Integrarea si Intensificarea Proceselor Chimice
Transcript of Integrarea si Intensificarea Proceselor Chimice
1
Integrarea si Intensificarea Proceselor Chimice
Michael Baldea The University of Texas at Austin
Facultatea de Chimie si Inginerie
Chimica
Universitatatea “Babes-Bolyai”
Cluj, 18 martie, 2015
TexPoint fonts used in EMF.
Read the TexPoint manual before you delete this box.: AAAAA
α
Tin
Ti
TC
TH
z=L z=0
A→B
TR Texit
QH
QR
Process and Energy Systems Engineering
Context
Process Integration (IP):
A holistic approach to process design, based on maximizing
recovery and utilization of energy and materials from within the
process, reducing the use of utilities and minimizing environmental
impact (cf. Hallale, 2001, Cussler and Moggridge, 2001, El-Halwagi,
2006)
2
Process intensification (PI):
“Any chemical engineering development that leads to substantially
smaller, cleaner, safer and more energy efficient technology”
(Reay et al., 2013) or “that combine[s] multiple operations into
fewer devices.” (Tsouris and Porcelli, 2003)
2
Process and Energy Systems Engineering
Context
Process intensification (PI):
“Any chemical engineering development that leads to substantially
smaller, cleaner, safer and more energy efficient technology” (Reay
et al., 2013) or “that combine[s] multiple operations into fewer
devices.” (Tsouris and Porcelli, 2003)
Multum in parvo (Lat.) : much in little
3
Stankiewicz and Moulijn, 2003 Reay et al., 2013
Process and Energy Systems Engineering
Context
Process intensification (PI):
“Any chemical engineering development that leads to substantially
smaller, cleaner, safer and more energy efficient technology” (Reay
et al., 2013) or “that combine[s] multiple operations into fewer
devices.” (Tsouris and Porcelli, 2003)
Multum in parvo (Lat.) : much in little
Paradigm
• Process should be governed by intrinsic rates
• Identify limiting factor(s) in a process (transport, transfer)
• Address them via changes in system operation (batch
continuous), device geometry, external energy fields
• Scale-up by “numbering-up”
4
3
Process and Energy Systems Engineering
PI: Multiple Phenomena, Scale-Independent
5
www.velocys.com
Courtesy of Bailee Roach
A,B,C
A
B
c
A,B,C
A
c
B
Reaction
mm channels
<1m length
cm diameter
Several meters tall
Separation
Integrated Intensified
ΔH<0
ΔH>0
Process and Energy Systems Engineering
PI: Multiple Phenomena, Scale-Independent
6
A,B,C
A
B
c
A,B,C
A
c
B
Reaction
mm channels
<1m length
cm diameter
Several meters tall
Separation
Integrated Intensified
ΔH<0
ΔH>0
30% capital savings,
use up to 40% less
energy
10x smaller size
Schultz et al., CEP, 2002, Kiss and
Bildea, CEPPI, 2011
Zanfir and Gavriilidis, CES, 2003
PI practice ahead of theory
4
Process and Energy Systems Engineering
0 2 4 6 8 100
2
4
6
V, m
30 2 4 6 8 10
a.u
. CAPEX OPEX
R/Fo
I II III
Integration vs. Intensification
7
RI
Fo,CAout=0
I
RII
Fo,CAo
II
Fo,CAout=0
Fo,CAo
III
Fo,CAout=0
Baldea, FOCAPD, 2014
Process and Energy Systems Engineering
Integration vs. Intensification
• Fundamental changes in design, operation 8
RI
Fo,CAout=0
I
RII
Fo,CAo
II
Fo,CAout=0
Fo,CAo
III
Fo,CAout=0
• “The front-runner of
industrial process
intensification” (Harmsen, 2007)
5
Process and Energy Systems Engineering
Integration vs. Intensification: Empirical
• Reduced number of units
• Reduced unit size and
holdup
• Reduced OPEX (no
recycling)
BUT
• Reduced number of degrees
of freedom for design and
control
9
RI
Fo,CAout=0
I
RII
Fo,CAo
II
Fo,CAout=0
Fo,CAo
III
Fo,CAout=0
Schembecker and Tlatlik, 2003; Nikacevic et al., 2012
Process and Energy Systems Engineering
Heat Integration vs. Intensification
• Often safety-critical
• Ample instrumentation and
actuation
• Multiple design DOF, several
possible control configurations
10
• Microchannel reactors
• Exothermic and endothermic
reactions occur in parallel
channels
ΔH<0
ΔH>0
ΔH>0
ΔH<0
ΔH>0
ΔH<0
6
Process and Energy Systems Engineering
Prototype System: Methane Steam Reformer
Assumptions
• Gas phase: 2D convection-diffusion-reacting flow (laminar)
• Wall: 2D heat conduction
• Catalyst layers: 1D reaction-diffusion
