Computational tools and models for carbon capture processes

David S. Mebane1 and David C. Miller2

1West Virginia University, Morgantown, WV USA2National Energy Technology Laboratory, Pittsburgh, PA USA

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

Development

10-15 years

Process Scale Up

20-30 years

Accelerated deployment timeline

Process Scale Up

15 years

1 M

We

10

MW

e

50

0 M

We

10

0 M

We

3

For Accelerating Technology Development

National LabsAcademia 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

6

CCSI Toolset to accelerate development and scale-upProcess Models

Solid In

Solid Out

Gas In

Gas Out

Utility In

Utility Out

Basic Data

Submodels

Carbon Capture 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

Process Synthesis, Design & Optimization

CFD Device ModelsCCSI CFD Validation Hierarchy for Sorbent Capture

NETLCarbonCaptureUnit(C2U)ReactingUnit

Upscaling

(heat transfer

filtering)

Coupled

adsorption

reaction, heat

transfer and flow

Non-reactive

flow and heat

transfer of

cooling tubes

Hydrodynamics

of Bubbling

Fluidized Bed

25MWe,100MWe,650MWe

SolidSorbentSystems

Upscaling

(flow filtering)

Upscaling

(energy

filtering)

Intermediate

Validation

(Adsorber without

reactions and heat

transfer)

1MWeCarbonCaptureSystem

Demonstration and

Full Scale Systems

Unit

Problems

Laboratory Scale

Subsystem

(Decoupled

benchmark cases)

Pilot Scale

Systems

Process Dynamics

and Control

7

1st Gen Basic Data Model

Alkylammonium carbamate formed by two molecules

of tetraethylene pentamine (TEPA) and CO2.

8

Least squares Fitting

Uncertainty Quantification

UQ with model calibration

Experimental DataModel ParametersProcess/Device-Scale Model

9

➢ Estimates model parameters by fitting to TGA data.

➢ Uses a particle swarm optimizer / Markov chain Monte Carlo routine on multiple processors.

➢ Can handle datasets of any size (Data Management Framework).

➢ Fits any combination of 13 parameters: all at once, or can split into dry CO2, H2O-only and humid

CO2 runs.

SorbentFit Tool: TGA

10

SorbentFit Tool: Fixed Bed

Simulated breakthrough curves

➢ Estimates model parameters by fitting to fixed bed data.

➢ Uses a particle swarm optimizer / Markov chain Monte Carlo routine on multiple processors.

➢ Can handle datasets of any size (Data Management Framework).

➢ Fits any combination of 13 parameters: all at once, or can split into dry CO2, H2O-only and humid

CO2 runs.Bivariate marginal distributions for

selected parameters

11

➢ Captures the complex

observed behavior of

the sorbent

➢ Three diffusing species;

two final adsorbed

states

➢ Reaction-diffusion

model order reduced by

over 10x using

stochastic functions:

Dynamic Discrepancy

Reduced Modeling

SorbentFit Tool: High Fidelity Sorbent Model

12

SorbentFit Tool: High Fidelity Sorbent Model

15

High Fidelity Process Models for Carbon CaptureBubbling Fluidized Bed (BFB) Model

➢ Variable solids inlet and outlet location

➢ Modular for multiple bed configurations

Moving Bed (MB) Model

➢ Unified steady-state and dynamic

➢ Heat recovery system

Compression System Model

➢ Integral-gear and inline compressors

➢ Determines stage required stages, intercoolers

➢ Based on impeller speed limitations

➢ Estimates stage efficiency

➢ Off-design and surge control

Solvent System Model

➢ Predictive, rate-based models

Membrane System Model

➢ Hollow fiber

➢ Supports multiple configurations

0 0.2 0.4 0.6 0.8 10.04

0.06

0.08

0.1

0.12

0.14

z/L

yb (

km

ol/km

ol)

