draft phd presentation robert brunet
TRANSCRIPT
Robert Brunet Page 1 of 45
ROBERT BRUNET SOLÉSupervisors: Dr. Gonzalo Guillén and Dr. Laureano Jiménez
Department of Chemical EngineeringUniversitat Rovira i Virgili, Tarragona
Tarragona, 19th December 2012
Optimal design of sustainable chemical processes via combined simulation-optimization approach
Robert Brunet Page 2 of 45
1. Introduction (Ch1)
2. Bioprocesses (Ch 2 & 3)
3. Thermodynamic cycles (Ch 4 & 5)4. Biofuels (Ch 6 & 7)
5. Conclusions (Ch 8)
Robert Brunet Page 3 of 45
1. Introduction
Robert Brunet Page 4 of 45
Math
em
ati
cal
Pro
gra
mm
ing
Eco
nom
ic
Evalu
ati
on
Ch
em
ical
Pro
cess
es
Life
Cycl
e
Ass
ess
men
t
Multi-objective optimization for sustainable
chemical process design
Pro
cess
S
imu
lati
on
Pack
ag
es
Case Study IndicatorsTools
Key basis of my research
Robert Brunet Page 5 of 45
Aim of the work •Develop systematic tools to achieve reductions in production costs and environmental impact of bioprocesses
• Systematic method based on the combined use of simulation and optmization tools
Main motivation
•Chemical companies need to develop more sustainable processes:
• Plant profitability increase• Emissions and enviromental impact reduction
Aim of the work
Robert Brunet Page 6 of 45
Chemical processes
Robert Brunet Page 7 of 45
Mathematical Programming
Algebraic eq. (f, h, g)
Linear
Non-linear
Variables (x, y)
Continuous
Discrete {0,1]
LPLinear
Programming
NLPNon-Linear
Programming
MILP Mixed Integer
Linear Programming
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
MINLPMixed Integer
Non-Linear Programming
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
}1,0{,
0),(
0),(..
),(min
yx
yxg
yxhts
yxf
Advanced customized solution methods required
Robert Brunet Page 8 of 45
Discrete variables (logical decisions denoting the potential existence of process units)
Process equations:• Non-linear performance of the system (mass and energy balances)• Thermodynamic properties
•Continuous variables:• Flows• Operating conditions (pressures, temperatures, etc.)• Sizes of equipments
•Design specifications (linear inequalities)
Objective functions (cost and environmental impact)
How can we measure the environmental impact
?
Mathematical formulation
Robert Brunet Page 9 of 45
Economical Evaluation (Net Present Value or Total Capital Investment or Operating Cost)
Life Cycle Assessment (LCA)
Economic and Environmental Analysis
Life Cycle Assessment
(LCA)
Life Cycle Assessment
(LCA)
Evaluate the environmental loads associated with a product or process
Evaluate the environmental loads associated with a product or process
Quantifying energy and materials used and waste released
Quantifying energy and materials used and waste released
to evaluate opportunities for improvements
to evaluate opportunities for improvements
Robert Brunet Page 10 of 45
Translate inventory into damage
• Human health
• Ecosystem quality
• Depletion of resources
Direct emissions from
the process
Express the life cycle inventory as a function of some continuous variables:
Damage in each impact indicator (11 indicators)
Damage in each damage category (3 damage
categories)
Waste generationProduction of raw materials
Operation phase
Construction phase
Process variables (pressures, temperatures,
flows, etc.)
LCA methodology
Robert Brunet Page 11 of 45
Combined Simulation-Optimization
Dependent variables:(Heat flow, Area,
Power)
Decision variables(Temperature, Pressure,
mass flow)
Using process simulators instead generic modeling systems…
Index calculator &Constraints evaluation(economic, LCA, etc.)
Optimization solvers
NLP solver (fmincon)
MILP solver (CPLEX)
Process Simulator
(Aspen Plus, Aspen HYSYS, SuperPro)
Robert Brunet Page 12 of 45
Environmental Impact
Cost
Multiobjective optimization problems (economic and environmental concerns)
Epsilon constraint methodology:
Solve a set of single objective problems for different values of ε
Epsilon constraint methodology
Robert Brunet Page 13 of 45
2. Bioprocesses
Robert Brunet Page 14 of 45
0
100
200
300
400
500
600
0 10 20 30 40 50 60
Annual Volume[m3/year] *103
Ma
rke
t P
ric
e [
M$
/kg
] *1
03
Pharma
Health Care
DetergentsFood/feed
Basic Chemicals
Bioprocesses
Robert Brunet Page 15 of 45
L-lysine production plant (Heinzle et al, 2006)
Raw materials preparation
Biomass removal and concentration
Bioreactor
Robert Brunet Page 16 of 45
Optimization problem (Mathematical formulation)
Mixed Integer Dynamic Optimization (MIDO)
The bioreactor is treated as dynamic, while the rest of the batch process with algebraic equations, involves also discrete decisions.
