Proof of Concept Test for a Real-Time Flare Combustion Efficiency Monitor

Paper from the AFRC 2013 conference titled Proof of Concept Test for a Real-Time Flare Combustion Efficiency Monitor by Yousheng Zeng

Real-time monitoring or even periodic testing of combustion efficiency (CE) of industrial flares remains a void in the current combustion and air emission measurement technology landscape. Zeng and Morris presented a new concept to monitor the flare CE in the American Flame Research Committee (AFRC) 2012 Annual Meeting in Salt Lake City, Utah.1 The Zeng and Morris paper described the theoretic basis of the concept and demonstrated feasibility of the approach through theoretic simulation. Since then the authors have conducted a proof of concept test to further evaluate the feasibility through experiment. This work is also supported by the U.S. EPA through its Small Business Innovative Research (SBIR) 2013 Phase I grant. The experiment and the results of the test are presented in this paper.

Page 1 Proof of Concept Test for a Real-Time Flare Combustion Efficiency Monitor Yousheng Zeng, PhD, PE and Jon Morris Providence, 1201 Main Street, Baton Rouge, LA 70802 Mark Dombrowski Surface Optics Corporation, 11555 Rancho Bernardo Rd., San Diego, CA 92127 INTRODUCTION Real-time monitoring or even periodic testing of combustion efficiency (CE) of industrial flares remains a void in the current combustion and air emission measurement technology landscape. Zeng and Morris presented a new concept to monitor the flare CE in the American Flame Research Committee (AFRC) 2012 Annual Meeting in Salt Lake City, Utah.1 The Zeng and Morris paper described the theoretic basis of the concept and demonstrated feasibility of the approach through theoretic simulation. Since then the authors have conducted a proof of concept test to further evaluate the feasibility through experiment. This work is also supported by the U.S. EPA through its Small Business Innovative Research (SBIR) 2013 Phase I grant. The experiment and the results of the test are presented in this paper. BRIEF REVIEW OF THE CONCEPT A typical flare burns unusable or unrecoverable hydrocarbons (HC). A generic expression of hydrocarbon combustion in a flare can be expressed as follows: CnHy + n O2 → nCO2 + H2O If the combustion is complete (i.e., CE=100%), the products of combustion would be CO2 and H2O. If the combustion is incomplete, there will be some unburned hydrocarbons and some intermediate combustion products, which commonly consist of CO. The flare CE can be determined by the following equation:2 (%) = [ ] Σ [ ] [ ] [ ] × 100% Eq. (1) 1 Y. Zeng and J. Morris, "A New Method to Measure Flare Combustion Efficiency in Real-Time", presented at the American Flame Research committee (AFRC) 2012 Annual Meeting, Salt Lake City, Utah, September 2012. 2 D.T. Allen and V.M. Torres, "2010 TCEQ Flare Study Project Final Report - Draft", Texas Commission on Environmental Quality PGA No. 582-8-862-45-FY09-04, May 23, 2011. Submitted to AFRC 2013 Industrial Combustion Symposium Page 2 Where ni is number of carbon atoms in hydrocarbon compound i (HCi). The terms in brackets [ ] on the right side of Equation (1) are for concentrations of compounds indicated by the subscripts and expressed in parts per million by volume (ppmv). Zeng and Morris demonstrated that the CE could be measured by the following equation: (%) = ( ) ( ) ̅ ( ) ̅( ) ( ) ( ) ( ) ( ) × 100% Eq. (2) Where: I = Intensity of IR radiance. Subscript 1 is used for hydrocarbons, 2 for CO2, and 3 for CO. α(λ) = Absorption coefficient of respective gases in the flare plume at wavelength of λ. B(T, λ) = Black body radiation at temperature T and wavelength λ. B(T, λ) is calculated by Plank's Law equation: The CE could be monitored in real-time through a multi-spectral IR imager that was capable of measuring IR radiance intensity at spectral bands corresponding to the wavelengths needed in Equation (2) above. EXPERIMENT The experiment set-up for the proof of concept test is illustrated by Figure 1. A bench scale flare was initially constructed from a propane cutting torch with a rosebud heating tip. While a cutting torch typically uses propane and oxygen to produce the desired flame, for our test bench the oxygen supply was replaced with nitrogen to reduce the British Thermal Unit (BTU) of the flare gas. This allowed us to use the regulators on the propane cutting torch to adjust the fuel (propane) and nitrogen mixture to achieve a range of CE. With our initial design the CE was quite high for all scenarios and we were not able to achieve poor CE. Although we mixed a large volume of nitrogen with the propane, we found that there was enough ambient air entrained with the propane/nitrogen