CO2

H2O

0 0.5 1200

400

600

800

1000

km

ol/

hr

z/L

Solid Component Flow Profile of MB Regenerator

Bicarb. Carb. H2O

16

Developing Detailed, Predictive Models of Solvent-Based Capture Processes

Steady-State and DynamicProcess Model

Process Sub-Models

Kinetic model

Hydrodynamic Models

Mass Transfer Models

Properties Package

Chemistry Model

Thermodynamic Models

Transport Models

UQ

Pilot/Commercial Scale Data

WWC/Bench/Pilot Scale Data

Lab Scale Data

UQ

Process UQ

Measurement Uncertainty

Measurement Uncertainty

Measurement Uncertainty

17

➢ Step 1: Define a large set of potential basis functions

➢ Step 2: Model reduction

ALAMO Surrogate Model Development

True error

Empirical error

Complexity

Err

or

Ideal Model

OverfittingUnderfitting

18

➢ 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 Superstructure Using Surrogate Models

Surrogate models for

each reactor and

technology used

19

Superstructure Optimization to Determine System Configuration

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

20

Simultaneous Simulation Based Optimization & Heat Integration

21

Optimization & Heat Integration

w/o heat

integrationSequential Simultaneous

Net power efficiency (%) 31.0 32.7 35.7

Net power output (MWe) 479.7 505.4 552.4

Electricity 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 efficiency

Constraint: CO2 removal ratio ≥ 90%

Decision Variables (17): Bed length, diameter, sorbent

and steam feed rate

22

CO2 Capture & Compression Systems Coupled with Steam Cycle

Integrated Model Enables Investigation of Entire Process

Compressor stages 1-6

Two-stage

BFB

adsorber 14-stage MB regenerator

TEG moisture

removal

Compressor stages 7-8

➢ CCS system requires power and steam

➢ Dynamics are important during operation

23

Ramp In Flue Gas Flowrate affects Power Plant, Power Demand

Resulting Power Consumption of CO2 CompressorsInitial Flue Gas Ramp

% CO2 Removed

24

Dynamic Reduced Model: Comparison with Detailed Model

% o

f C

O2

Cap

ture

d

So

rben

t F

eed

Flo

w

(kg/h

r)

Flu

e G

as

Flo

w

(km

ol/

hr)

Time (sec)

Single-Pole Result (Old) Double-Pole Result (New)

ACM

D-RM

Fast Response Slow Response

ACM Flowsheet Model

Inaccurate

Input Step Changes

Input Step Changes

Output Profile

25

APC Framework: Performance Comparison on 2-Stage BFB Adsorber

(ACM)

26

Framework for Optimization, Quantification of Uncertainty and Surrogates

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

SimulationAspen

gPROMS

Excel

SimSinter Config GUI

Resu

lts

FOQUS

Framework for Optimization Quantification of Uncertainty and Surrogates

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

Sa

mp

les

Simulation

Based

Optimization

UQ

ALAMO

Surrogate

Models

TurbineParallel simulation execution management

system

Desktop – Cloud – Cluster

iREVEAL

Surrogate

Models

Optimization

Under Uncertainty

D-RM

BuilderHeat Integration Data Management

Framework

27

Dynamic Reduced Models: D-RM Builder

Select I/O

Variables

Set Input

Ranges

Create Step

Change

Sequence

Perform

Dynamic

Simulation

Create/

Train D-RM

Plot/

Validate

D-RM

Perform UQ

Analysis

Export

D-RM

SinterConfigGUI SimSinter

FOQUS/D-RM Builder GUI

D-RM CodeACM Model

Toolbar

Menu

28

➢ Generate samples from different probability distributions▪ Produce sample input graphs (1D and 2D)

▪ Modify sample points

➢ Launch ensemble simulations (locally or via Turbine)▪ Track run progress

▪ Perform sample refinement

➢ Parameter screening (identify “unimportant” parameters)▪ Use parametric (e.g., linear) or nonparametric methods

➢ Response surface analysis▪ Provide a large selection of curve-fitting functions

▪ Perform response surface validation and visualization

➢ Uncertainty/sensitivity analysis ▪ Perform on raw sample or on response surfaces, visualize results

▪ Compute probability distributions: aleatory or mixed (aleatory+epistemic)

