You are on page 1of 9

Validation of Numerical Models of Flow Through SCR

Units
A.N. Sayre
M.G. Milobowski
Babcock & Wilcox
Barberton, Ohio, U.S.A.

Presented to: BR-1678


EPRI-DOE-EPA Combined Utility Air Pollutant Control
Symposium
August 16-20, 1999
Atlanta, Georgia, U.S.A.

Abstract allow the flow and mixing properties in geometrically complex


Numerical modeling is used by McDermott Technology, Inc. equipment to be numerically modeled economically.
(MTI) and The Babcock & Wilcox Company (B&W) to support In this study, the flue gas flow through the SCR units in Tai-
the design of products, including selective catalytic reduction wan Power Company’s Taichung Units 5-8 was numerically
(SCR) units. Accurate determination of the pressure drop through modeled. The accuracy of the model predictions was compared
the SCR system is required to stay within plant fan capacities to data obtained from a 1/10 scale physical flow model study.
for retrofit applications. Furthermore, uniform flows and mix- FLUENT-UNS was used to predict the flow using an unstruc-
ing of NH 3 and NOx at the entrance to the catalyst bed are re- tured tetrahedral mesh. The small features of the ammonia in-
quired to meet the design specifications of the catalyst vendor. jection grid were accurately resolved using this advanced geo-
Traditionally, physical flow models have been used to ensure metric model. Pressure drops obtained from numerical model-
the proper flow properties entering the catalyst bed. Advances ing compared well with those experimentally determined from
in computer processor speed and grid generation techniques now the physical flow model study.

Babcock & Wilcox 1


Introduction
Numerical modeling has been used as an effective, economi-
cal design and analysis tool by MTI and B&W for simulating
such items as flow, heat transfer and combustion phenomena in
boiler components and auxiliary equipment. Furnaces, burners,
windboxes, steam drums, pulverizers, precipitators, coal nozzles
and piping are only a few examples of the many components
and equipment that have been successfully numerically mod-
eled by MTI/B&W.
With the increasing demand for SCR systems, coupled with
compressed project design/fabrication/erection timetables, the
utilization of numerical modeling appeared attractive for the
following reasons:
1. While numerical modeling of a SCR system can be used
to confirm the same criteria that is typically checked with a
physical flow model (i.e. flue gas velocity distribution, pres- Figure 1 Outline of left-hand side of SCR unit.
sure drop), numerical modeling can also be used to economi-
cally predict flue gas temperature distribution, flue gas NO x monia injection grid (AIG) used in the 1/10th scale physical
distribution, and ammonia/NO x distribution. flow model was also obtained from NELS (O’Sullivan, 1998).
2. Due to its flexibility, numerical modeling permits the The geometry of the full-scale unit, including the AIG, was ob-
study of an increased number of geometric arrangements or tained from the contract’s design erection/arrangement drawings.
modifications in a more timely manner than is possible with Since the SCR unit has two flow paths which are mirror im-
physical flow modeling. ages of each other, only half the SCR unit was modeled and a
3. Numerical modeling can accurately represent complex symmetry boundary condition was employed at the center of
devices, such as static mixers, and their effect on flue gas ho- the plenum chamber.
mogeneity and ammonia/NOx mixing. A “water-tight” surface representation of the ductwork was
MTI/B&W’s experience with numerical modeling required created in MSC-PATRAN, Version 7.5. PATRAN is a thermal
that the results of the first SCR system numerically modeled be analysis program which is flexible in the creation of model ge-
validated against the results of a physical flow model. To this ometry. A triangular surface mesh was then applied. PATRAN
end, the latest B&W SCR project, Taiwan Power Company’s gave unsuitable meshes (highly skewed) for the AIG, so a pro-
Taichung Units 5-8, was selected for the validation work. The gram was written in FORTRAN which created a surface mesh
SCR systems for Units 5 and 6 were placed in service in 1996, for the AIG in a PATRAN-neutral file format. The surface mesh
while those for Units 7 and 8 began operation in 1997. for the AIG was then merged into the rest of the model, and
Each unit of Taichung 5-8 is equipped with a natural circu- exported in PATRAN-neutral file format. A view of the numeri-
lation, balanced draft, sub-critical pressure B&W Carolina-type cal model’s surface mesh of the AIG is shown in Figure 2.
boiler having single stage reheat. Each boiler is rated at 550
MW when fired on coal, oil, or any combination of coal and oil.
One SCR system is provided for each unit. Each SCR system
utilizes two fixed-bed, parallel passage type vertical downflow
reactors, each handling 50% of a unit’s total flue gas flow. Each
SCR reactor is provided with an ammonia injection grid
equipped with 284 injection ports. The SCR systems are de-
signed for 50% minimum NOx removal, and 5 ppmvd maxi-
mum ammonia slip.
The accuracy of the Taichung Units 5-8 SCR numerical mod-
eling was compared to data obtained on a 1/10 scale physical
flow model study conducted by NELS Inc., a flow modeler/con-
sultant. In the NELS study, scaling relationships were used to
scale the physical flow model results to full scale. The MTI/
B&W validation of the Taichung 5-8 SCR units was performed
by numerical modeling both the 1/10th scale model and full scale
actual design. An outline of the Taichung SCR system, left-hand
side when viewed from the economizer, is shown in Figure 1.
Figure 2 Detail of the Ammonia Injection Grid (AIG).

