CAPD LYANG 031112 - Carnegie Mellon Universityegon.cheme.cmu.edu/esi/docs/pdf/LLYang_ESI.pdf · 5...
Transcript of CAPD LYANG 031112 - Carnegie Mellon Universityegon.cheme.cmu.edu/esi/docs/pdf/LLYang_ESI.pdf · 5...
Linlin Yang Advisor: Ignacio E. Grossmann
Department of Chemical Engineering
Carnegie Mellon University
March 11, 2012
“Water is the fastest growing market at the moment, with a size of $500 billion globally.” “If nothing is done, there will be a 40 percent gap between supply and demand by 2030.”
2
Water and energy are important resources in the process industries
3 Ahmetovic & Grossmann
Conven>onal water network
Freshwater
Boiler Feedwater treatment
Steam System
Wastewater Boiler Blowdown
Steam
Condensate Losses
Boiler
Water-‐using unit 1
Water-‐using unit 2
Water-‐using unit 3
Raw Water Raw Water Treatment
Wastewater
Cooling Tower
Cooling Tower Blowdown
Water Loss by Evapora2on
Discharge
Wastewater Treatment
Storm Water
Process uses
Other Uses (Housekeeping)
Wastewater
No regenera>on
reuse
No reuse
Process Unit
Contaminants
Process Unit
Contaminants
» Integrated water network with reuse, recycle, and regeneraLon schemes » superstructure is formulated using a nonconvex NLP model
4 Karuppiah & Grossmann (2006); Ahmetovic & Grossmann (2010)
Superstructure based water network design
Freshwater Discharge
Treatment Unit
Contaminants
SpliTer
Mixer
Treatment Unit
Contaminants
5
Freshwater targe>ng formula>on
outinkj
kpj
ij
iin
pout
iout
pin
kinout
ij
kj
insi
ik
outmi
fwifw
ij
ikj
k
outmi
ik
fw
pkpiPUpjCFLCFpiPUpPFpkPUpPF
sksiSUsjCC
skSUsFF
mkMUmjCFCFCF
mkMUmFF
FZ
out
in
in
∈∈∈∀∀=+∈∈∀=
∈∈∀=
∈∈∀∈∀∀=
∈∈∀=
∈∈∀∀+≥
∈∈∀=
=
∑
∑
∑
∈
∈
∈
,,,,,,,
,
,,)(
,s.t.
min
,
max,max,
SpliUers mass balances
Process unit mass balances
Mixer mass balances
This formula>on provides target for a network consists of a set of water-‐using process units using linear constraints
(LP)
AssumpLon: for some contaminant j that reaches its concentraLon upper bound at a given unit, it also reaches the upper bound at all other process units from which reuse streams have non-‐zero flowrate
Goal: determine minimum freshwater consump>on
Use heat and water network formulaLon (MINLP model) to obtain network structure
6 Bogataj & Bagajewicz (2007)
Heat-‐integrated WN reported in the literature
749 conLnuous variables 115 binary variables
7
Extension: heat-‐integrated water network
SUssCjspTtTSUssHispTT
ttcf
TTCFQQc
TT
TTCF
TTt
TTtcfQ
outminjs
pout
inis
p
injs
outjsjsjssCjs
outis
inisisissHisH
poutis
pinisisissHis
mpin
js
mpout
jsjsjssCjsH
out
out
out
out
∈∀∪∈=∀Δ+=
∈∀∪∈=∀=
−Σ−
−Σ+=
−−
−Σ−
Δ−−−
Δ−−Σ≥
∪∈
∪∈
∪∈
∪∈
)(
)(
)(
}],0max{
)},0[max{
}](,0max{
)}(,0[max{
PU1, T1
PU2, T2
Freshwater Tfw
Discharge Tdis
All black streams can par>cipate in heat integra>on
outin
kj
kpj
ij
iin
pout
iout
pin
kinout
ij
kj
insi
ikout
mifw
ifw
ij
ikj
k
outmi
ik
pkpiPUpjCFLCF
piPUpPFpkPUpPF
sksiSUsjCC
skSUsFFmkMUmj
CFCFCF
mkMUmFF
out
in
in
∈∈∈∀∀=+
∈∈∀=
∈∈∀=
∈∈∀∈∀∀=
∈∈∀=
∈∈∀∀
+≥
∈∈∀=
∑
∑
∑
∈
∈
∈
,,,
,,,,
,
,,
)(
,
,
max,max,
fwfwCCHH FcQcQc ++=φ.minObjec>ve fcn
