Scheduling technology either commercial or homegrown in today’s crude-oil refining industries relies on a complex simulation of scenarios where the user is solely responsible for making many different decisions manually in the search for feasible solutions over some limited time-horizon i.e., trial-and-error heuristics. As a normal outcome, schedulers abandon these solutions and then return to their simpler spreadsheet simulators due to: (i) time-consuming efforts to configure and manage numerous scheduling scenarios, and (ii) requirements of updating premises and situations that are constantly changing. Moving to solutions based in optimization rather than simulation, the lecture describes the future steps in the refactoring of the scheduling technology in PETROBRAS considering in separate the graphic user interface (GUI) and data communication developments (non-modeling related), and the modeling and process engineering related in an automated decision-making with built-in problem representation facilities and integrated data handling features among other techniques in a smart scheduling frontline.
Crude-Oil Scheduling Technology: moving from simulation to optimization
1. Crude-Oil Scheduling Technology: moving from
simulation to optimization
1Refining Optimization, PETROBRAS Headquarters, Rio de Janeiro, Brazil.
2Center for Information, Automation and Mobility, Technological Research Institute, São Paulo, Brazil.
Brenno C. Menezes,1,2 Marcel Joly,1,2 Lincoln F. L. Moro1
Upstream Downstream Distribution Gas & Energy Biofuels
ESCAPE25, Copenhagen, Jun 2nd, 2015
4. 1- Scheduling Technology in PETROBRAS (home-grown solution SIPP)
2- Workshop on Commercial Scheduling Technologies in Oct, 2013
3- Refactoring/Remaking of SIPP: GUI + IT Developments
Modeling + Engineering Advancements
4- Applications of Optimization (CTA+ISW, DBCTO, MOVPath, Demi-Water)
5- Opportunities (CTA+ISW+DBCTO, Bottleneck Scheduling, Smart Operations)
6- Conclusions
Outline
5. 5
Scheduling Technology in PETROBRAS
Space
Time
Supply
Chain
Refinery
Process
Unit
second hour day month year
RTOControl
on-line off-line
Scheduling
Operational
Planning
Tactical
Planning
Strategic
Planning
Simulation
Petrobras
NLP Optimization
Commercial (Aspentech)
LP Optimization
Petrobras
Operational Corporate
SIPP: Integrated System for Production Scheduling
week
6. 6
What to do?
How and When to do?
Crude transf./receiving/diet
Process unit operations
Blending
Inventories
Deliveries
SheWhart or PlanDoCheckAct (PDCA) Management Cycle
Scheduling Technology in PETROBRAS
(Joly et al., 2015)
estimation
7. 7
Operational Planning (MINLP): (Neiro and Pinto, 2005)
Strategic Planning (MILP and MILP+NLP): (Menezes et al., 2015ab)
(Menezes , Kelly & Grossmann, 2015a): Phenomenological Decomposition Heuristic , ESCAPE25
(Menezes , Kelly, Grossmann & Vazacopoulos, 2015b): Generalized Capital Investment Planning, CACE
Goal: Multi-Site Scheduling
8. SIPP and Other Initiatives for Scheduling
SIPP
ARAUCARIASMART
Crude Oil
Transferring
Refinery Units Fuels
Deliveries
Fuels
Blending
Inventories
Crude Oil
Blending
SMART:
- Genetic Alg. model
using non-optimized
starting points
ARAUCARIA
- Continuous-time
impossible to be
executed in practice
Crude Oil
Receiving
Initiative Pitfalls:
9. Crude Oil
Transferring
Refinery Units Fuels
Deliveries
Product
Blending
Crude Oil
Receiving Inventories
Inventory control
Yields updated by hand
Crude heavy/light and sour/sweet
Blending indices from literature
Scheduling is
Worst Case Best Case
Crude, Units, Inventories, Deliveries
Yields updated automatically
Crude in several properties/yields
Blending using daily data/interp.
Crude Oil
Blending
11. 11
As a normal outcome, schedulers abandon these solutions and then
return to their simpler spreadsheet simulators due to:
(i) efforts to model and manage the numerous scheduling scenarios
(ii) requirements of updating premises and situations that are
constantly changing
(iii) manual scheduling is very time-consuming work.
SIPP’s or Simulation-based Solution Problems
“Automation
-of-Things”
(AoT) Automated Data Integration = IT Development
Automated Decision-Making = Optimization
Automated Data Integrity = Data Rec./Par. Est.