Model
Both the full scale actual Taichung SCR design and the NELS The surface mesh was imported into Fluent Inc.’s TGRID
1/10th scale design were numerically modeled to provide vali- (V. 3.0), where the tetrahedral volume mesh was generated. The
dation on the use of numerical models for providing flow guar- tetrahedral meshes for both geometries were composed of ap-
antees. Geometrical information used in the 1/10th scale model proximately 682,000 elements. The small features of the AIG
generation was obtained from Plates 1-4 from the NELS report were thus accurately resolved using this advanced geometric
which are reproduced in Appendix A. The geometry of the am- model.

2 Babcock & Wilcox


The flow solution was obtained using Fluent Inc.’s FLUENT- cates the location of the SCR catalyst bed. A continuous gradi-
UNS (V. 4.2), a general-purpose, unstructured flow solver. The ent in static pressure can be seen across the SCR bed. The com-
flow was modeled as isothermal, and two scalars were used for bined effect of the porous plate and AIG can be observed in the
flue gas and ammonia. A standard formulation of the k-e model step-wise change in static pressure.
was used for turbulence, with a standard “log-of-the-wall” used
for boundary conditions at the walls. Pressure-velocity coupling
was obtained through the standard SIMPLEC algorithm, and
second order numerics were used to converge the flow solu-
tion.
A 60% open porous plate, which lies in the flow path up-
stream of the AIG to aid in flow distribution, was modeled us-
ing the “porous jump” capability. This allows a pressure drop
to be modeled over a membrane surface and provides a source
term in the momentum equation as:
Static Pressure (in. H 2O)

(1) 0.0
-0.5
-1.0
-1.5
-2.0

Figure 3 Contours of static pressure on a plane through the


SCR system.

Note that the first term is a viscous resistance term and is


neglected, and the second term is an inertial resistance term.
The coefficient C2 was calculated from the data in the NELS Mass-averaged surface integrals of static pressure were cal-
report in the direction normal to the bulk flow. All momentum culated at measurement planes corresponding to those reported
sources were modeled as isotropic with these values. in the NELS report. Measurement locations are shown in Plate
The SCR bed was modeled as a “porous zone”. The SCR 4, Appendix A. The mass-averaged static pressure was calcu-
bed in the 1/10th scale model was modeled as isotropic. A power- lated as:
law relationship provided a better fit to the data for the full-
scale unit, and the following relationship was used:

(2)

The power-law model is isotropic in nature, requiring only


one set of coefficients.
The flow into the numerical model from the economizer out-
let was modeled as uniform with an ambient temperature air Comparison of the 1/10th scale numerical model predictions
density and velocity of 1.18 kg/m3 and 3.7 m/s for the one-tenth to the physical flow model data is presented in Figure 4. Com-
scale model, and a hot flue gas density and velocity of 0.55 kg/ parison of full-scale numeric model predictions to NELS full-
m3 and 5.5 m/s for the full-scale model. Turbulence intensities scale predictions derived from physical flow model data using
were assumed to be 10%, and the length scale set to the hydrau- scaling techniques is shown in Figure 5. The comparisons are
lic diameter of the economizer outlet for both geometries. The in excellent agreement. It should be noted the exact locations
velocity of the ammonia/air mixture was set equal to the sur- of measurement positions were not given in the NELS report
rounding flue gas flow at the AIG, and the turbulence intensity and were estimated from Plates 1-4.
set to 10% with the hydraulic diameter based on the nozzle out- • The pressure drop labeled “T1-T2” is the pressure drop
let diameter. between the economizer outlet (model inlet) and a planar sur-
face halfway between the porous plate and the AIG.
• The pressure drop labeled “T2-T3” is the pressure drop
Results between T2 and a planar surface halfway between the AIG and
Contours of static pressure distribution are shown in Figure the leading edge of the first set of turning vanes.
3 for the 1/10th scale numerical model. The location of the plane • The pressure drop labeled “T3-T4” is the pressure drop
was taken at an arbitrary location through the left-hand side of between T3 and a planar surface located in the horizontal sec-
the flow path. The rectangular surface color-coded in red indi- tion of the hood immediately before the vertical flue.

