Optimizer as a CAPE-OPEN Unit Operation - · PDF fileOptimizer as a CAPE-OPEN Unit Operation...
Transcript of Optimizer as a CAPE-OPEN Unit Operation - · PDF fileOptimizer as a CAPE-OPEN Unit Operation...
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah
©IF
P
Extended reserves | Clean refining | Fuel-efficient vehicles | Diversified fuels | Controlled CO2
Optimizer as a CAPE-OPENUnit Operation
2007 AIChE Annual Meeting – Topical D: 4th US CAPE-OPEN Conference
Authors:Martin Gainville, Pascal Roux: IFPAlain Vacher, Philippe Baudet: ProSim SAMichel Pons: Michel Pons Technologie SA (for TOTAL)Didier Paen: RSI - Dynamic Simulations Solutions
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 2
©IF
P
Content
� Introduction
� Problem statement
� Optimizer integration and validation
� Case study
� Conclusion
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 3
©IF
P
Introduction Current situation
Riser
Riser base
PumaFlow
ECLIPSEREVEAL...
OLGA
LEDA
TACITE
PIPESIM, PIPEPHASE, GAP
...
PROSPER
WELLFLO
...
HYSYSINDISSD-SPICEPRO/IIUniSim...
Reservoir
Well Head
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 4
©IF
P
Introduction
� Design and operation of oil & gas production systems require more and more advanced process simulations� Multiple software solutions to address multiple knowledge domains
� For integrated simulations from reservoir to process, seamless integration of different provider solutions is needed
� Numerous issues to be resolved when coupling these solutions� Components interoperability� Data consistency
� Numerical consistency
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 5
©IF
P
Introduction
� To address these requirements, TOTAL and IFP have led a research collaborative project� TINA project based on INDISS Process Modelling Environment
� Purpose of the study� Extends modular strategy of INDISS for performing steady-state
optimizations� Using a 3rd party optimization module within INDISS architecture
� In a CAPE-OPEN interoperability framework� CAPE-OPEN PME: INDISS
� TINA pipeline modules are CAPE-OPEN compliant� Optimizer wrapped as a CAPE-OPEN unit operation to be easily
integrated within INDISS architecture
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 6
©IF
P
Problem statement� Network variables: x
� Unit operation parameters (ex: choke opening), pressure, flow rate, partial flow rate, ...
� Function to optimize: f(x)� Min of cost, max of Net Present Value, max of production, ...
� Equality constraints� N(y) = 0, y ⊂⊂⊂⊂ x, network equations
� Conservation equations, recycling equations, homogeneous pressure at nodes, ...
� Z(x) = 0, other specifications
� Inequality constraints� C(x) ≤ 0
� Lower bound ≤ x ≤ Upper bound
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 7
©IF
P
Problem statement� Solving mechanism: a sequential modular approach and
an "Unfeasible path" strategy
C(x), Z(x), N(y), f(x)
Optimizer
C(x) ≤≤≤≤ 0 ? Z(x) = 0 ? N(y) = 0 ?
Lower bound ≤≤≤≤ x ≤≤≤≤ Upper Bound ?
min f(x) ?
