Optimal synthesis of batch separation processes
description
Transcript of Optimal synthesis of batch separation processes
Optimal synthesis of batch separation processes
Taj Barakat and Eva Sørensen
University College London
iCPSE Consortium Meeting, Atlanta, 30-31 March 2006
2
Motivations
Many valuable mixtures are difficult to separate
Need to optimise efficiency of current processes
Select most economical separation process Explore novel techniques and alternatives
3
Objectives
Development of models/superstructure to determine the best design configuration, operating policy and control strategy for hybrid separation (distillation/membrane) processes.
Develop general guidelines for design, operation and control of such processes
4
Project Features
Economics objective function Rigorous dynamic models Encompassing (most of) the available
decision variables Considering novel configurations
5
Outline
1. Optimal synthesis of batch separation processes
2. Multi-objective optimisation of batch distillation processes
3. Concluding remarks
6
Optimal synthesis of
batch separation processes
7
Configuration Decisions
Separation problem
Process Superstructure
?
Batch Distillation Batch Pervaporation Batch Hybrid
8
Design and Operation DecisionsDesign Alternatives Operational Alternatives
Mincapital
cost
Minrunning
cost
• Trays• Membrane stages• Membrane modules
• Vapour loading rate• Reflux/reboil ratios• Recovery/No. batches• Withdrawal rate• Task durations
9
Process Superstructure
Feed
Retentate
Permeate
Offcut
Nt
Rc
Qr
Rp
Ns , Nm,s
P
Rr
Lr
Fs
Qs
10
Batch Distillation
Product 1
Product 2
Offcut
Reboiler
Nt
Rc
Qr
Rp
11
Batch Pervaporation
Offcut
Feed
Separation Stage
Retentate
Permeate
Ns
Nm,s Rr
Rp
P
Qf
12
Hybrid Distillation I
Feed
Product
PermeateReboiler
Offcut
Nt
Rc
Qr
RpP
Ns Nm,s
13
Hybrid Distillation II
Feed
Retentate
Permeate
Offcut
Nt
Rc
Qr
Rp
P
Ns Nm,s
14
Hybrid Distillation III
Retentate
Permeate
Offcut
Feed
Nt
Rc
Qr
Rp
Ns
, Nm,s
P
Rpr
Lr
Fs
Rr
15
Problem Formulation – Objective Function
Maximise
Annual Profit = Revenues – Operating Costs
Batch Processing TimeAv. Time – Capital Costs
Subject to :
Model equations DAE/PDAE, nonlinear
Design variable bounds discrete and continuous
Operational variable bounds continuous
To determine :
Design variables
Operation variables (time dependent)
Nonlinear, (OC/CC, Guthrie’s correlations)
16
Problem Formulation - Solution
DAE gPROMS (Process Systems Enterprise Ltd., 2005)
MIDO Genetic Algorithm (GA)
• Mixed integer dynamic optimisation (MIDO) problem • Complex search space topography (local optima, nonconvex)• Need robust, stable and global solution method
17
Optimisation Implementation
GeneticAlgorithmModule
Batch Distillation/Pervap
Model
ThermodynamicsModel
Genome Set
Model State
Simulation Output
Physical Properties
GAlib
gPROMS
Multiflash
18
Case Study
19
Case Study ( Acetone – Water )
Separation of a binary tangent-pinch mixture Acetone dehydration system ( 70 mol % acetone feed ) 20,000 mole feed
Subject to: Purity ≥ 97% Recovery ≥ 70%
Maximise: Annual profit
Assuming: Single membrane stage Single retentate recycle location
20
Case Study Superstructure
Retentate
Permeate
Offcut
Feed
Nt
Rc
Qr
Rp
NsNm,s
P
Rr
Lr Fs
21
Optimal Process - Hybrid
Feed
Retentate
Permeate
Offcut
Rp
0.79 – 1.8%
1.00 – 96.3%
0.88 – 1.9%
Rr
1.00 – 1.8%
0.83 – 96.3%
0.24 – 1.9%Lr =3
Nt = 30
Fs = 9
VReb = 5 mole/s
Fside = 2.5 mole/s
P = 300 Pa
Nm = 2
Profit 18.07 M£/yr
tf = 5119 s
To = 330 K
22
Fixed Configuration – Distillation only
Product 1
Product 2
Offcut
Reboiler
Rp
1.00 – 0.10%
1.00 – 99.7%
0.00 – 0.20%
Rr
1.00 – 0.10%
0.68 – 99.7%
0.70 – 0.20%
Nt = 30
VReb = 5 mole/s
tf = 8964 s
Profit 14.30 M£/yr
-26%
23
Case Study Summary
Approach for process selection based on overall economics
Allows determination of best process alternative for maximum overall profitability
Company specific costing can easily be included
24
Multi-objective optimisation of
batch distillation processes
25
Batch Distillation
Product 1
Product 2
Offcut
Reboiler
Nt
Rc
Qr
Rp
26
Problem Formulation – Objective Function
Minimise
Investment Costs
Subject to :
Model equations DAE/PDAE, nonlinear
Design variable bounds discrete and continuous
Operational variable bounds continuous
To determine :
Design variables
Operation variables (time dependent)
Minimise
Operating Costs&
27
Optimisation
Single-objective optimisation:
To find a single optimal solution x* of a single objective function f(x)
Multi-objective optimisation:
To find array of “Pareto optimal” solutions with respect to multiple objective functions
xx*
f(x)
0
28
Multiobjective Optimization Problem
))(...,),(),(()( 21 xxxxf kfffMaximize
Xxsubject to
Several Pareto-optimal sets Pareto Optimal Solutions
Min
imis
e
Minimise)(1 xf
)(2 xf
29
Ranking
3
2)(1 i
ni k
gf
c if solution is infeasible
if solution is feasible but dominated
if solution is feasible and non-dominated
30
Ranking
3
F2
F1
3
better
bett
er
3
2
22
2
Max = 1
3
3
3
31
Problem Formulation - Solution
DAE gPROMS (Process Systems Enterprise Ltd., 2005)
MO-MIDO Multi-Criteria Genetic Algorithm (MOGA)
• Multi-objective Mixed integer dynamic optimisation (MO-MIDO) problem• Need robust, stable and global solution method
32
Case Study
33
Case Study ( Acetone – Water )
Separation of a binary tangent-pinch mixture Acetone dehydration system ( 70 mol % acetone feed ) 20,000 mole feed
Subject to: Purity ≥ 97% Recovery ≥ 70%
Minimise: Investment costs Annual operating costs
34
Case Study Summary
35
Case Study Summary
Approach for multi-criteria process optimisation using Genetic Algorithm
Allows determination of process alternatives through Pareto optimality
Company specific costing can easily be included
36
Concluding RemarksFor hybrid batch separation processes: Optimum synthesis and design procedure Multi-criteria optimisation
Simple extension to continuous hybrid processes