Multi-­scale  modeling  of  carbon  capture  systemsDavid  C.  Miller,a,*  Joel  D.  Kress,e David  S.  Mebane,g Xin  Sun,f Curtis  B.  Storlie,e Sankaran Sundaresan,h Debangsu Bhattacharyya,g Nikolaos  V.  Sahinidis,b John  C.  Eslick,b Charles  H.  Tong,d Deb  A.  AgarwalcaNational Energy  Technology  Laboratory,  Pittsburgh,  PA  USA,  bCarnegie Mellon  University,  Pittsburgh,  PA  USA,  cLawrence Berkeley  National  Laboratory,  Berkeley,  CA  USA,  dLawrence Livermore  National  Laboratory,  Livermore,  CA  USA,  eLos Alamos  National  Laboratory,  Los  Alamos,  NM  USA,  fPacificNorthwest  National  Laboratory,  Richland,  WA  USA,  gWest Virginia  University,  Morgantown,  WV  USA,  hPrinceton University,  Princeton,  NJ  USA

8-­9  September  2015

2

Challenge:  Accelerate  Development/Scale  Up

2010                              2015                              2020                              2025                              2030                              2035                              2040               2045                              2050

1  MWe1  kWe 10  MWe 100  MWe 500  MWe

Traditional   time   to  deploy  new  technology  in  the  power  industry  

Laboratory  Development10-­15  years

Process  Scale  Up20-­30  years

Accelerated  deployment   timeline

Process  Scale  Up15  years

1  MWe

10  MWe

500  MWe

100  MWe

3

For Accelerating Technology Development

National  Labs Academia Industry

Rapidly  synthesize  optimized  processes  to  identify  promising  

concepts

Better  understand  internal  behavior    to  reduce  time  for  troubleshooting

Quantify  sources  and  effects  of  uncertainty   to  guide  testing  &  reach  larger  scales  faster

Stabilize  the  cost  during  commercial  

deployment

3

4

• Develop new  computational  tools  and  models  to  enable  industry  to  more  rapidly  develop  and  deploy  new  advanced  energy  technologies– Base  development  on  industry  needs/constraints

• Demonstrate the  capabilities  of  the  CCSI  Toolset  on  non-­proprietary  case  studies– Examples  of  how  new  capabilities  improve  ability  to  develop  capture  technology

• Deploy the  CCSI  Toolset  to  industry– Initial  licensees,  CRADA

Goals  &  Objectives  of  CCSI

5

Advanced  Computational  Tools  to  Accelerate  Carbon  Capture  Technology  Development

Lab  &  Pilot  ScaleExperiments  &  Data

Device Scale  Models  Validated  3-­D,  CFD

Process  SystemsDesign,  Optimization  &  Control

Physical  PropertiesKinetics

Thermodynamics

6

Data  ManagementFramework

Process  Models

  Solid  In

Solid  Out

Gas  In

Gas  Out

Utility  In

Utility  Out

CCSI  Toolset  Workflow  and  Connections

Basic  Data  Submodels

Optimized  Process

GHX-­‐001CPR-­‐001

ADS-­‐001

RGN-­‐001

SHX-­‐001

SHX-­‐002

CPR-­‐002

CPP-­‐002ELE-­‐002

ELE-­‐001

Flue  Gas

Clean  Gas

Rich  Sorbent

LP/IP  Steam

HX  Fluid

Legend

Rich  CO2  Gas

Lean  Sorbent

Parallel  ADS  Units

GHX-­‐002

Injected  Steam

Cooling  Water

CPT-­‐001

1

2

4

7

8

5 3

6

9

10

11

S1

S2

S3

S4

S5

S6

12

13

14

15

16

17

18

19

21

24

2022

23

CYC-­‐001

Algebraic  Surrogate  Models

Uncertainty

Uncertainty

Uncertainty

Uncertainty

Superstructure  Optimization

Uncertainty

Simulation-­Based  Optimization

Uncertainty

Process  Dynamics  and  

Control

CFD  Device  Models

Uncertainty  Quantification

Framework  forOptimization

Quantification  of  Uncertainty  &  Sensitivity

7

Solid  Sorbent:  Process  Systems  Example

Lab  &  Pilot  ScaleExperiments  &  Data

Device Scale  Models  Validated  3-­D,  CFD

Process  SystemsDesign,  Optimization  &  Control

Physical  PropertiesKinetics

Thermodynamics

8

Basic  Data  Model

Alkylammoniumcarbamate formed  by  two  molecules  of  tetraethylenepentamine (TEPA)  and  CO2.