mixture to completely burn the propane and achieve CE near 100%. As a result, we modified the design by replacing the rosebud heating tip with a larger diameter (~ 2.5 inches) iron pipe. This reduced the exit velocity and allowed for a wider range of CE, but not low enough for our purposes. We further modified the design to add a copper tube around the flame connected to a commercial steam cleaner to provide a low pressure steam curtain around the flame. The combination of a lower exit velocity and steam curtain allowed us to reduce the amount of ambient air entrained with the flame and achieve CE ranging from upper 50% up to 100%. Figure 2 shows a picture of the scale model flare with steam ring. Submitted to AFRC 2013 Industrial Combustion Symposium Page 3 Figure 1. Illustration of experiment set up for the proof of concept test. Figure 2. Scale model of industrial flare. The flare was placed on a QuickSet pedestal mount which allowed us to move the flare around and adjust the height. To capture and analyze the combustion gases from the flare, a four inch duct (flue) was placed above the flare with an extractive sampling probe. The flue height was Nitrogen Burner SOC750 Flue Imager CO2 and CO Analyzer HC Analyzer Datalogger Propane Steam Submitted to AFRC 2013 Industrial Combustion Symposium Page 4 also adjustable and was typically placed at a height just above the top of the flame. A thermocouple measured the temperature of the combustion gases as they entered the flue. Baffles were added inside the flue to create turbulence and ensure that the combustion gases were well mixed before they were sampled at the top of the flue. The sampling probe was placed approximately 14 inches above the bottom of the flue. Figure 3 shows the flue and sampling probe attached to the test bench. Figure 3: Flue and sampling probe. A Testo 350XL was used to sample the combustion gases and provided measurements for CO2, CO, O2, NOx, H2, and Temperature. The Testo provided data at a one second interval. Although the Testo 350XL also provided the ability to measure hydrocarbons, the accuracy was not high enough to meet the objectives of this test. A 3010 MINIFID Portable Heated FID THC Analyzer was added with a heated sample line to provide an accurate measurement of unburned hydrocarbon (propane) in the exhaust plume. The 3010 provided an analog voltage output 0 to 10V DC. Figure 4 shows the Testo 350XL (left) and the 3010 MINIFID (right). Submitted to AFRC 2013 Industrial Combustion Symposium Page 5 Figure 4: Testo 350XL (left) and 3010 Mini FID (right) A Campbell Scientific CR1000 data logger was added to the test bench to record the analog voltage from the 3010 Mini FID and the thermocouple at a scan rate of once per second to match the data interval of the Testo 350XL. The data collected from the extractive sampling system provided a high resolution measurement of the combustion efficiency, which served as a control for the data collected by the hyperspectral imager. Real time remote combustion efficiency monitoring requires both high frame rates and high spectral resolution. A staring Hyperspectral Imager operating in the proper wavelengths can provide the required imagery. To prove the concepts, we utilized the Surface Optics SOC750 Hyperspectral Imager. This imager is a cross between a scanning and a staring imager, providing a high frame rate (up to 32 frames or data cubes per second) and high spectral resolution (42 channels, 73 nm wide). For this test, the frame rate was set at 22 frames per second. It requires liquid nitrogen to bring the focal plane array to operating temperatures. The SOC750 was placed approximately 23 feet from the flare with an unobstructed line of sight to the entire flame. Figure 5 shows the SOC750 mounted on a tripod. Submitted to AFRC 2013 Industrial Combustion Symposium Page 6 Figure 5: SOC750 Hyperspectral Imager Several experiments were conducted to obtain the spectral imagery and extractive sampling data needed to complete the proof of concept test. The Testo 350XL (CO2, CO and O2) and the 3010 MiniFID (propane) were calibrated each day using certified calibration gases. The test bench was configured to provide real time extractive sampling data which allowed us to adjust the propane and nitrogen mixture to achieve various levels of CE between upper 50% and 98%. We found that steam was required to reduce CE below 97% so it was used in each scenario at a constant rate of 1.4 L/hour. A total of 28 scenarios were recorded for various experiment purposes. DATA REDUCTION AND ANALYSIS Using the experiment set up described above, a total of 28 spectral imagery data sets were collected. Some of them did not have matching extractive sampling data. Each