A Nonlinear Integrated Model for Operational Planning of Multi-Site Refineries

Post on 25-May-2015

389 views 2 download

Tags:

Transcript of A Nonlinear Integrated Model for Operational Planning of Multi-Site Refineries

A Nonlinear Integrated Model for Operational Planning

of Multi-Site Refineries

Brenno C. Menezes, Lincoln F. Moro Refining Optimization PETROBRAS Petroleo S.A. Rio de Janeiro, RJ

Ignacio E. Grossmann Department of Chemical Engineering Carnegie Mellon University Pittsburgh, PA

1

Jeffrey D. Kelly Toronto, ON

industrIALgorithms

Thesis Overview

Summary

2

1st- Operational Planning of Sao Paulo Refineries

2nd- Swing-Cuts Improvements

Why Nonlinear?

Why Integrated?

Why Operational Planning?

Why Multi-Site Refineries?

A Nonlinear Integrated Model for Operational Planning

of Multi-Site Refineries

5 min

5 min

5 min

Quantitative Methods for Investment and Strategic

Planning in the Oil-Refining Industry

Brenno C. Menezes, Lincoln F. Moro Refining Optimization PETROBRAS Petroleo S.A. Rio de Janeiro, RJ

Ignacio E. Grossmann Department of Chemical Engineering Carnegie Mellon University Pittsburgh, PA

3

Jeffrey D. Kelly Toronto, ON

Fernando Pellegrini, Ricardo Medronho Department of Chemical Engineering Federal University of Rio de Janeiro Rio de Janeiro, RJ

industrIALgorithms

The thesis aims to develop a quantitative method to predict necessaries

structural modifications in the Brazilian refining and logistics assets through time

PETROBRAS Current Tool for Strategic Planning (PLANINV) – LP Tool

No Framework Synthesis

Optimize only the streams transfers (fuel and petroleum import/export, fuel local market supply)

PLANINV Framework OT

4

The necessity to develop a strategic level supply chain planning models in order to address issues in a quantitative manner rather than the qualitative approaches used till now is acknowledge by the industry and still remains an active research area. (Shapiro, 2004; Papageorgiou, 2008)

Still need modification on the figure. I want to show a scheme with current Investment Methodology

Jul-13

GAMS Distillation Models (CDU/VDU) Swing-Cuts Improvements

Sep-12 Nov-12

IMPRESS

PIMS

Jan-13 May-13 Aug-12

Nonlinear Operational Planning of Sao Paulo Refineries

EWO

Multi-plant Operational Planning of Sao Paulo Refineries

REPLAN Investments Cases for 2020-2030

Multi-period Investment Planning with NPV as Goal (refining only) MINLP

CAPD & EWO

Multi-period Strategic and Logistics Planning with NPV as Goal (refining, transportation and terminals) MILP

Distillation Models CDF least-square Interpolation

Mar-13

Binary variables

Blending

Refining Processes

fixed recipes

variable recipes

fixed yields

swing cuts (fixed properties)

Which Surface?

GAMS

IMPRESS

Planning the Future & Existing Refining Units (MINLP; DICOPT++)

Planning the Future & Existing Refining Units and Logistics (MILP; UOSS/QLQP) 8

Basic Equations for Modeling Process Unit in a Refinery

Mixer:

u u',s,u

(u',s)

QF = Q

uUS

Feed Properties:

u,p u,p u',s,u u',s,p(u',s) (u',s) ,pPF =f Q ,PF

u u u,sUS US PO

Products from Units:

u,s u,s u u,p u,v vpQS =f QF ,PF ,V

uu VOPI

Products Properties from Units:

u,s,p u,s,p u,p u,v vpPS =f PF ,V

uu VOPI

Splitter:

u u,s,u'

(u',s)

QS = Q

uUS

Mixer Unit Splitter

Splitter

Splitter

9

Qu’,s,u

Qu’,s,u

Qu’,s,u

Qu,s,u’

Qu,s,u’

Qu,s,u’

Qu,s,u’

Qu,s,u’

Qu,s,u’

Qu,s,u’

QSu,s

QSu,s

QSu,s

QFu

QFu Feed Flow

QSu,s Product Flow

Qu,s,u’ Transfer Stream Flow

PFu,p Feed Property

PSu,s,p Product Property PFu,p

PSu,s,p

PSu,s,p

PSu,s,p u units

s streams

p properties

NLP Monoperiod (Operational Planning)

Find the best quantity of Kero and VGO from REPLAN to: Kero => REVAP VGO => REVAP/RBPC - Today maximum permitted for Kero is 1500 m3/d - From the NLP Model the best value is 2300 m3/d.

