Electrostatic Precipitator Performance Improvement Through Computational Fluid Dynamic Modeling PowerGen International December 11, 2002 Authors: Brian J. Dumont Robert G. Mudry, P.E. 1
Introduction Flow modeling is an established practice to optimize gas flow patterns within industrial equipment Little published data exists on the accuracy of flow models, particularly for electrostatic precipitators (ESPs) 2 Available data was analyzed in detail to compare model results to actual plant test data
Fluid Flow Modeling Computational Fluid Dynamics (CFD) Physical scale modeling 3
Testing Methods ESP cold-flow velocity distribution measurement Flow Flow 4
Objectives of Analysis Assess all available data for ESP testing and modeling acquired over the past 7 years ESP field test data CFD model results Physical model results Perform statistical comparisons of the data Obtain quantitative information relating model correlation to test data 5
Cases Analyzed Ten cases where field data and CFD data exist for the same configuration Five cases where corresponding physical model data also exist All cases from coal-fired electric power stations U.S. and Canadian plants Unit size ranges from 326 MW to 952 MW One case study is presented in detail, 326MW Unit in the Western U.S. 6 All cases are summarized
Data Comparisons Contour plots Flow distribution statistics % RMS ICAC Standards Point-by-point deviations Overall Correlation Factor (%RMS of point-by-point deviations) 7
Case Study 2 - Inlet Plane Normalized Streamwise Velocity Component Test Data Physical Both models capture trends CFD model captures peak velocity better, but overpredicts the size of the high velocity region CFD 8
Case Study 2 - Inlet Plane Flow Statistics - Deviations from Goal Velocities 100 90 80 70 60 Physical Model Test Data CFD Model 86.5 63.5 61.5 99.7 93.7 91.0 91 Physical model overpredicts flow compliance to goal as shown by all 3 analyses 50 40 30 20 20.1 20.1 33.4 33.4 42.0 42 CFD model underpredicts flow compliance to goal as shown by all 3 analyses 10 9 0 %%RMS Modified ICAC ICAC Modified ICAC m 115% ICAC m 140%
Case Study 2 - Inlet Plane Histograms Point-By-Point Deviations from Test Data 10 Correlation Factors: CFD: 37.0 Physical: 32.4 65% of CFD points within +/-25% band 61% of physical model points within same band Percentage of Data Points 30 25 20 15 10 5 CFD Model Physical Model 0-100 -80-60 -40-20 0 20 40 60 80 100 Percent Deviation from Test Data
Case Study 2 - Outlet Plane Normalized Streamwise Velocity Component Test Data Physical Both models capture trends CFD model captures peak velocity better CFD 11
Case Study 2 - Outlet Plane Flow Statistics - Deviations from Goal Velocities 100 90 80 70 60 Physical Model Test Data CFD Model 72.6 72.6 78.8 76.7 92.0 90.6 86.8 Both models underpredict %RMS, especially the physical model 50 40 30 20 17.8 17.8 34.7 27.8 27.8 Both models predict modified ICAC conditions fairly well 10 12 0 %%RMS Modified ICAC ICAC Modified ICAC m 115% ICAC m 140%
Case Study 2 - Outlet Plane Histograms - Point-By-Point Deviations from Test Data 13 Correlation Factors: CFD: 27.2 Physical: 40.8 75% of CFD points within +/-25% band 48% of physical model points within same band Percentage of Data Points 30 25 20 15 10 5 CFD Model Physical Model 0-100 -80-60 -40-20 0 20 40 60 80 100 Percent Deviation from Test Data
Case Study 2 - Summary Œ Œ Œ ESP Inlet Both models correlate fairly well Physical model has a lower correlation coefficient CFD model matches flow statistics better and has a larger number of points within +/-25% deviation band on histogram Œ Œ ESP Outlet CFD model agrees better with test data under all comparisons Both models capture correct trends 14
All Case Studies - Correlation Factor In some cases, the CFD model has a better correlation factor In others, the physical model is better There is no clear trend as to why this occurs The CFD model correlates better on average Correlation Factor Correlation Factor 45 40 35 30 25 20 15 10 5 0 Correlation Factor Summary Correlation Factor Summary CFD Model Physical Model 1I 1O 2I 2O 3I 3AB 3O 4AB 5O 6I 6O 7I 7O 8I 8O 9I 9O 10AB Case 15
All Case Studies - Overall Summary On average, the Correlation Factor for CFD models is 23.5 (27.8 using only studies where the physical model also existed) On average, the Correlation Factor for physical models is 32.6 16
Analysis Conclusions On the whole, correlation is not as strong as desired Experience indicates that this level of correlation is enough to make CFD modeling a useful engineering tool for ESPs Additional research and development in CFD technology will allow for increased accuracy 17
Case Study Two identical 790 MW units Western U.S. Cold side chevron ESPs SCA=125 Avg. vel 7.2 ft/s Both units regularly derated by 240 MW to operate within opacity limits 18
Baseline CFD Model Results Velocity pattern at ESP inlet shows non-uniform High velocity on top and bottom, low velocity in center RMS deviation = 26 % Outlet plane poor distribution as well RMS deviation = 20 % 19
Design Optimization Design modifications developed using CFD model New inlet and outlet perforated plates Slight alteration to inlet duct turning vanes 20
Final Design CFD Model Results ESP inlet velocity profile now uniform RMS deviation = 7 % Outlet plane also improved RMS deviation = 7 % Change to system pressure drop = +0.3 H 2 O 21
Resulting ESP Performance ESP efficiency testing performed 23% reduction in particulate emissions compared to original ESP geometry At same opacity, plant output increases by 150 MW per unit Payback for model study and installation costs = 1 year 22
Questions 23