Time invariant equality and inequality constraints
Objective function (cost and environmental impact)
Set of differential and algebraic equations (DAEs) describes the dynamic system
Initial conditions
Enforce conditions must be satisfied at specific time instances
Problem posed as a MIDO
Robert Brunet Page 17 of 45
PRIMAL PROBLEM (NLPk)
END
MASTER PROBLEM (MILPk)Determine plant topology
Initial (NLP)Fixed topology
k=k+1Sup. hyp. + int. cuts
No
COM
NLP solver (determine operating conditions)
Set of differential equations (bioreactor model)
Process model1. Mass & energy balances2. Economic & environmental analysis
COM
Yes
NLP worsening?
Reduced space method
Robert Brunet Page 18 of 45
Optimization results Process of L-Lysine Production
Objective function- maximize NPV- minimize Environmental Impact
Decision variable - Threonine Concentration
- Glucose Concentration- Vo reactor- Reaction time- Equiments in parallel (discrete)
Constrains- Production = Demand- Product Purity
NPV improved 13.1%
Combine
Article 1. Hybrid Simulation-Optimization based approach for the Optimal Design of Single-Product Biotechnological Processes. Computers and Chemical Engineering 2012.
Results Base Case
Optimal Case
Net present value [M$] 172.003 195.688
Total capital investment [M$] 101.766 79.885
Operating cost [M$/year] 10.631 8.830
Production rate [ tons MP/year]
6,202 6,202
Batch Throughput [tons MP/batch]
29.647 44.30
Recipe Cycle time [h] 37.51 55.81
Fermentors [equipment] 3 2
Robert Brunet Page 19 of 45
PRIMAL PROBLEM (NLPk)
END
MASTER PROBLEM (MILPk)Determine plant topology
Initial (NLP)Fixed topology
k=k+1Sup. hyp. + int. cuts
Yes
No
New epsilon value
COM
NLP solver (determine operating conditions)
Set of differential equations (bioreactor model)
Process model1. Mass & energy balances2. Economic & environmental analysis
COM
Yes
NLP worsening?
Termination criterionNo
Multi-objective Reduced space method
Robert Brunet Page 20 of 45
PCA
Reduction 2-dimensional Pareto sets
Robert Brunet Page 21 of 45
Minimum EI
EI (YOA )↓Glucose ConsumptionNPV (STY )↑Volume equip. ↑Batch time
Maximum NPV
NPV (STY )↓ Volume equip. ↓ Batch timeEI (YOA )↑Glucose Consumption
Reduced Pareto Set of optimal solutions
Article 2. Cleaner design of single-product biotechnology facilites through the integration of process simulation, multi-objective optimization, LCA and principal component analysis. Industrial & Engineering Chemistry Research 2012.
Robert Brunet Page 22 of 45
3. Thermodynamic Cycles
Robert Brunet Page 23 of 45
Energy consumption increase in the last 25 years
Increase of 66% in the last 25 years
1981: 6,600 Mtones oil eq.
2006: 11,000 Mtones oil eq.
Robert Brunet Page 24 of 45
Optimization of Thermodynamic Cycles
Develop a systematic method for the optimal design of thermodynamic cycles based on the combined use of process simulation and optmization tools
Reduce cycle costsMake a better use of resources
Thermodynamic CyclesPower production Rankine CycleCooling and refrigeration Absorption Cycle
Aim of the work
Robert Brunet Page 25 of 45
Decision variables:
(continuous variables) Pressure, Mass flow, Temperature, Composition(discrete variables) Number of trays, Feed tray
Absorption cooling cycle
Absorber
Pump
Desorber
Condenser & subcooler
Evaporator
Cooling capacity
Robert Brunet Page 26 of 45
PRIMAL PROBLEM (NLPk)
END
MASTER PROBLEM (MILPk)Determines new cycle topology
Initial (NLP)Fixed topology
k=k+1Sup. hyp. + int. cuts
Yes
No
New epsilon value
NLP solver (determine operating
conditions)
Process model
1. Mass & energy balances2. Economic & environmental analysis
COM
Yes
NLP worsening?
Termination criterionNo
Combined Simulation-optimization
Robert Brunet Page 27 of 45
Results Absorption cycle optimization
TAC = 9.35%
Article 3. Combined simulation-optimization methodology for the design of environmental conscious absorption systems. Computers and Chemical Engineering 2012.
