dynamic modeling and control of an integrated solid

58th Annual ISA Power Industry Division Symposium7-11 June 2015, Kansas City, Missouri

Kansas City Marriott Hotel

Dynamic Modeling and Control of an Integrated Solid Sorbent Based CO2 Capture Process Benjamin Omell, Debangsu Bhattacharyya*Department of Chemical EngineeringWest Virginia University, Morgantown, WV

David C. Miller, National Energy Technology LaboratoryUS DOE, Pittsburgh, PA

Jinliang Ma, Priyadarshi Mahapatra, Stephen E. ZitneyNational Energy Technology LaboratoryUS DOE, Morgantown, WV

11

58th Annual ISA POWID Symposium, 7-11 June 2015, Kansas City, Missouri

Short Biography of Benjamin Omell

• Post-Doctoral Fellow in the Department of Chemical Engineering at West Virginia University

• PhD from Illinois Institute of Technology, Chicago• Research interest is in the area of steady-state and

dynamic modeling and advanced process control of energy-generating and associated processes

• Member of AICHE• Hobbies- Biking, hiking

2


CO2 Capture & Compression SystemsCoupled with the Steam Cycle

• Solid-sorbent systems are energetically superior to typical MEA-based solvent systems

• CCS system requires power and steam

• Dynamics are important during load-following

Bubbling Fluidized

Bed (BFB) Adsorber

CO2 Compression System

Moving Bed (MB)

Regenerator


Goals and Objectives

• Develop accurate and flexible steady-state and dynamic solid sorbent models for CO2 capture– BFB and MB adsorber/regenerator– Solids heat exchangers– CO2 compression – Balance of the plant

• Create toolset to simplify implementation of advance control strategies that can perform efficiently in the presence of unmodeleddisturbances, noisy measurements, and unmeasured variables

• Use these models and tools for optimization, transient studies, and control system design as part of DOE’s Carbon Capture Simulation Initiative (CCSI)

4


CCSI Toolset Framework

5


Challenge: Limitations in Previous Bubbling Fluidized Bed Models

• Typical simplifications to facilitate analytic solutions– Isothermal– Simplistic reaction kinetics and

transport– Steady-state– Embedded heater/cooler neglected

• Limited support in commercial tools

• Minimal application to CO2 capture in literature

6


A Flexible BFB Model

• Flexible configurations– Dynamic or steady-state– Adsorber or regenerator– Under/overflow– Integrated heat exchanger for

heating or cooling

• Supports complex reaction kinetics

• Compatible with CCSI uncertainty quantification (UQ) tools

1-D, two-phase, pressure-driven and non-isothermal models developed in both Aspen Custom Modeler (ACM) and gPROMS

7


Model Development

• Gaseous species : CO2, N2, H2O• Solid phase components: bicarbonate, carbamate, and

physisorbed water.• Transient species conservation and energy balance

equations for both gas and solid phases in all three regions.

8


Single-Stage BFB Model: Steady-State Results

0 0.2 0.4 0.6 0.8 10.05

0.06

0.07

0.08

0.09

0.1

0.11

0.12

0.13

Z/L

Mole

Frac

tion

H2O

CO2

Flue GasInlet

Flue GasOutlet

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

1.2

1.4

Z/L

Mol

ar L

oadi

ng (m

ol/k

g so

lids)

Solids overflow exit type configuration

BicCarH2O

SolidsInlet

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Z/L

Mole

Frac

tion

H2O

CO2

Gas Inlet Gas Outlet 0 0.2 0.4 0.6 0.8 10.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Z/L

Mol

ar L

oadi

ng (m

ol/k

g so

lids)

BicCarH2O

Solids Outlet Solids Inlet

Adsorber

Regenerator

9


Dynamic Results – Increase Inlet GasFlow by 20.6%

Gas CO2 Concentration Gas H2O Concentration


Challenge: Limitations of ExistingMoving Bed Models

Very few references in literature, little application to CO2 capture

Previous applications in literature include MB furnaces for iron pellet reduction Dryers Non-catalytic gas-solid reactions

Lack of mathematical model with large amount of heat transfer for solid sorbent regeneration

Embedded heater/cooler not modeled Mainly steady-state model Hardly any model available in the

commercial software

Solid In

Solid Out

Gas In

Gas Out

Utility In

Utility Out

11


CCSI’s Moving Bed Models

Flexible dynamic and steady-state models Integrated heat exchanger that can be

used interchangeably for heating or cooling

Flexible enough for adsorber application

A1-D, two-phase, pressure-driven and non-isothermal regenerator model developed in both ACM and gPROMS