dx

dt= k

x

hs2p

c

� xw/x

i(1)

db

dt= k

b

hsap

c

� bw/b

i(2)

da

dt= k

a

hph

� a/a

i� k

b

hsap

c

� bw/b

i(3)

1 = s+ x+ w (4)

w = x+ b/nv

(5)

TGA wt. % =⇥Mn

v

(x+ b) +H(b+ a)⇤/⇢ (6)

x ⌘ carbamate site fraction (7)

s ⌘ free amine site fraction (8)

w ⌘ alkylammonium site fraction (9)

b ⌘ bicarbonate concentration (10)

a ⌘ adsorbed water concentration (11)

pc

⌘ partial pressure CO2 (12)

ph

⌘ partial pressure H2O (13)

nv

⌘ site density of amines (14)

kx

, kb

, ka

⌘ rate constants (15)

x

, b

, a

⌘ equilibrium constants (16)

x

= exph S

x

RT

iexp

h�H

x

RT

i/P (17)

b

= exph S

b

RT

iexp

h� H

b

RT

i/P (18)

a

= exph S

a

RT

iexp

h�H

a

RT

i/P (19)

kx

= ⇣x

exph�H‡

x

RT

i(20)

kb

= ⇣b

exph�H‡

b

RT

i(21)

ka

= ⇣a

exph�H‡

a

RT

i(22)

1

9

Process  ModelsBubbling  Fluidized  Bed  (BFB)  Model• 1-­D,  nonisothermal with  heat  exchange• Unified  steady-­state  and  dynamic• Adsorber  and  Regenerator• Variable  solids  inlet  and  outlet   location• Modular   for  multiple  bed  configurations

Moving  Bed  (MB)  Model• 1-­D,  nonisothermal with  heat  exchange• Unified  steady-­state  and  dynamic• Adsorber  and  Regenerator• Heat  recovery  system

Compression  System  Model• Integral-­gear   and  inline  compressors• Determines  stage  required  stages,  intercoolers• Based  on  impeller   speed  limitations• Estimates  stage  efficiency• CO2 drying  (TEG    absorption  system)• Off-­design  performance.• Includes  surge  control  algorithm

0 0.2 0.4 0.6 0.8 10.04

0.06

0.08

0.1

0.12

0.14

z/L

yb (k

mol

/km

ol)

CO2H2O

0 0.5 1200

400

600

800

1000

kmol

/hr

z/L

Solid Component Flow Profile of MB Regenerator

Bicarb. Carb. H2O

10

• Discrete  decisions:     How  many  units?  Parallel  trains?  What  technology  used  for  each  reactor?

• Continuous  decisions:  Unit  geometries• Operating  conditions:    Vessel  temperature  and  pressure,  flow  rates,  

compositions

Carbon  Capture  System  Configuration

Surrogate  models  for  each  reactor  and  technology  used

11

Superstructure  Optimization

Optimal   layout  

Mixed-­integer  nonlinear  programming  model  in  GAMS• Parameters• Variables• Equations

• Economic  modules• Process  modules

• Material  balances• Hydrodynamic/Energy  balances

• Reactor  surrogate  models• Link  between  economic  modules  and  process  modules

• Binary  variable  constraints• Bounds   for  variables

12

Optimization  &Heat  Integration

w/o heat  integration Sequential Simultaneous

Net power efficiency (%) 31.0 32.7 35.7Net power output (MWe) 479.7 505.4 552.4Electricity  consumption b (MWe) 67.0 67.0 80.4

Base case w/o CCS: 650 MWe, 42.1 %Chen,  Y.,  J.  C.  Eslick,  I.  E.  Grossmann  and  D.  C.  Miller  (2015).  "Simultaneous  Process  Optimization  and  Heat  Integration  Based  on  Rigorous  Process  Simulations."  Computers  &  Chemical  Engineering.  doi:10.1016/j.compchemeng.2015.04.033