Needs of
12. 12
Simulation X Optimization
Simulation
Pros
• Wide-refinery simulation
• Familiar to Scheduler
• Quick solution (can be
rigorous)
Cons
• Trial-and-error
• Only feasible solution
Optimization
Pros
• Automated search for a
feasible solution
• Optimized solution (Local)
Cons
• Optimization of subsystems
• Solution time can explode
• High-skilled schedulers
• Global optimal (dream)
13. Workshop on Commercial Scheduling
Technologies in Oct, 2013
(Joly et al., 2015) M3Tech
Honeywell
SIMTO
Production Scheduler
Out of the market
14. GAMS
Pre-Formatted (Simulation) Modeling Platform (Optimization)
Soteica
IMPL
AIMMS
Off-Line
On-Line
Average
Price
10k (dev.) and 20k (dep.) +20% year100 k/year
(per tool)
Modeling Built-in
facilities
Without
facilities
Black
Box
Demanded Tools 1 13
Configuration Coding Configuration
Workshop on Commercial Scheduling
Technologies in Oct, 2013
OPL
15. - Drawer to generate flowsheet structures (Visual Prog. Lang.)
- Upper and lower bounds for yields (more realistic)
- Pre-Solver to reduce problem size and debug "common" infeas.
- Proprietary SLP to solve large-scale NLPs (called SLPQPE)
- Names-to-numbers to generate large models very quickly
- Ability to add ad-hoc formula (e.g., blending rules)
- Generates analytical quality derivatives using complex numbers
- Initial value randomization to search for better solutions
- Digitization/discretization engine (continuous-time data input)
IMPL Important Techniques/Features
(Industrial Modeling and Programming Language)
16. Modeling and Programming Languages Aspects
- Same process unit models for planning and scheduling
- Planning & scheduling with data-mining, MPC, data rec., RTO
- CDU(N) and VDU(M) as hypos, pseudo-components or micro-
cuts for any NxM arrangement (towers in cascade)
- Hierarchical Decomposition Heuristics HDH (Kelly & Zyngier, 2008)
- Phenomenological Decomposition Heuristics PDH: the MINLP
model is partitioned in MILP and NLP (Menezes, Kelly & Grossmann,
ESCAPE25, 2015)
17. 1- APS (Advanced Planning and Scheduling):
Planning: Aspen, Soteica
Scheduling: Aspen, Princeps, Soteica, Invensys
Blending: Aspen, Princeps, Invensys
2- APC (Advanced Process Control): Aspen, gProms
3- RTO (Real-Time Optimization): Aspen, Invensys
4- Data Reconciliation and Parameter Estimation: Aspen, KBC, Soteica
5- Hybrid Dynamic Simulation: Aspen, KBC, Invensys
6- Differential Equation Solution (ODE and PDE): gProms
Applications in IMPL
18. 1st STEP: separate (GUI + IT) from (Modeling + Engineering)
2nd STEP: prototype (ModEng) using easy-to-use modeling language
3rd STEP: prototype (GUI+IT) in a reactive iteration with 2nd STEP
30% 30%30%
GUI
(Graphic User Interface)
Interfacing/database Modeling+Engineering
10%
Solver
GUI + IT Modeling + Engineering
Refactoring/Remaking of SIPP
4th STEP: integrate (GUI + IT) and (Modeling + Engineering)
19. GUI + IT Developments
30%30%
GUI
(Graphic User Interface)
Interfacing/database
GUI + IT
Plant
(Visio)
Database
(Oracle)
Simulation
(Visual C++)
IHM
(Delphi)
Movement and Mixing
Optimization Management
GOMM
New GUI in C#
20. Modeling + Engineering Advancements
30%
Modeling+Engineering
10%
Solver
Modeling + Engineering
1st: Refinery Teams should be
involved in the modeling
Demand: easy-to-use tools
2nd: Optimize subsystems and
integrate them incrementally
HQ R&D
Center
Refineries
Universities
IT Develp.
