372 CDU 544 508 400 400153 VDU 202 223 180 18222 RFCC 82 25 15 1576 FCC 52 52 72 6110 HCC 97 114 45 4574 DC 128 126 114 10515 KHT 18 15 15 1568 D2HT 122 116 65 6554 LCNHT 72 42 47 4034 CLNHT 72 72 54 4034 ST 72 72 54 4012 REF 33 13 24 18
Profit (mi US$/d) 38.992 33.376 21.432 13.420CPU (s) 0.733 0.795 0.468 0.546
Expansions, QE (k m3/d) 2016 Unit GLNC GLNCETH GLNC GLNCETH
372 CDU 592 554 492 468153 VDU 204 267 206 21822 RFCC 107 22 49 2276 FCC 76 91 76 5410 HCC 53 76 54 9474 DC 102 117 105 10015 KHT 26 19 18 1068 D2HT 130 122 111 10054 LCNHT 99 61 67 4234 CLNHT 48 55 49 4734 ST 48 55 49 4712 REF 18 26 21 23
NPV (bi US$) 8.387 5.2465 11.624 6.451Investment (bi US$) 25.000 24.681 19.170 21.312Profit (mi US$/d) 27.123 23.100 22.747 20.701CPU (s) 1.06 1.746 1.077 0.734
2009-2012 trends 4.2% p.a.
2020
2009-2012 trends 4.2% p.a.
2020
CONOPT
DICOPT (NLP: CONOPT)
7 EWO Meeting – March 2014
Multi-Site Approach – PDH (MILP+NLP)
PDH = Phenomenological Decomposition Heuristics = Quantity + Quality problems Decomposed
MILP NLP Quantity Problem: Logic + Quantity variables Quality Problem: Quantity + Quality Variables
REVAP CDU.1 VDU.1 FCC.1 PDA.1 DC.1 LCNHT.1 CLNHT.1 KHT.(1,2) DHT.(1,2) REF.1
12 units
REPLAN CDU.(1,2) VDU.(1,2) FCC.(1,2) DC.(1,2) LCNHT.(1,2) CLNHT.(1,2) DHT.(1,2) REF.1
15 units
RPBC CDU.(1,2,3) VDU.(1,2) FCC.1 DC.(1,2) LCNHT.(1,2) CLNHT.1 DHT.(1,2) REF.1 ALK.1
15 units
CDU.2.2 VDU.1.2 FCC.1.1 DHT.2.1 LCNHT.1.1 CLNHT.1.1 CDU.3.1 VDU.3.1
6 expans 2 install
CDU.2.1 VDU.1.1 FCC.1.1 LCNHT.1.1 CLNHT.1.1 KHT.1.1 KHT.2.2 DHT.2.1 CDU.3.2
8 expans 1 install
CDU.1.1 VDU.1.1 FCC.1.1 LCNHT.1.1 CLNHT.1.1 DHT.1.1
6 expans
GAMS @ Intel i7-3820QM 2.7GHz 16GB
(R,U,N,T) Refinery Unit Type Number of the Unit type Time
8
U.N.T U.N.T U.N.T (20 exp / 3 inst)
MILP/NLP (T=3)* step 1 2
CPU (s) MILP (CPLEX) 0.22 0.14
NLP (CONOPT) infeas 87.14
NLP (SNOPT) 92.09 -------
NLP (IPOPT) 345.68 -------
NPV (bi U$)
MILP (CPLEX) 30.995 30.995
NLP (CONOPT) infeas 31.273
NLP (SNOPT) 31.273 -------
NLP (IPOPT) 31.273 -------
Discrete Var.
Eq. Var. Non-zero
Non-Linear
--- 1,538 1,986 14,888 9,964
444 5,960 7,620 28,693 ---
--- 7,223 10,751 70,376 45,446
MILP (T=3)
NLP (T=3)
NLP (T=1)
*MINLP (DICOPT) problem does not converge due to NLP solver infeasibilities
9 EWO Meeting – March 2014
Conclusions
Novelty:
• Aggregated multi-site approach for capacity expansion of a country/company
• Nonlinearities from processing and blending to evaluate the capability
• Includes project execution time (excluding the production from expanded units during this period)
• Expansion and Installation to control the capacity increment of units
• Phenomenological decomposition (quantity + quality problems segregated)
• More realistic approach (in a quantitative manner) for strategic and investment planning in the oil-refining industry
10
Impact for industrial applications:
• Aggregated model (NLP and MINLP cases) used to define the overall capacity increment per type of oil-refinery unit demanded in the conceptual projects
• Realistic formulation to predict investments in oil-refinery units
• Avoids overestimating/underestimating capacity expansion/installation
• Evaluates the capability (not only the capacity) by including nonlinearities
EWO Meeting – March 2014
Conclusions
top related