x for new estimate
or
x+dx for gradient estimate
Sequential modular calculation with a specific calculation order
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 8
©IF
P
Optimizer integration� OPTI module: optimizer
� Native optimizer of ProSimPlus PME (ProSim SA)� Applied to convergence of NLP optimization problems
� Designed for a sequential modular approach
� Unfeasible path strategy is recommended
� Jacobian Matrix is estimated by numerical differentiation� Physical models are complex and implicit therefore analytical
derivatives are not available
� OPTI wrapped as a CAPE-OPEN unit operation (CO-OPTI)� Specific parameter to check the optimizer convergence� Data transfer using CAPE-OPEN information streams
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 9
©IF
P
Optimizer integration� INDISS PME has been adapted: INDISS-TINA
� To check the convergence of the optimizer� To manage tear streams solved by CO-OPTI� To establish convenient calculation order
� CO-OPTI made last UO in the maximum recycle sequence
� CO-OPTI ICapeUnit.Calculate( ) method� Retrieves values of constraints and objective function through
information streams� Transfers data to OPTI that gives new iterative values in return� Updates CO-OPTI outputs variables with the new iterative values
� Including recycle material streams which are flashed
� Updates the convergence status (flag parameter)
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 10
©IF
P
Validation on a simple linear problem� Within ProSimPlus PME
� ProSimPlus is a CAPE-OPEN compliant PME� Native OPTI module in ProSimPlus and CO-OPTI used as CAPE-
OPEN UO in ProSimPlus give identical results
� Within ProSimPlus and INDISS using CO-OPTI as a CO UO� Convergence path and final results are identical in both PMEs
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 11
©IF
P
Case study
( )γβα ++= DDLC Line2
CPumpPump WBAC +=
PQW Pump ∆=
� Cost optimization of a sub-sea production system� Target function: minimize the design cost of the system� Cost of pipe installation + cost of multiphase pump
� Optimization parameters� Line diameter� Pump useful power (or differential pressure)
� Equality constraint (operational constraint)� Topside arrival pressure equal to 15 bars
� Inequality constraints� Maximum Pump useful power: 1.2 MW� Line diameter: from 8 to 12 inches
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 12
©IF
P
Case study� Demo: CO-OPTI within INDISS-TINA
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 13
©IF
P
8
9
10
11
12
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15Iterations
Line
dia
met
er (i
nch)
0
0.2
0.4
0.6
0.8
1
1.2
Use
full
pum
p po
wer
(MW
)
DIAMETER POWER
Case study� Results
CONSTRAINTS VIOLATION
0.00001
0.0001
0.001
0.01
0.1
1
10
100
1000
10000
0 2 4 6 8 10 12 14 16
Iterations
OBJECTIVE FUNCTION
3.54E+07
3.56E+07
3.58E+07
3.60E+07
3.62E+07
3.64E+07
3.66E+07
0 2 4 6 8 10 12 14 16
Iterations
Upper bounds
Lower bounds
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 14
©IF
P
Conclusion
� CO-OPTI integrated and used successfully in INDISS as a CAPE-OPEN Unit Operation� Use of CAPE-OPEN specifications has reduced development time for
integrating CO-OPTI into INDISS-TINA� CO-OPTI easily plugged into INDISS-TINA PME
� Recycle material streams easily handled� CAPE-OPEN information streams useful for data transfer
� CAPE-OPEN enables optimizer integration in sequential modular PMEs
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 15
©IF
P
Thank you for your attention
� Acknowledgements to TOTAL, RSI, ProSim SA for their contribution to this work
� Questions?
� Additional explanations will be provided tomorrow morning during the CAPE-OPEN Interoperability Workshop session
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 16
©IF
P
Case study� Sensitivity study to the transportation distance
� Optimizations are performed for several distances� 5, 10, 15, 20 and 25 km
� Three design options� Option 1: system designed with the maximum specified pipe
diameter (Pipe diameter of 12 inches), � Option 2: system designed with the maximum pump unit
capacity (Pump unit of 1.2 MW),� Option 3: optimum solution calculated by CO-OPTI
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 17
©IF
P
Case study� Normalized overall cost
0.5
0.6
0.7
0.8
0.9
1
0 5 10 15 20 25 30Tie-back distance [km]
No
rmal
ized
ove
rall
cost
Option 1 - Pipe diameter of 12 inches Option 2 - Pump unit of 1.2 MW
Option 3 - 0ptimum solution calculated by CO-OPTI
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 18
©IF
P
Case study� Over cost in percentage of the optimum cost
0
2
4
6
8
10
12
0 5 10 15 20 25 30Tie-back distance [km]
Co
st b
enef
it in
per
cen
tag
e o
f th
e o
pti
mu
m o
vera
ll co
st [
%]
Option 1 - Pipe diameter of 12 inches Option 2 - Pump unit of 1.2 MW
2007 AIChE Annual Meeting - Topical D: 4th CAPE-OPEN US Conference - Salt Lake City, Utah 19
©IF
P
Optimizer integration� Schematic view
� Calculation order� {UO1, UO2, UO3, Evaluation module, UO5, UO4, CO-OPTI}
ICapeUtilities.Edit