Center
Petrobras case:
- HQ + CMU + São Paulo/Rio
Universities
- R&D
Center
Several Brazilian
Universities
+
Research Phase Development Phase
(5-10 years) (1-3 years)
dataflow or diagrammatic programming
21. IMPL’s UOPSS Visual Programming Language using DIA
Variable Names:
v2r_xmfm,t: unit-operation m flow variable
v3r_xjifj,i,t: unit-operation-port-state-unit-operation-port-state ji flow variable
v2r_ymsum,t: unit-operation m setup variable
v3r_yjisuj,i,t: unit-operation-port-state-unit-operation-port-state ji setup variable
VPLs (known as dataflow or diagrammatic programming) are based on the idea of "boxes and
arrows", where boxes or other screen objects are treated as entities, connected by arrows,
lines or arcs which represent relations (node-port constructs). (Bragg et al., 2004)
x = continuous variables (flow f)
y = binary variables (setup su)
j
22. 𝐯𝟐𝐫_𝒙𝒎𝒇 𝐦,𝒕 ≥ 𝑳𝑩𝒇 𝒎 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ∀ 𝐦, 𝐭 (1)
𝐯𝟐𝐫_𝒙𝒎𝒇 𝐦,𝒕 ≤ 𝑼𝑩𝒇 𝒎 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ∀ 𝐦, 𝐭 (2)
𝐣∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≥ 𝑳𝑩𝒇 𝒎 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ∀ (𝐢, 𝐦), 𝐭
(3)
𝐣∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≤ 𝑼𝑩𝒇 𝒎 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ∀ (𝐢, 𝐦), 𝐭
(4)
𝐢∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≥ 𝑳𝑩𝒇 𝒎 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ∀ (𝐦, 𝐣), 𝐭
(5)
𝐢∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≤ 𝑼𝑩𝒇 𝒎 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ∀ (𝐦, 𝐣), 𝐭
(6)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≥ 𝑳𝑩𝒇𝐣,𝒊 𝐯𝟑𝐫_𝒚𝒋𝒊𝒔𝒖𝐣,𝐢,𝒕 ∀ (𝐣, 𝐢), 𝐭
(7)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≤ 𝑼𝑩𝒇𝐣,𝒊 𝐯𝟑𝐫_𝒚𝒋𝒊𝒔𝒖𝐣,𝐢,𝒕 ∀ (𝐣, 𝐢), 𝐭 (8)
j
Semi-continuous
equations for units
Semi-continuous
equations for streams
Mixer for each i, but
using lo/up bounds
Splitter for each j, but
using lo/up bounds
23. 𝐣∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≥ 𝑳𝑩𝒚𝐢,𝒎 𝐯𝟐𝐫_𝒙𝒎𝒇 𝐦,𝒕 ∀ (𝐢, 𝐦), 𝐭
(9)
𝐣∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≤ 𝑼𝑩𝒚𝐢,𝒎 𝐯𝟐𝐫_𝒙𝒎𝒇 𝐦,𝒕 ∀ (𝐢, 𝐦), 𝐭
(10)
𝐢∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≥ 𝑳𝑩𝒚 𝐦,𝒋 𝐯𝟐𝐫_𝒙𝒎𝒇 𝐦,𝒕 ∀ (𝐣, 𝐦), 𝐭
(11)
𝐢∈(𝐣,𝐢)
𝐯𝟑𝐫_𝒙𝒋𝒊𝒇𝐣,𝐢,𝒕 ≤ 𝑼𝑩𝒚 𝐦,𝒋 𝐯𝟐𝐫_𝒙𝒎𝒇 𝐦,𝒕 ∀ (𝐣, 𝐦), 𝐭
(12)
𝐦(𝐦∈𝐮)
𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ≤ 𝟏 ∀ 𝐮, 𝐭
(13)
𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝒎′,𝒕 + 𝐯𝟐𝐫_𝒚𝒎𝒔𝒖 𝐦,𝒕 ≥ 𝟐 𝐯𝟑𝐫_𝒚𝒋𝒊𝒔𝒖𝐣,𝐢,𝒕∀ 𝒎′
, 𝒋 , (𝐢, 𝐦) (14)
(Menezes , Kelly, Grossmann & Vazacopoulos, 2015b): Generalized Capital Investment Planning, CACE
xX
xX
x
x
j
Several unit feeds
(treated as yields
with lower and
upper bounds)
Selection of modes
in one physical unit
Structural
Transitions
26. Crude Tank Assignment + Improved Swing Cut
(CTA) (ISW)
Kerosene
Light Diesel
ATR
CDU
C1C2
C3C4
SW1
SW2
SW3
VR
VDU
N
K
LD
HD
D1HT
Naphtha
Heavy Diesel
LVGO
HVGO HTD2
D2HT
HTD1
to hydrotreating
and/or reforming
(To FCC)
Crude C
Crude D
(To Delayed Coker)
to hydrotreating