Solid In

Solid Out

Gas In

Gas Out

Utility In

Utility Out

12


Step Test: Sorbent Temperature

13


Multi-Stage Moving Bed Design to Overcome Fluidization Concerns

ConcernTwenty-seven 9 m diameter MB regenerators in parallel required to maintain flow in moving bed regime*

Solution: Multi-Stage Moving Bed Regenerator• Some CO2 (gas) removed between

stages• Reduced gas flowrate at the top of the

MB• Requires an advanced control strategy

Solid in

Stage 1

Stage 2

HX steam

downcomer

CO2 draw-off

Steam Regenerated solids

CO2 to compressor

*Assumptions for preliminary analysis: 12% CO2 flue gas @ 2000 mol/s with 90% capture rate

14


Challenge: Limitations of Existing Compression Systems Models

• Has been mostly developed for non-CO2 systems• For CO2 systems, developed mainly for sCO2 cycles

– Pressure ratio of 1 to 2.6 (about 150 for CCS)– Fixed inventory (variable for CCS)– Composition change, especially water content, is not a

major concern for sCO2 cycles• Typically steady-state

– Dynamics are essential for load-following in power systems

• Lack of performance curves for CO2 systems• Surge detection and control algorithms hardly studied for

these systems

15


CCSI CO2 Compression System ModelDynamic model of multi-stage integral-gear compression system

Surge Control Valves

TEG Regenerator & AbsorberFlash Vessels

Surge detection and control algorithm also developed

16


Integrated Model Enables Investigationof Entire Process Dynamics

Integrated Model in gPROMS

(also ACM version)


Ramp In Flue Gas

Flow into compressor train

Time Hours

Inle

t flo

w k

mol

/hr

0.0 20.0 40.0

5600

.058

00.0 CO2 % Removed

Time Hours

%

0.0 20.0 40.0-0.2

-0.1

0.0

0.1

0.2

0.3

0.4Ramp in flue gas

Time=0 : 15.8% increaseTime=20: 28.8% decrease

Total Power

Time Hours

kW

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0

2300

0.0

2400

0.0

18


Dynamics of Integrated Process

• Transient effect of the capture and compression process on the power plant and vice versa

• Can result in temporal variation in CO2 capture target-different time constants depending on the type of the bed

• CCSI tools DRM-builder and APC toolset can improve system dynamics



20


Motivation

• High-fidelity dynamic models (e.g. ACM dynamic models) are computationally expensive– Need to solve many Differential Algebraic Equations

(DAEs)– May require small time steps due to stiffness of DAEs– Not fast enough to catch up with real time

• Dynamic reduced models (D-RMs)– Speed up the dynamic simulations– Capture dynamic systems with reasonable accuracy– Can be used for Advanced Process Control (APC) and

Real Time Optimization (RTO)

21


D-RM Builder Workflow

Generate D-RM Based on

Simulation Results

1u

2u1

2

3

4

I/O Variable Selection

GUI for Configuring Inputs/Outputs GUI for Configuring

Training SequenceRun Training

Sequence

Analyze Reduced Model using UQ Tools, Validation Sets, Plots

D-RM in Form of MATLAB Code

SinterConfigGUI


D-RM for the BFB Adsorber

D-RM generated based on open-loop ACM model Inputs:

• Flue gas flow rate: 6,075 to 7,425 kmol/hr• Sorbent flow rate: 540,000 to 660,000 kg/hr

Output:• CO2 removal (Fraction of CO2 in flue gas removed)

DABNet* model with pole values optimized CPU time required for ACM simulations

• Approximately 50 minutes for 2500 sampling steps (Sampling time interval at 0.1 second)

*Sentoni, G.B., L.T. Biegler, J.B. Guiver and H. Zhao, “State-Space Nonlinear Process Modeling: Identification and Universality,” AIChE Journal, 44(10), 2229-2239, 1998.

23


Validation Input Data

Time (sec)

Sorb

ent F

low

(k

g/hr

)Fl

ue G

as F

low

(k

mol

/hr)

ACM

D-RM

24


Validation Output Data

Time (sec)

Rel

ativ

e Er

ror

Frac

tion

of

CO

2R

emov

ed

25



26


APC Framework’s Capabilities

Utilize proven techniques from control relevant studies in literature and combine them synergistically in an efficient and robust process control framework.