Water targe>ng (LP) Heat targe>ng (LP)
8
Revisit: heat-‐integrated water network u>lity targe>ng
Freshwater Tfw = 20°C
Discharge Tdis = 30°C
PU1 T1 = 40°C
PU3 T3 = 75°C
PU2 T2 = 100°C
PU4 T4 = 50°C
Parameter
CHU ($/kW a) 260 THUin (°C) 126
CCU ($/kW a) 150 THUout (°C) 126
CFW ($/t) 2.5 TCUin (°C) 15
HRAT (°C) 10 TCUout (°C) 20
Use heat and water targe>ng formula>on:
Minimum hea>ng u>lity: 3767 kW Minimum cooling u>lity : No cooling u>lity required Minimum freshwater consump>on: 324 ton/h
Same result as network approach
206 conLnuous variables 229 constraints
9
Simultaneous op>miza>on strategy
HEAT TARGETING
PROCESS FLOWSHEET
WATER TARGETING
Cold streams
Hot streams
MUC*
*MUC – minimum uLlity consumpLon
Hot uLlity
Cold uLlity
MUC*
Water streams
Wastewater
Freshwater
ULlity networks
PROCESS STRUCTURE
WN STRUCTURE HEN STRUCTURE
PROCESS FLOWSHEET WITH HEN AND WN
VvUuXxFvg
QQugvuxgvuxh
FcQcQcvuxF
fwWN
CHHEN
P
fwfwCUj
jC
jC
HUi
iH
iH
∈∈∈
≤
≤
≤
=
+++= ∑∑∈∈
,,
0),(
0),,( 0),,(
0),,( s.t.
),,(.min φ
10
Simultaneous op>miza>on: methanol synthesis from syngas
+ cooling cycle + boiler loop
Freshwater requirement
Hea>ng requirement
Cooling requirement
Duran & Grossmann (1987)
SEQUENTIAL SIMULTANEOUS Profit (1000 $/yr) 62,695 73,416 Investment cost (1000 $) 1,891 1,174 OperaLng parameters
electricity (KW) 6.59 1.84 freshwater (kg/s) 36.43 29.25
heaLng uLlity (109 KJ/yr) 0.293 0 cooling uLlity (109 KJ/yr) 67.3 72.7
Steam generated (109 kJ/yr) 2448 1965 overall conversion 0.68 0.88
Material flowrate (106 kmol/yr) feedstock 48.04 37.13 product 10.89 10.89
11
Sequen>al vs. simultaneous result comparison
17% improvement Solved with BARON 9
12 R. Karuppiah, A. Peschel, I. E. Grossmann, M. Marln, W. son, and L. Zullo, “Energy opLmizaLon for the design of corn-‐based ethanol plants,” AIChE Journal, vol. 54, no. 6, 2008, pp. 1499–1525.
Example 2: Bioethanol produc>on
Water Consump>on and genera>on
Fermenta>on
Solid Removal
Water/EtOH Separa>on
Pretreatment
13
Water network superstructure
Cjin,max (ppm) TSS TDS ORG
Boiler loop 2 100 10
Cooling cycle 10 500 10
1-‐Βjt
Screens 95% 0 0
Reverse osmosis 0 90% 0
Anaerobic tank 0 0 99%
PU 1
Cooling requirement (Qc)
Cooling Tower
Freshwater
blowdown
boiler
HeaLng requirement (Qs)
Steam loss Steam
blowdown
Coldwater outlet
Warmwater inlet
Condensate
Feedwater
PU 3
Screens
ANA
Reverse osmosis
PU 2
Formula>on ˃ Dew point equaLon -‐ condenser
temperature ˃ Bubble point equaLon -‐ feed and reboiler
temperature ˃ Fenske equaLon -‐ # of trays ˃ Watson's equaLon – heat of vaporizaLon ˃ Mass balance ˃ Energy balance
Assump>ons ˃ Constant relaLve volaLlity ˃ Ideal soluLon ˃ Water is the only component contribuLng
to heat of vaporizaLon ˃ Temperature change due to pumps is
negligible 14
Mul>effect columns
Feed
Distillate
Bottom
1 atm Feed
LP column
HP column
Distillate
Bottom
Result
NLP solver: CONOPT 3 MINLP solver: BARON 9 GAMS 23.7
15
Reboiler duty reduced by ~36% by with mul>effect column
Even though the objecLve funcLon did not improve using simultaneous method, we can see that the soluLon Lme did not increase drasLcally