• Boundary conditions: no flux (outlet), equal flux (fluid-solid interface),
symmetry (channel center)
Implementation: gPROMS
• Discretization: axial: finite differences, radial: OCFEM
11
Zanfir and Gavriilidis, 2003, Zanfir et al., 2011, Pattison and Baldea, 2013
Process and Energy Systems Engineering
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1800
900
1000
1100
1200
1300
1400
1500
Dimensionless z
Tem
pera
ture
(K
)
Pre-disturbance
Post-disturbance
Temperature Control Problem
12
∆H>0
∆H<0
Reactor subject to operational disturbances:
Flow rate
Pressure
Temperature
Composition
May result in formation of “hotspots” that damage the reactor or catalysts
Goal: develop distributed control concepts that prevent the formation of hotspots
Pattison, R.C. and Baldea, M., AIChE J., 2013
7
Process and Energy Systems Engineering
Concept 1: Segmented Catalyst
13
Alternating catalytically active and inactive segments
Emulates a distributed feed system : increase number of design degrees of freedom
Modulates the rate of heat generation axially
Formulate optimization to select:
the optimal parametric temperature trajectory and
the optimal catalyst segmentation to track the trajectory
Pattison, R.C. and Estep, F.E. and Baldea, M., IECR., 2013
Tem
pera
ture
z
Base profile
Desired profile
Eigenberger et al., CES, 2000
Process and Energy Systems Engineering
Catalyst Segmentation Results
14
Base Case
Segmented Catalyst
200oC reduction in maximum temperature
Higher average temperature
Increase methane conversion in both sets of channels
Pattison, R.C. and Estep, F.E. and Baldea, M., IECR., 2013
8
Process and Energy Systems Engineering
Concept 2: Thermal Fuse
15
T
z
∆H>0
∆H<0
Tmelt
z
Phase change material (PCM) layer absorbs heat at constant temperature when melting, prevents formation of hotspots
Pattison, R.C. and Baldea, M., AIChE J., 2013; Patent WO2014042800 A1
Process and Energy Systems Engineering
16
PCM Controller Tuning
Single degree of freedom: PCM layer thickness
• Geometry is fixed and must be determined at design stage
• Online tuning is not possible
Model-based controller design:
• Shape dynamic behavior (melting/solidification time) via
stochastic optimization of PCM size
• Identification-based optimization: represent disturbances as
multi-level random signals, impose on system model during
dynamic optimization iterations
ΔH<0
ΔH>0
ref ref v =v (v,σ)
Wang and Baldea, Comput. Chem. Eng., 2014
9
Process and Energy Systems Engineering
Optimization Under Uncertainty
17
• Problem θ: Uncertain variables
d: design variables
u: controls
Pistikopoulos, 1995
Infinite-dimensional problem due to 𝜽, 𝒖
- Discretization to convert to finite-dimensional (MI)NLP
- Problem solution:
Multi stage (design vs. resource variables), generalized Benders
decomposition
- Chance constrained formulation, parametric programming
Acevedo and Pistikopoulos, 1997, Clay and Grossmann, 1997, Dua et al., 2002,
Benerjee and Ierapetritou, 2002, Li et al., 2008, Faisca et al., 2008
min 𝒖,𝒅
𝐸(𝐽 𝒙 , 𝒙, 𝒖, 𝜽 𝑡 , 𝑡 )
s.t. 𝒉 𝒙 , 𝒙, 𝒖, 𝜽 𝑡 , 𝑡 = 0
𝒈 𝒙 , 𝒙, 𝒖, 𝜽 𝑡 , 𝑡 ≤ 0
Process and Energy Systems Engineering
Discretization
18
-4 -3 -2 -1 0 1 2 3 4-4 -3 -2 -1 0 1 2 3 4
• Sample Average Approximation
(SAA)
Ierapetritou et al., 1996, Acevedo and Pistikopoulos, 1998 Linderoth et al.2006, Constantinescu et al., 2008
• Infinite-dimensional problem converted to (MI)NLP
• Computationally expensive when number of scenarios increases
• Similar to a system identification experiment
min𝒖,𝒅
𝑤𝑝
𝑃
𝑝=1
𝐽 𝒙 , 𝒙, 𝒖, 𝒅, 𝜽𝑝, 𝑡
s.t. ∀𝑝 𝒉 𝒙 , 𝒙, 𝒖, 𝜽𝑝 , 𝑡 = 0
𝒈 𝒙 , 𝒙, 𝒖, 𝜽𝑝 , 𝑡 ≤ 0
min𝒖,𝒅
1
𝑁 𝐽 𝒙 , 𝒙, 𝒖, 𝒅, 𝜽𝑘 , 𝑡
𝑁
𝑘=1
s.t. ∀𝑘 𝒉 𝒙 , 𝒙, 𝒖, 𝜽𝑘 , 𝑡 = 0
𝒈 𝒙 , 𝒙, 𝒖, 𝜽𝑘 , 𝑡 ≤ 0
• Multiperiod
10
Process and Energy Systems Engineering
Identification-Based Optimization
19
• Nonlinear systems: pseudo-random multi-level signal
(PRMS)
Step testing: generate scenarios
to quantify input/output
relationship
Train of signal changes (e.g.:
PRBS): increase efficiency
Godfrey, 1993, Barker., 2004
• IBO concept:
- represent uncertain variables as pseudo-random
multi-level signals (PRMS)
- apply PRMS to efficiently evaluate the response of
the system to fluctuations in uncertain variables.