FCC FCC,RCRFCC FCC FCC,s,RCR FCC,RCR

FCC,s,TRX FCC,s,TCC

QS QF .[Y Y .(PF PF )

Y .TRX Y .TCC] s

FCCSO

CDi,sCDi,s CDi CDi,s i iQS QF . (Y Y . HOT ) s , CD CDiSO CD

PDA,ASFR PDAQS QF .(1 EXT)

k k kHT ,HTs,S HT ,S HT kPF =PF 1 SEV HT HT

Crude recipe: Yields Sulfur Gravity Acidity

Swing Cuts Fractionation-Index Interpolation Regressed CDF

(Moro, Zanin & Pinto, 1998)

Refining Framework Modeling

11

CDU Models Yields Properties Modeling Reference

Fixed yields Fixed Fixed LP conventional approach

Delta Base Base+Delta Base+Delta NLP Moro, Zanin & Pinto (1998)

Swing-Cuts Pre-Cut Fixed LP Zhang et. Al. (2001)

Swing Cuts Modificated Pre-Cut with

operational modes

LS, Prop=f(Cum. Yie) NLP Li, Hui & Li (2005)

Fractionation-Index

(Heaviside function to

control FIR and FIS)

Geddes eq.

K=y/x

K=f(T,FIR,FIS)

Non-distribution in T

FIR=FI rectifying section

FIS=FI stripping section

NLP Alattas, Grossmann & Rivera (2011)

Fractionation-Index

(Binary logic to control FIR

and FIS)

Geddes eq.

K=y/x

K=f(T,FIR,FIS)

Non-distribution in T

FIR=FI rectifying section

FIS=FI stripping section

MINLP Alattas, Grossmann & Rivera (2012)

Hybrid

(mass/energy equations +

empirical PLS relations)

Mass/Energy

balance

Tray Temperature

Measurements

PLS only for TBP SLP Mahalec & Sanchez (2012)

Swing-Cuts Improvement Cutting & Blending

Hypos-Swing

Ordination

Volume-Mass weighted

interpolation

NLP Menezes, Kelly & Grossmann (2013)

Linear & Monotonic Spline

Interpolation

Yield=f(T) Prop=f(T) NLP Menezes, Kelly & Grossmann (2013)

Least-Squares Fit of CDF Yield=CDF(T) Prop=CDF(T) NLP Menezes, Kelly & Grossmann (2013)

Fractionation-Index

(Heaviside function to

control FIR and FIS)

Same as above Prop=f(T,FIR,FIS) NLP Menezes & Grossmann (2013)

Fractionation-Index

(Binary logic to control FIR

and FIS)

Same as above Prop=f(T,FIR,FIS) MINLP Menezes & Grossmann (2013)

A

B

C

Cut-to-Mix Mix-to-Cut

Assay

CDF/Interpolation

Hypos Cutting and blending

Hypos Tcut Functions

Mass Bal + Constrains (Corrections)

Volume/Mass weighted Interpolation

Conventional Approach

Cut-to-Mix Hypos-Swing Ordination

Mix-to-Cut Hypos-Swing Ordination

Hypos -> Cuts -> FCuts Swing-Cuts flows as variables

Hypos -> FCuts (need Hypos-Swing-Cuts Ordination) Hypos flows as variables

Big CDU Hypo

Hypos -> FCuts Tcut as variables [TISW,TESW]

CDF (Weibull Extreme)

Linear Interp.

Monotonic Spline

CDU Hypos per crude

Fraction Index

Yields Distribution (K=y/x)

Properties Distribution

Hypos -> FCuts Tcut as variables [TISW,TESW]

GAMS is not supporting monotonic splines (piecewise Hermite polynomial)

Motivating Example: Swing-Cuts Model This example is the well-known Swing-Cuts model applied in commercial tools for operational planning like as PIMS. The Swing-Cuts (SW1, SW2, SW3, SW4) are treated as a normal Cuts (LN,HN,K,LD,HD) with constants properties, so if the SWi is going to the upper or lower adjacent cut they will affect the Final Cuts properties (LN,HN,K,LD,HD) equally.