DABNet-based Nonlinear System Identification

Analytical Jacobians and Hessians

Multiple Model Predictive Control (MMPC)

Auto Covariance Estimation (ALS Technique)

IPOPT-based Optimization Technique

UKF-based State Estimation

Enhanced User-Friendly GUI

Poor control response for highly nonlinear plants using traditional PID or MPC

Nonlinear MPC becomes computationally expensive for large-scale plant

Noisy measurements along with unmodeled disturbance and unmeasured variables provides poor control response

Controller is too sensitive to user-provided tuning parameters (inexperienced operators)

Typical APC interface overwhelms the operators (setting up control-model, tuning parameters, etc.)

Typi

cal p

robl

ems

with

APC

app

licat

ions

APC

Framew

ork

27


Time [seconds]0 500 1000 1500 2000 2500 3000

CO

2 C

aptu

re [%

]

85

86

87

88

89

90

SetpointConventional MPCDAB-Net NMPC

Time [seconds]0 500 1000 1500 2000 2500 3000

Fso

rben

t [k

g/hr

]

10 5

3

4

5

6

7

8

9

Time [seconds]0 500 1000 1500 2000 2500 3000

Fflu

egas

[km

ol/h

r]

6000

6200

6400

6600

6800

7000

Performance Comparison on 2-Stage BFB Adsorber (ACM)

Controller responses to drastic plant-load changes –Comparison with standard MPC controller

Large changes in flue-gas flowrate

(input-disturbance)

Active Constraints

Better disturbance rejection

Note: Max. Control Calculation Time << Sample Time (Ts = 20 sec), Real-Time Operation with APC

Controller Parameters

Prediction Horizon = 50 Control Horizon = 10

Wu/Wy = 0.2

Cumulative Control

Calculation Time (sec)

Max Control Calculation Time (sec)

Total Simulation Time (min)

Conventional MPC 0.019 9.54 0.33 18.39

DAB-Net NMPC 0.007 19.97 3.43 18.92

Computational CostCumulative

ResidualAlgorithm

28


0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.01 0.1 1 10

Resi

dual

= ∑

(r-y

)TW

y(r-y

)

Wu/Wy

MATLAB's MPC Toolbox

APC Framework's DABNet-based NMPC

Performance Comparison on 2-Stage BFB Adsorber (ACM)

Benefits of APC Framework1. Low residuals – tracks

setpoint better over time

2. Low sensitivity for user-provided tuning parameters (Wu/Wy in this case)

Sensitivity of Control Performance to Tuning Parameter –Comparison with standard MPC controller

29


Developed high-fidelity steady-state and dynamic solid sorbent models for CO2 capture

Utilization of DRM-builder tool to generate reduced ordered models Use reduced order model to generate control strategy with

APC framework to handle moving boundary problem of multi-stage moving bed regenerator

APC performance is relatively insensitive to tuning parameters and results in efficient disturbance rejection and load-following characteristics

Summary


Acknowledgement

DOE’s Carbon Capture Simulation Initiative for funding. Part of this technical effort was performed under the RES contract DE-FE0004000 through National Energy Technology Laboratory’s Regional University Alliance (NETL-RUA.

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.

Disclaimer

31


Thank You

32


Optimized Process Developed using CCSI Toolset

GHX-001

CPR-001

ADS-001

RGN-001

BLR-001

Flue Gas from Plant

Clean Gas to Stack

CO2 to Sequestration SHX-01

SHX-02

CPR-002

LP/IP Steam from Power Plant

CPP-002

ELE-002

ELE-001

Flue GasClean Gas

Rich Sorbent

LP SteamHX Fluid

Legend

Rich CO2 Gas

Lean Sorbent

Parallel ADS Units

Parallel RGN Units

Parallel ADS Units

GHX-002

CPP-000

Cooling Water

Injected Steam

Cooling Water

Saturated Steam Return to Power Plant

Mild SteamDirectly Injected into Reactor

Compression Train

Cond

ense

d W

ater

Cond

ense

d W

ater

∆Loading1.8 mol CO2/kg

0.66 mol H2O/kg

Solid Sorbent MEA27

(∆10°C HX)MEA27

(∆5°C HX)Q_Rxn (GJ/tonne CO2) 1.82 1.48 1.48Q_Vap (GJ/tonne CO2) 0 0.61 0.74Q_Sen (GJ/tonne CO2) 0.97 1.35 0.68

Total Q 2.79 3.44 2.90

dynamic modeling and control of an integrated solid

Documents