Process and Energy Systems Engineering
Pseudo-Random Multi-Level Signal representation:
• time-unfolding using Galois Sequence support
N
PCM max,wall,j melt
j=1 0
ref,j ref
1h =argmin (T -T )ds
N
s.t. reactor model equations
v =v (v,σ)
St
Problem Reformulation
20
PRMS-based formulation
PCM max,wall melt
0
ref
h =argmin (T -T )d
s.t. reactor model equations
v (t)=v
MPRSt
MPRS
Wang and Baldea, Comput. Chem. Eng., 2014
𝑁𝑡𝑠 ≤ 𝑡𝑀𝑃𝑅𝑆
11
Process and Energy Systems Engineering
Identification-Based Optimization: Algorithm
21
Make initial guess for
optimization
variables
Set initial
conditions
for model
Integrate for given time
horizon
Reset system states
to initial conditionOptimal? Display results
Calculate the gradients of
the objective function and
constraints
Integrate for given time
horizon
Compute new decision
variablesCheck optimality criteria
and constraints
NO
YESIn
itia
l ite
ratio
nIte
ratio
ns 2
...N
Generate MPRS
of all disturbances
• Assumptions
- Time horizon T is sufficiently long
- System is feedback stabilizable
- Disturbance distributions known
PRMS
min𝒅,𝒌
𝐽 𝒙 , 𝒙, 𝒖, 𝒅, 𝜽𝑃𝑅𝑀𝑆(𝑡), 𝑡
s.t. 𝒉 𝒙 , 𝒙, 𝒖, 𝒅, 𝜽𝑃𝑅𝑀𝑆(𝑡), 𝑡 = 0
u=k(x)
𝒈 𝒙 , 𝒙, 𝒖, 𝒅, 𝜽𝑃𝑅𝑀𝑆(𝑡), 𝑡 ≤ 0
0<t<T
Wang and Baldea, Comput. Chem. Eng. 2014
Process and Energy Systems Engineering
Thermal Fuse: Results
22
Segmented Catalyst: reduced nominal temperature profile
0.8 mm Thermal Fuse: prevents the formation of hotspots
Cannot account for sustained disturbances Pattison and Baldea, AICHE J. 2013
12
Process and Energy Systems Engineering
Application: Stranded and Associated Gas
Stranded gas: Small remote
deposits that are too expensive to
extract & transport to consumers
7000 tcf worldwide
Associated gas: found with oil,
typically flared or reinjected
4.5 trillion megajoules wasted
in 2011
Monetization: pipeline,
liquefaction, Gas-To-Liquids
(GTL):
Not viable at small scales
23
image source: www.gereports.com
Microchannel reactor-based
systems can have large impact
Process and Energy Systems Engineering
Intensification of Systems without Recycle
• Dividing wall column (DWC): separate ternary
mixture using a single physical device
Courtesy of Bailee Roach
A,B,C
A
B
c
A,B,C
A
c
B
13
Process and Energy Systems Engineering
DWC at the University of Texas at Austin
University of Texas at Austin
Pickle Research Center
August 2014
Project Partners:
UT, Eastman, Emerson,
Packing from Sulzer, Koch-Glitsch
Courtesy of Bailee Roach
Process and Energy Systems Engineering
DWC Pilot Plant
Support El 21' 5.5"
3rd
Platform El 35' 6"
2nd
Platform El 23' 6"
Sump
Section
Stripping
Section
Lower DW
Section
Upper DW
Section
Rectifying
Section
Head
1
2
49
50
34 56
7
8
910
12 11
1416 13 15
2830 27 29
2122 23
1718
26 25
32 31
36 35 37
40 39
41 43
45
4746
48
51
24
3334
38
1920
4244
52
Courtesy of Bailee Roach
14
Process and Energy Systems Engineering
PI: Future Challenges
Theory: break unit ops paradigm
Synthesis of intensified processes: Phenomena-
based Superstructure? (Ismail et al., 2001, Arizmendi-
Sánchez and Sharratt, 2008, Lutze et al., 2013)
Flowsheet “co-simulation” / optimization (Lang et al.,
2009)
Embed control capability 27
A,B,C
A
B
c
A,B,C
A
c
B ΔH<0
ΔH>0
Models readily available in
process simulator “parts bin” ?
Process and Energy Systems Engineering
28
Conclusions
• Intensification fosters dynamic complexity
- Better economics/improved efficiency: more difficult control
- Scale independent
• Accomplishments
- “cool” applications and commercial success
• Future
- Theory: new process synthesis, simulation, optimization framework; will likely lead to new applications
- Embed control considerations at the control stage
- Applications: smarter manufacturing, interaction with power system
15
Process and Energy Systems Engineering
• Students, in particular R.C. Pattison and S. Wang
• American Chemical Society- Petroleum Research Fund
• US Department of Energy, Advanced Manufacturing Office
• W.A. “Tex” Moncrief Grand Challenges Award
• Eastman Chemical, Emerson Process Management, ABB, Dow Chemical, Corning, Center for Operator Performance, Texas Wisconsin California Control Consortium
29
Acknowledgements