No integra>on
Sequen>al single column
Sequen>al w/ mul>effect
Simultaneous w/ Mul>effect
Cost (MM$/yr) 14.91 11.77 8.57 8.57
Cooling water use (kg/s) 2895.6 1998.3 1127.3 1124.8
Freshwater use (kg/s) 40.8 127.6 90.0 90.0
Steam use (kg/s) 35.1 28.3 21.2 21.3
CPU(s) 387 387 470 563
# eqns 2,232 2,232 3,213 5,221
# cont var 2,921 2,921 3,914 5,392
Cooling water
U>lity integra>on – power, water, & heat
blowdown
Makeup water
H1-C1
H1-C2
H2-C1
H2-C2
H1-C1
H1-C2
H2-C1
H2-C2
Stage 1 Stage 2
Temperature location k=1
Temperature location k=2
Temperature location k=3
s1
s2
W1
W2
C1
C2
H1
H2
tC1,1 tC1,2 tC1,3
tC2,1 tC2,2 tC2,3
tH1,1 tH1,2 tH1,3
tH2,1 tH2,2 tH2,3
CoolersHeatersblowdown
blowdown
Condensate return
HP MP LP
Warm water return
Raw Water
SiO2 Removal
discharge
Condensate loss
Cooling Tower
Process Load (HEN)
Sand Filter
Reverse Osmosis
Water Treatment
Scrubber
ULlity System
16
fwfws
ssst d
stdtur
d
extd
fixext
st d
std
fixtur
st
stbb
st
stb
fixb FcFcWcYcYcFcYc ++++++= ∑∑∑∑∑∑∑∑ varvarφ
U>lity system § Logical constraints § Demand constraints § Power balances § Mass balances
17
Problem statement
ULlity system • Existence of boiler • Existence of turbine • Back pressure turbine • ExtracLon turbine (addiLonal cost $20,000)
• Flowsheet power demand (7500kW) • 70% condensate return
HEN • 2 hot streams/ 2 cold streams • Inlet and outlet temperature can vary within +/-‐ 10 K
• Heat capacity flowrate can vary within 20%
• Two treams have assigned costs • Hot uLlity -‐ HP, MP, and LP steam • Cold uLlity -‐ cooling water
WN • HP boiler has more stringent feedwater requirement
• HP boiler/MP boiler have different blowdown rates
• RO consumes electricity • Raw water needs treatment • TSS, TDS, GAS present in freshwater • Discharge limit imposed
Water network § Mass balances § Power demand constraint
Boiler cost Turbine cost freshwater cost
Flowsheet stream cost
Objec>ve func>on
Mul>ple hot u>lity targe>ng (Duran & Grossmann) § HeaLng uLliLes targets § Cooling uLlity target
18
Result
Sequen>al Simultaneous Cost (1000 $ / yr) 884.2 641.5
U>lity
HP boiler flowrate (kg/s) Yes 17.66 Yes 18.20
MP boiler flowrate (kg/s) No No
Power demand external (kW) HP à LP 7500 ExtracLon 7500
Reverse osmosis power demand (kW) MP à LP 62.0 MP à LP 63.89
HEN U>lity (kW)
Cooling 1463.8 751.1
HP steam 3820.2 5727.2
MP steam 13628.2 21065.7
LP steam 4743.4 19110.2
Fcp,H1 (kW/K) 48 32
Fcp,C2 (kW/K) 144 216
WN flowrate (kg/s)
Freshwater 7.26 6.47
Sand filter 7.2 6.4
Reverse osmosis 5.6 5.8
Scrubber 2.4 1.2
» Developed LP formulaLons for targeLng minimum freshwater consumpLon for a set of water-‐using process units under a specific condiLon
» Extended the water targeLng formulaLon to nonisothermal water network
» Targe>ng method can be used to improve objec>ve func>on and computa>onal effort under the simultaneous approach for flowsheet op>miza>on
» The interac>on among power use, heat use, and water use can be exploited to achieve beUer flowsheet design
19
Conclusion
Thank you!