▪ Compute first, second or total order sensitivity analysis

➢ Parameter inference (calibration)▪ Infer with or without discrepancy

▪ Generate posterior plots and posterior sample

▪ Facilitate hierarchical (multi-stage) UQ

FOQUS Uncertainty Quantification Capabilities

30

Process Models

Basic Data Fitting Tools

log10(k0_1)

De

nsity

3 4 5 6 7

0.0

1.0

2.0

log10(k0_2)

De

nsity

7 8 9 10 11

02

46

log10(k0_3)

De

nsity

7 8 9 10 11

0.0

0.4

0.8

EA_1

De

nsity

0 50 100 150 200

0.0

00.0

20.0

4

EA_2

De

nsity

0 50 100 150 200

0.0

00

.04

0.0

8

EA_3

De

nsity

0 50 100 150 200

0.0

00

0.0

10

log10(D0)

De

nsity

-11.0 -10.5 -10.0 -9.5 -9.0

01

23

45

a

De

nsity

-1.0 -0.8 -0.6 -0.4 -0.2

02

46

8

b

De

nsity

-3.0 -2.5 -2.0 -1.5 -1.0 -0.5

0.0

1.0

2.0

APC Framework

d1d3

y3

Process 1

Process 2

Process 3

D-RM 1

APC 3

D-RM 3

3uk-1

u3

3uk

3dk-1

y2

3yk

APC 2

D-RM 2

2yk

2uk

2uk-1

u2

2dk-1

d2

2rk

3rk

r2

r3

1uk

1uk-1

u1

1dk-1

1yk

y1

1rk

r1

APC 1

APC Framework

Process Simulator /

Real Plant

CFD Models

CCSI CFD Validation Hierarchy for Sorbent Capture

NETLCarbonCaptureUnit(C2U)ReactingUnit

Upscaling

(heat transfer

filtering)

Coupled

adsorption

reaction, heat

transfer and flow

Non-reactive

flow and heat

transfer of

cooling tubes

Hydrodynamics

of Bubbling

Fluidized Bed

25MWe,100MWe,650MWe

SolidSorbentSystems

Upscaling

(flow filtering)

Upscaling

(energy

filtering)

Intermediate

Validation

(Adsorber without

reactions and heat

transfer)

1MWeCarbonCaptureSystem

Demonstration and

Full Scale Systems

Unit

Problems

Laboratory Scale

Subsystem

(Decoupled

benchmark cases)

Pilot Scale

Systems

Gas In

Liquid In

Gas Out

Liquid Out

PackedBed

Film Flow: O (mm)

WWC:

O(mm)

Device Scale:

O(m)

Domain (60´50 mm2)

Superstructure Optimization

Oxycombustion System

Optimization Framework

Solvent Blend Model

1E-2

1E+0

1E+2

1E+4

1E+6

0.1 0.3 0.5 0.7 0.9

P*CO2

(Pa)

loading (mol CO2/mol alk)

40°C7m2° Amine

LeLi,2015

pKa =8

=9

=10

pKa =11

CCSI TOOLSET PRODUCTS

31

Acknowledgements

Funding provided by U.S. Department of Energy, Office of Fossil Energy

▪ SorbentFit• David Mebane, Brian Logsdon, Kuijun Li (West Virginia University)• Joel Kress (LANL)

▪ Process Models• Solid sorbents: Debangsu Bhattacharyya, Srinivasarao Modekurti, Ben Omell (West Virginia University), Andrew Lee, Hosoo Kim, Juan Morinelly, Yang

Chen (NETL)• Solvents: Joshua Morgan, Anderson Soares Chinen, 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, Qianwen Gao (NETL)• Heat Integration: Yang Chen, Ignacio Grossmann (CMU), David Miller (NETL)• UQ: Charles Tong, Brenda Ng, Jeremey Ou (LLNL)• OUU: DFO Team, UQ Team, Alex Dowling (CMU)• D-RM Builder: Jinliang Ma (NETL)• Turbine: Josh Boverhof, Abdelrahman Elbashandy, Deb Agarwal (LBNL)• SimSinter: Jim Leek (LLNL), John Eslick (CMU)

▪ APC Framework• Priyam Mahapatra, Steve Zitney (NETL)

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

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.

31