Objective:    Max.  Net  efficiencyConstraint:    CO2 removal ratio  ≥  90%  Decision  Variables   (17):  Bed  length,  diameter,  sorbent  and  steam  feed  rate

13

Uncertainty  Quantification  for  Prediction  Confidence§ Now  that  we  have

• A  chemical  kinetics  model  with  quantified  uncertainty• A  process  model  with  other  sources  of  uncertainty• Surrogates  with  approximation  errors• An  optimized  process  based  on  the  above

§ UQ  questions• How  do  these  errors  and  uncertainties  affect  our  prediction  confidence  (e.g.  operating  cost)  for  the  optimized  process?

• Can  the  optimized  system  maintain  >=  90%  CO2  capture  in  the  presence  of  these  uncertainties?

• Which  sources  of    uncertainty  have  the  most  impact  on  our  prediction  uncertainty?

• What  additional  experiments  need  to  be  performed  to  give  acceptable  uncertainty  bounds?

CCSI  UQ  framework  is  designed  to  answer  these  questions

14

Perform  statistical  analyses  with  FOQUSEnsemble  AnalysesØUncertainty  analysisØSensitivity  analysisØCorrelation  analysisØScatterplots  for  visualization

Response  Surface  (RS)  AnalysesØRS  validationØRS  visualizationØRS-­based  uncertainty  analysisØRS-­based  sensitivity  analysisØRS-­based  Bayesian  inference

Ensemble  UA

RS-­based  UA

RS-­based  SA

RSvalidation

RS  visualizationRS-­based  Bayesian  inference

15

Solid  Sorbents:  Validated  CFD  Model  Example

Lab  &  Pilot  ScaleExperiments  &  Data

Device Scale  Models  Validated  3-­D,  CFD

Process  SystemsDesign,  Optimization  &  Control

Physical  PropertiesKinetics

Thermodynamics

16

C2U  BatchUnit

Building  Predictive  Confidence  for  Device-­scale  CO2  Capture  with  Multiphase  CFD  Models

17

Quantitatively  predicting  scale  up  performance

18

Advanced  Computational  Tools  to  Accelerate  Carbon  Capture  Technology  Development

Lab  &  Pilot  ScaleExperiments  &  Data

Device Scale  Models  Validated  3-­D,  CFD

Process  SystemsDesign,  Optimization  &  Control

Physical  PropertiesKinetics

Thermodynamics

19

Carbon Capture Simulation for Industry Impact

• Work  closely  with  industry  partners  to  help  scale  up– Large  scale  pilots

• Help  ensure  success  at  this  scale– Employ  simulation  to  predict  performance,  potential  issue– Help  resolve  issues  using  simulation  tools

• Maximize  learning  at  this  scale– Data  collection  &  experimental  design– Develop  &  Validate  models– UQ  to  identify  critical  data

• Help  develop  demonstration  plant  design– Utilize  optimization  tools  (OUU,  Heat  Integration)– Quantitative  confidence  on  predicted  performance– Predict  dynamic  performance

– Partnership  via  CRADA

20

Acknowledgements– SorbentFit

• David  Mebane  (NETL/ORISE,  West  Virginia  University)• Joel  Kress  (LANL)

– Process  Models• Solid  sorbents:  Debangsu Bhattacharyya,  SrinivasaraoModekurti,  Ben  Omell (West  Virginia  University),  Andrew  Lee,  

Hosoo Kim,  Juan  Morinelly,  Yang  Chen  (NETL)• Solvents:  Joshua  Morgan,  Anderson  SoaresChinen,  Benjamin  Omell,  Debangsu Bhattacharyya (WVU),  Gary  Rochelle  

and  Brent  Sherman (UT,  Austin)• MEA  validation  data:  NCCC  staff  (John  Wheeldon and  his  team)