to caustic and
amines treating
JET
GLN
FG
LPG
VGO
FO
Final Products
MSD
HSD
LSD
Crude A
Crude B
(Menezes, Kelly & Grossmann, 2013)(IAL, 2015)
Clusters or Crude Tanks
Crude
Min cr,pr(Crude-Cluster)2
cr crude
pr property
pr ou yields: naphtha-yield (NY), diesel-yield (DY), diesel-sulfur (DS) and residue-yield (RY)
Improve the flexibility in the search for
optimized diet/recipe/blend
27. Distillation Blending and Cutpoint Temperature
Optimization (DBCTO) (Kelly, Menezes & Grossmann, 2014)
From Other
Units
From CDU
Kerosene
Light Diesel
ATR
C1C2
C3C4
N
K
LD
HD
Naphtha
Heavy Diesel
Crude
CDU
ASTM D86
TBP
Inter-conversion
Evaporation
Curves
Interpolation
Ideal Blending
Evaporation
Curve
Multiple
Components
Final
Product
ASTM D86
Interpolation
Inter-conversion
TBP
𝐘𝐍𝐓𝟗𝟗 = 𝟎. 𝟗𝟎 +
𝟎. 𝟗𝟗 − 𝟎. 𝟗𝟎
𝐎𝐓𝟗𝟗 − 𝐎𝐓𝟗𝟎
𝐍𝐓𝟗𝟗 − 𝐎𝐓𝟗𝟎
𝐘𝐍𝐓𝟎𝟏 = 𝟎. 𝟏𝟎 −
𝟎. 𝟏𝟎 − 𝟎. 𝟎𝟏
𝐎𝐓𝟏𝟎 − 𝐎𝐓𝟎𝟏
𝐎𝐓𝟏𝟎 − 𝐍𝐓𝟎𝟏
𝐃𝐘𝐍𝐓𝟎𝟏 = 𝟎. 𝟎𝟏 − 𝐘𝐍𝐓𝟎𝟏
𝐃𝐘𝐍𝐓𝟗𝟗 = 𝐘𝐍𝐓𝟗𝟗 − 𝟎. 𝟗𝟗
𝐎𝐥𝐝 𝐓𝐞𝐦𝐩𝐞𝐫𝐚𝐭𝐮𝐫𝐞: 𝐎𝐓
New Temperature: NT
New Yield: YNT
Difference in Yield: DYNT
28. Crude Oil
Transferring
Refinery Units Fuels
Deliveries
Product
Blending
Crude Oil
Receiving
Inventories
Opportunities in CTA+ISW+DBCTO
CTA
ISW DBCTO
New-SIPP with optimization
GOMMCrude Oil
Blending
New-SIPPOT
inside GOMM
to register the
execution of
the scheduling
29. Bottleneck Scheduling
Step 1: Identify Key Bottlenecks (see below)
Step 2: Design Optimization Strategy
Step 3: Determine Information Requirements
Step 4: Prototype and Implement, etc.
Quantity-related:
Inventory containment
Hydraulically constrained
Logic-related (Physics):
Mixing, certification delays, run-lengths, etc.
Sequencing and timing
Quality-related (Chemistry):
Octane limits on gasoline
Freeze and cloud-points on
kerosene and diesels, etc
Step 5: Capture Benefits Immediately
(Harjunkoski, 2015)
Scheduling Solution Development Curves
30. Smart Operations
(Qin, 2014)(Christofides et al., 2007)
(Davis et al., 2012)
(Huang et al., 2012)
(Chongwatpol and Sharda, 2013)
(Ivanov et al., 2013)
Smart Process Manufacturing Big Data RFID in APS and Supply Chain
Opportunity for Molecular Scheduling for a selected crude feed
Example: when crude is selected for 2-4 days, after the 1st shift of 8h update all
data using Information and Communication Technologies (ICT) integrated with
Data-Mining applications and then use this in the Decision-Making
31. 31
• Partnership Industry-Academia is fundamental for modeling advances.
Our vision it is missing some RPSE section, initiative, journal, meeting, etc.
• Automated DMs (Decision-Making and Data Mining)
• Permit schedulers to model using VPL in diagrammatic programming
• When moving from simulation to optimization:
Conclusions
- Optimize subsystems and then, if necessary, integrate them
incrementally
- Integrate distillates cutpoints and blending using daily data in
today’s operations as well as hydrotreating severity, etc.
- Be sure the data is accurate otherwise the decision is bad despite
the modeling