spectral data set was collected in 30 seconds, and contained 670 data cubes (about 2 GB of data per set, a total of 56 GB of data). Thirteen data sets were considered primary data sets for proof of concept. Other data sets were collected for exploratory purposes and to aid the design of the prototype of the flare CE monitor in the next step of the development. Presented in this paper are the preliminary results of the 13 primary data sets. Our data reduction and analysis effort is still in progress, and results of other data sets will be reported elsewhere. Submitted to AFRC 2013 Industrial Combustion Symposium Page 7 The preliminary results of the 13 test runs are presented in Table 1. The data included in Table 1 is divided into two sections, one section for results based on extracted samples analyzed by conventional analyzers, and the other section for results based on data measured by the IR imager and the method developed for this technology. In the "Results Based on Analyzers" section, the "Temp. (F)" column is a 30-sec average temperature, in Fahrenheit, measured by the Testo 350XL at the sample probe, i.e., sample extraction point inside the flue (ref. Figure 3). The 30 seconds match the spectral data collection period of 30 seconds. The "CE (%)" column is calculated per Equation (1) using the CO2 and CO concentrations measured by the Testo 350 XL and the propane concentrations measured by 3010 Mini FID in the same 30-sec period. In the "Results Based on IR Imager and New Method" section, results presented are based on the IR spectral data collected by SOC750 and proprietary software and methods. The "CE w/o Calibration (%)" column is the results calculated straight from the IR radiance intensity measured in respective spectral bands selected for CO2, CO, and hydrocarbon. As demonstrated in Figure 6, this un-calibrated data shows a good correlation with true CE measured by conventional analyzers, indicating promising potential for this new technology. Table 1. Summary of proof of concept test Test ID Results Based on Analyzers Results Based on IR Imager and New Method Temp. (F) CE (%) CE w/o Calibration (%) Highest IR Radiance (W/m2-sr-μ) CE w/ Preliminary Calibration (%) Temp. Estimated by Imager (F) 1 887 98.1% 94.2% 142.40 98.8% 856 2 754 93.7% 89.2% 63.14 93.5% 755 3 819 90.1% 84.2% 67.30 88.3% 763 4 767 81.7% 74.7% 59.57 78.3% 747 5 315 57.2% 50.1% 0.71 52.6% 198 12 569 97.6% 90.3% 30.80 94.7% 666 13 591 95.3% 86.5% 22.78 90.7% 628 14 494 83.4% 70.9% 10.24 74.4% 529 15 420 75.9% 67.5% 4.85 70.8% 436 16 631 92.2% 83.5% 30.50 87.6% 664 16A 635 89.3% 82.8% 30.00 86.8% 662 16B 634 89.4% 80.1% 23.40 83.9% 632 16C 662 90.5% 86.2% 34.19 90.4% 679 Submitted to AFRC 2013 Industrial Combustion Symposium Page 8 Figure 6. Correlation between true CE measured by analyzers and CE determined by the new method without calibration. Based on the correlation, a preliminary calibration is applied to the un-calibrated CE and the results are listed in Table 1 under column heading of "CE w/ Preliminary Calibration". This calibrated CE is also compared with true CE in Figure 7. It is our intention to develop and evaluate more refined and practical calibration approaches that can be implemented for a commercial product of this technology. Figure 7. Comparison of true CE measured by analyzers and CE determined by the new method with a preliminary calibration. y = 1.0487x - 0.1149 R² = 0.9617 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 110.0% CE by New Method without Cal. True CE by Analyzers 50.0% 55.0% 60.0% 65.0% 70.0% 75.0% 80.0% 85.0% 90.0% 95.0% 100.0% 1 2 3 4 5 12 13 14 15 16 16A 16B 16C Combustion Efficiency Test Run CE by Analyzer CE by New Method Submitted to AFRC 2013 Industrial Combustion Symposium Page 9 Based on this preliminary study, we have observed that the new method works better when the highest IR radiance is sufficiently high, generally corresponding to a higher temperature. This can be seen in data included in Table 1. When the "Highest IR Radiance" listed in Table 1 is lower than 30 W/m2-sr-μ (Tests 5, 13, 14, 15, 16B), the IR signals are very week, particularly for CO2, which tends to result in a larger error in CE calculation. The IR radiance is a function of temperature. We have developed an approach to estimate the flare combustion temperature. The results are listed in Table 1 under the column heading of "Temp. Estimated by Imager (F)". A comparison of temperature estimated by this method and the actual temperature measured by Testo 350 XL is presented in Figure 8. This temperature estimation approach is preliminary. It will be further evaluated and more refined approach may be developed in the next phase of the technology development. Figure 8. Comparison