– FOQUS• ALAMO:  Nick  Sahinidis,  Alison  Cozad,  Zach  Wilson  (CMU),  David  Miller  (NETL)• Superstructure:  Nick  Sahinidis,  Zhihong Yuan  (CMU),  David  Miller  (NETL)• DFO:  John  Eslick (CMU),  David  Miller  (NETL)• Heat  Integration:  Yang  Chen,  Ignacio  Grossmann  (CMU),  David  Miller  (NETL)• UQ:  Charles  Tong,  Brenda  Ng,  JeremeyOu (LLNL)• OUU:  DFO  Team,  UQ  Team,  Alex  Dowling  (CMU)• D-­RM  Builder:  JinliangMa  (NETL)• Turbine:  Josh  Boverhof,  Deb  Agarwal (LBNL)• SimSinter:  Jim  Leek  (LLNL),  John  Eslick (CMU)

– Data  Management• Tom  Epperly (LLNL),  Deb  Agarwal,  You-­Wei  Cheah (LBNL)

– CFD  Models  and  Validation• Xin Sun,  Jay  Xu,  Kevin  Lai,  Wenxiao Pan,  Wesley  Xu,  Greg  Whyatt,  Charlie  Freeman  (PNNL),  Curt  Storlie,  Peter  

Marcey,  Brett  Okhuysen (LANL),  S.  Sundaresan,  Ali  Ozel (Princeton),  Janine  Carney,  Rajesh  Singh,  Jeff  Dietiker,  Tingwen Li  (NETL)  Emily  Ryan,  William  Lane  (Boston  University)

Disclaimer This  presentation  was  prepared  as  an  account  of  work  sponsored  by  an  agency  of  the  United  States  Government.  Neither  the  United  States  Government  nor  any  agency  thereof,  nor  any  of  their  employees,  makes  any  warranty,  express  or  implied,  or  assumes  any  legal  liability  or  responsibility  for  the  accuracy,  completeness,  or  usefulness  of  any  information,  apparatus,  product,  or  process  disclosed,  or  represents  that  its  use  would  not  infringe  privately  owned  rights.  Reference  herein  to  any  specific  commercial  product,  process,  or  service  by  trade  name,  trademark,  manufacturer,  or  otherwise  does  not  necessarily  constitute  or  imply  its  endorsement,  recommendation,  or  favoring  by  the  United  States  Government  or  any  agency  thereof.  The  views  and  opinions  of  authors  expressed  herein  do  not  necessarily  state  or  reflect  those  of  the  United  States  Government  or  any  agency  thereof.

21

Framework  for  Optimization,  Quantification  of  Uncertainty  and  Sensitivity

D.  C.  Miller,  B.  Ng,  J.  C.  Eslick,  C.  Tong  and  Y.  Chen,  2014,  Advanced  Computational  Tools  for  Optimization  and  Uncertainty  Quantification  of  Carbon  Capture  Processes.  In  Proceedings  of  the  8th  Foundations  of  Computer  Aided  Process  Design  Conference  – FOCAPD  2014.  M.  R.  Eden,  J.  D.  Siirola and  G.  P.  Towler Elsevier.

SimSinterStandardized  interface  for  simulation  software

Steady  state  &  dynamic

SimulationAspengPROMSExcel

SimSinter ConfigGUI

Results

FOQUSFramework  for  Optimization  Quantification  of  Uncertainty  and  Sensitivity

Meta-­flowsheet: Links  simulations,  parallel  execution,  heat  integration

Samples

SimulationBased  

OptimizationUQ

ALAMO  Surrogate  Models

Data  ManagementFramework

TurbineParallel  simulation  execution  

management  systemDesktop  – Cloud  – Cluster  

iREVEALSurrogate  Models

Optimization  Under  

Uncertainty

D-­RMBuilder

Heat  Integration

22

• Organizational  Meetings:  March  2010  -­ October  2010• Technical  work  initiated:  Feb.  1,  2011• Preliminary  Release  of  CCSI  Toolset:  September  2012

– Initial  licenses  signed• CCSI  Year  3  starts  Feb.  1,  2013

– Began  solvent  modeling/demonstration  component• 2013  Toolset  Release:  October  31,  2013

– Multiple  tools  and  models  released  and  being  used  by  industry

• 2014  Toolset  Release:  October  31,  2014  

• Future– Final  IAB  meeting:  Sept.  23-­24,  2015  (Reston,  VA)  – Final  major  release  November  2015– Commercial  licensing  late  2015  or  early  2016

CCSI  Timeline

22