LN

K

LD

HN

HD

ATR

CDU C1C2

C3C4

AGBAMI

BARRACUDA

LULA

MARLIM

PCONCHAS

RONCADOR

SW1

SW2

SW3

SW4

VR

HVGO

VDU

LVGO

PFO

PVGO

PHDS

PLDS

PJFUEL

PGLNLN

K

LD

HN

HD

C3C4

C1C2

VR

HVGO

LVGOCUTS=LN,SW1,HN,SW2,K,SW3,LD,SW4,HD

FCUTS=LN,HN,K,LD,HD

SWINGS=SW1,SW2,SW3,SW4

Using this example as a baseline, different approaches are proposed to improve the properties accuracy of the final cuts:

•Swing-Cuts: Cut-to-Mix with Corrections •Swing-Cuts: Cut-to-Mix with Corrections + Volume/Mass Weighted Interpolation •Swing-Cuts: Cut-to-Mix with Hypo-Swing-Cuts Ordination

•Swing-Cuts: Mix-to-Cut with Hypo-Swing-Cuts Ordination

20 140 160

LN

SW1 HN

180

SW2

LN

HN

Vol

T(ºC) 210

C1C2

C3C4

LN

HN

SW1 SW2

K

SW3

LD

TI TE

C1C2 -273 -50

C3C4 -50 20

LN 20 140

SW1 140 160

HN 160 180

SW2 180 210

K 210 240

SW3 240 260

LD 260 360

SW4 360 380

HD 380 420

ATR 440 850

LVGO 440 580

HVGO 580 620

VR 620 850

CUTS Final Pools

Tcuts Assay

Cutting

blending

FCUTS HYPOS

Big HYPOS FCUTS Final Pools

Hypos-Averaged inside the CDU

Trange

Cutting

blending

HYPOS

Hypos-Swing Ordination

1- Cut-to-Mix with Corrections (Constrains and Mass Balance)

This approach is just a numerical correction once the properties are averaged values. Appling a mass balance and a sulfur mass balance and a set of constrains in the SWis :

(0.778*463+0.796*1999)/(463+1999)=0.793

(0.870*488+0.913*764)/(488+764)=0.896

Conventional Aproacch

2- Cut-to-Mix Correction and Interfacial Interpolation Interpolating the SW1L between the layers (LN,HN).

QSW1UP

(0.778*463+0.796*1999)/(463+1999)=0.793

(0.878*488+0.907*764)/(488+764)=0.896

(0.778*463+0.796*1999)/(463+1999)=0.793

(0.870*488+0.913*764)/(488+764)=0.896

normal approach SWis

properties constants

hypos->Cuts>Final Cuts Ready

1 Cut-to-Mix Mass bal and constrains hypos->Cuts>Final Cuts Ready

2 Cut-to-Mix Mass bal and constrains +

Interfacial Interpolation

hypos->Cuts>Final Cuts Ready

3 Cut-to-Mix Swing-Hypos ordination hypos->Final Cuts Need Ordination

4 Mix-to-Cut hypos->Final Cuts Ready

5 Mix-to-Cut Swing-Hypos ordination hypos->Final Cuts Ready

6 Mix-to-Cut Swing-Hypos ordination,

Mass bal and constrains +

Interfacial Interpolation

hypos->Final Cuts Need Ordination

+ interfacial Interpolation

7 CDF Weibull Extreme Tcuts [TISW,TESW] Ready

8 CDF with sin/cos

Weibull Correction

Weibull Extreme

+correction

Tcuts [TISW,TESW] Need Interpolation in GAMS

9 Linear Interpolation Tcuts [TISW,TESW] Need Interpolation in GAMS

10 Spline Tcuts [TISW,TESW] Not started yet

Fraction Index 11 Fraction Index Tcuts [TISW,TESW] Need properties correction

baseline

Swing Cuts

Regressed

Models

Sahinidis, N. V., Grossmann, I. E., Fornari, R. E., Chathrathi, M. (1989). Optimization model for long range planning in the chemical industry. Computers and Chemical Engineering, 13(9), 1049-1063. Moro, L.F.L., Zanin, A.C. e Pinto, J.M. (1998). A planning model for refinery diesel production. Computers and Chemical Engineering, 22 (1), 1039-1042. Li, W., Hui, C.W. e Li, A. (2005). Integrating CDU, FCC and blending models into a refinery planning. Computers and Chemical Engineering, 29, 2010-2028. Alattas, A. M., Grossmann, I. E., Paulo-Rivera, I. (2011). Integration of nonlinear crude distillation unit models in refinery planning optimization. Industrial and Engineering Chemistry Research, 50, 6860-6870. Alattas, A. M., Grossmann, I. E., Paulo-Rivera, I. (2012). Refinery production planning: multiperiod MINLP with nonlinear CDU model. Industrial and Engineering Chemistry Research (Accepted Aug 23rd). Zyngier, D., Kelly, J. D. (2012). UOPSS: A new paradigm for modeling planning and sheduling systems. ESCAPE 22, June 17-20, London.

References

29