Design COP [-] TAC [€/yr] ECO99 [Points]
Cooling
ECO99 0.686 23,445 15,601
Cost 0.629 21,916 16,926
Refrigeration
ECO99 0.516 32,293 20,807
Cost 0.453 28,771 23,451
EI = 7.82%
TAC = 10.90%
TAC = 11.27%
Robert Brunet Page 28 of 45
Decision variables:
Pressure, Mass flows, Temperature
(continuous variables)
Modified Steam Rankine Cycle
Robert Brunet Page 29 of 45
NLP
ENDYes
New epsilon value
NLP solver (determine operating
conditions)
Process model
1. Mass & energy balances2. Economic & environmental analysis
COM
Termination criterionNo
Combined Simulation-optimization
Robert Brunet Page 30 of 45
TAC HH EQ DR
Parallel coordinates plot
minTAC
minHH
minEQ
MinDR
Cost [€] 659.876 689.017 678.386 689.017
HH [Poitns] 18.849 17.901 18.106 17.901
EQ [Points] 10.294 9.881 9.767 9.767
NR [Points] 197.993 189.894 187.646 187.646
EI [Points] 227.136 217.675 215.520 215.314
Min TAC↓ Exchange area↓Turbine size↑Energy consumptionMin impact↑Exchange area↓Energy consumption
Article 4. Minimization of the LCA impact of thermodynamic cycles using a combined simulation-optimization approach. Applied Thermal Engineering 2012.
Robert Brunet Page 31 of 45
4. Biofuels
Robert Brunet Page 32 of 45
Objectives• Reducing the energy consumption of biofuel plants
through their integration with a solar thermal energy system that generates steam
• Bi-criteria NLP for the simultaneous minimization of cost and energy consumption.
• Two different biofuel processes are optimized a alkali-catalyzed biodiesel process using vegetable oil and a dry-grind corn to bioethanol.
Main motivation• Petroleum-based fuels play a vital role in industrial
development, transportation, agricultural sector and many other human needs.
• To be a viable alternative, a biofuel should provide a net energy gain, have environmental benefits, be economically competitive, and be producible in large quantities without reducing food supplies.
Aim of the work
Robert Brunet Page 33 of 45
Process combined with solar collectors
Robert Brunet Page 34 of 45
Computer implementation
Robert Brunet Page 35 of 45
Solar assisted steam generation system
tBkTESkGFHkcolkQQQQQ tktktktktk ,''','',',','
_
',''
_
1','
_
'',
_
,
_
tcolkAGQ tkkttk ,,,
tCAPQ tTES ,
_GFHkLHVmQ kNGk ··
tcolkGTT
aGTT
aTTat
ambt
avt
t
ambt
avtamb
tavtttk
,,
2
321, 0
tcolkTT
T kk INOUTavt
,
2
Robert Brunet Page 36 of 45
Biodiesel production from vegetable oil
Robert Brunet Page 37 of 45
Pareto set of biodiesel production plant
Robert Brunet Page 38 of 45
Summary of the different design alternatives
Article 5. Reducing the environmental impact of biodiesel production from vegetable oil using a solar assisted steam generation system with heat storage. Industrial & Engineering Chemistry Research 2012.
Robert Brunet Page 39 of 45
Dry-grind corn bioethanol production
Robert Brunet Page 40 of 45
Pareto set of bioethanol production plant
Robert Brunet Page 41 of 45
Summary of the different design alternatives
Article 6. Minimization of the energy consumption in bioethanol production processes using a solar assisted steam generation system with heat storage.
Robert Brunet Page 42 of 45
5. Conclusions
Robert Brunet Page 43 of 45
Conclusions
General•A new methodology for optimization of chemical processes based on a combined use of simulation and optimization tools
•The methodology introduces the environmental impact (measured following the LCA principles) in the multi-objective optimization
•Very efficient with “non-standard” unit operations (complex reaction kinetics,…) modeled and optimized via external solver
Bioprocesses•The capabilities of this method have been tested in a typical fermentation process and the production of the amino acid L-lysine. From numerical results, we concluded that it is possible to significantly improve the economic and environmental performance of bioprocesses by optimizing them as a whole.
Robert Brunet Page 44 of 45
Thermodynamic cycles•The capabilities of this approach were tested in two thermodynamic cycles: a steam power cycle and an ammonia-water absorption cooling cycle, for which we minimized the total annualized cost and a set of environmental impacts measured in three LCA damage categories.
Biofuel•We demonstrate the capabilities of this strategy with two case studies in which we address the design of a 12,000 ton/year alkali-catalyzed biodiesel process using vegetable oil modeled in Aspen Plus and a 120,000 tones/year dry-grind corn-to-ethanol production plant modeled in SuperPro Designer.
•The results obtained show that is possible to achieve reductions in environmental impact up to 15 % for the biodiesel and energy consumption of up to 25% for the bioethanol with respect to the minimum cost design.
Conclusions
Robert Brunet Page 45 of 45
Systematic methods based on combined simulation-optimization for the optimal design of chemical processes
Thanks for your Thanks for your attention!attention!
ROBERT BRUNET SOLÉ
Supervisors: Dr. Gonzalo Guillén and Dr. Laureano Jiménez
Department of Chemical Engineering, URV, Tarragona
SUSCAPE