of temperature measured by Testo 350 XL and temperature estimated by this new method. One of the distinctive features of this new technology is that it can image the entire flare flame and determine CE at a pixel level, therefore it can map out the combustion efficiency across the flare flame, as illustrated in Figure 9. The left side of Figure 9 is a colorized IR imagery of Test 2 run averaged over its entire 30 seconds run. The same imagery can also be made frame by frame. As an example, the first frame of this series (i.e., the first 1/22 seconds or ~45 ms of test) is presented in Figure 10. The right side of Figures 9 and 10 is the CE map generated by this new technology. The middle rectangular part is blank because the method does not apply due to the presence of the metal duct (i.e., the flue). The Region of Interest (ROI) # 1 indicated by the black border and labeled as ROI #1 represents the combustion product gases that exit the flue. This portion of combustion gases is also measured by the analyzers. Therefore the results from this ROI are tabulated in Table 1 and compared with the analyzer results. The CE maps in 0 100 200 300 400 500 600 700 800 900 1000 1 2 3 4 5 6 7 8 9 10 11 12 13 Temperature (F) Test Run Actual T (F) Est. T (F) Submitted to AFRC 2013 Industrial Combustion Symposium Page 10 Figures 9 and 10 are un-calibrated CE's. Several additional ROI are also indicated in the right side of Figure 9. ROI #2 is on the side of flue and has similar CE as the top of the flue. ROI #3 is closer to the lower center of the flame. The combustion in ROI #3 has not completed yet, thus a lower CE of 75%. ROI #4 is close to the top of the burner and shows even lower CE of 50%. Figure 9. Imagery and full flame CE map of Test 2 (30-sec average). Left: colorized IR imagery acquired by SOC750 hyperspectral imager. Right: CE map determined by this new method. Figure 10. Imagery and full flame CE map of Test 2, Frame 1. Left: colorized IR imagery acquired by SOC750 hyperspectral imager. Right: CE map determined by this new method. ROI #1: CE=89% ROI #2 CE=89% ROI #3 CE=75% ROI #4 CE=50% ROI #1: CE=82% ROI #2 CE=79% ROI #3 CE=69% ROI #4 CE=60% Submitted to AFRC 2013 Industrial Combustion Symposium Page 11 Figure 11 provides a visual comparison of the three species of interest, i.e., CO2, propane, and CO in the flame. Figure 11 includes images generated by SOC750 multispectral imager. Colors are added to spectral bands for CO2 (red), propane (green), and CO (blue). The image on the left has the same gain for red, green, and blue (RGB=1:1:1), and it shows domination by red (CO2). For the image in the middle, the signal for propane (green) is magnified by 16 times (RGB=1:16:1) in order to show the much smaller amount of propane. For the image on the right, the signal for CO (blue) is magnified by 64 times (RGB=1;1:64) in order to show the trance level of CO in the flame. It shows overwhelming amount CO2 forming an envelope around the flame (left image), higher concentrations of propane just above the tip of the burner and gradually fading as the combustion progresses (middle image), and trace levels of CO in a similar distribution as the CO2 envelope but with a smaller "footprint" than the CO2 envelope. Figure 11. IR images generated by SOC750 multispectral imager during Test 2. CONCLUSION An experiment was conducted to prove the concept of monitoring CE of industrial flare in real-time. Through the test, the following conclusions are drawn: 1. The concept of monitoring flare CE using a multi-spectral, high frame rate IR imager is valid. The CE determined by this method correlated well with the true CE measured by extracted samples analyzed by conventional analyzers. With a further refined calibration approach, more accurate CE measurement is expected. 2. This new technology can determine CE at a pixel level. Each pixel represents a portion of the flare flame in the field of view. With CE calculated for all pixels in a contemporaneous timeframe, a CE map can be generated to show the CE distribution across the flare flame and therefore the progression of the combustion process. 3. As a side benefit, the technology will also provide temperature mapping of the flare flame, which is also related to the performance of the flare. Submitted to AFRC 2013 Industrial Combustion Symposium Page 12 This work is just the first step in developing this technology. We will design and construct a prototype to further test the technology. We will also refine the calibration approaches, or consider additional calibration approaches, to account for the influence of atmospheric attenuation and other factors to make this technology more accurate and practical for large scale industrial deployment.