Babcock & Wilcox 3


“The catalyst outlet velocity profiles were mea-
sured initially at test location T7. However, this
location contains a large number of structural mem-
bers, which interfere with the velocity measuring
locations. In order to measure a more realistic cata-
lyst outlet distribution, the measurements were sub-
sequently taken at location T6A, which is within
the catalyst section, but downstream of the synthetic
filter material used to simulate the catalyst beds.”
The higher pressure drop obtained in the physical flow model
indicates that the structural members in the bottom of the cata-
lyst bed are contributing to the pressure drop and are not ac-
counted for in the CFD model.

Conclusions
The excellent agreement between the predicted pressure drop
Figure 4 Computational Fluid Dynamics (CFD) predictions in the numerical model and the measured pressure drop in the
and physical flow model results for the 1/10 scale geometry. physical flow model provides validation of numerical model-
ing for predicting flow through SCR units.

Recommendations
With the proven accuracy of numerical modeling for flows
through SCR units, customers should be urged to use in-house
numerical modeling instead of sub-contracted physical flow
modeling for numerous reasons. Four of these reasons are the
economic advantage, the flexibility to model numerous configu-
rations including complex geometries, the ability to confirm
more design criteria, and the safeguarding of proprietary tech-
nology.

References
1. NELS INC., “Air Flow Model Study Selective Catalytic
Reduction System Taiwan Power Company Taichung 5 – 8,
NELS Project No. P046.92, January 1993.
2. O’Sullivan, K., personal communication, November 1998.
Figure 5 CFD predictions and NELS predictions based on
scaling criteria for full-scale geometry.

Copyright © 1999 by The Babcock & Wilcox Company,


• The pressure drop labeled “T4-T6” is the pressure drop All rights reserved.
between T4 and the top of the SCR catalyst bed.
• The pressure drop labeled “T6-T6A” is the pressure drop No part of this work may be published, translated or reproduced in any
between T6 and a planar surface located within the SCR cata- form or by any means, or incorporated into any information retrieval
system, without the written permission of the copyright holder.
lyst bed, near the bottom.
Permission requests should be addressed to: Market Communications,
• The pressure drop labeled “T6A-T8” is the pressure drop The Babcock & Wilcox Company, P.O. Box 351, Barberton, Ohio,
between T6A and the model exit. U.S.A. 44203-0351.
The counter-balancing errors in the pressure drops between
locations T2-T3 and T3-T4 may be explained by an incorrect Disclaimer
indication in the NELS report on the location of the AIG. In Although the information presented in this work is believed to be
B&W Drawing No. 444500E, the location of the AIG is indi- reliable, this work is published with the understanding that The
cated as 914 mm downstream of the expansion joint; whereas Babcock & Wilcox Company and the authors are supplying general
in NELS Plate 2, Appendix A, the perforated plate is located at information and are not attempting to render or provide engineering or
professional services. Neither The Babcock & Wilcox Company nor any
914 mm downstream of the expansion joint and the AIG is indi-
of its employees make any warranty, guarantee, or representation,
cated as being located approximately half the distance between whether expressed or implied, with respect to the accuracy,
the perforated plate and the leading edge of the five turning completeness or usefulness of any information, product, process or
vanes (approximately 1057 mm downstream of the perforated apparatus discussed in this work; and neither The Babcock & Wilcox
plate). Company nor any of its employees shall be liable for any losses or
damages with respect to or resulting from the use of, or the inability to
The small difference in the pressure drop for T6A-T8 is at- use, any information, product, process or apparatus discussed in this
tributed to the fact that the model is not geometrically accurate work.
in the SCR region. In the NELS report it is stated that:

4 Babcock & Wilcox


Appendix A

Babcock & Wilcox 5


6 Babcock & Wilcox
Babcock & Wilcox 7
8 Babcock & Wilcox
Babcock & Wilcox 9

You might also like