COMBUSTION EFFICIENCY OF FLARES AND THE EFFECT OF WIND By John H. Pohl, Energy International, Laguna Woods, CA Nicholas R. Soelberg, Idaho National Laboratory, Idaho Falls, ID James A. Seebold, Chevron (retired), Atherton, CA Presented at the 14th AFRC Symposium, Houston, TX, September 2014. ABSTRACT Wind can have a major effect on flare performance. Flares such as those on the gas fields between Calgary and Edmonton, Alberta, Canada have major impacts of wind, with wind speeds approaching 100 mph. A number of factors that control the efficiency of industrial flares have been partially indentified. Some for which experimental evidence is available are: * exit velocity, * fuel composition, * flare head design, * amount and kind of assist fluid and how it is added, * external wind speed and turbulence. This paper sets out the effect of these factors on flare performance. First, the factors other than wind are discussed from theory and experimental data. Next, the effect of wind is discussed from literature data and theory. The effect of wind and the magnitude of its effect are discussed and summarized. 1.0 INTRODUCTION Flare flames are used to destroy waste gases in the petroleum production industry, the petroleum refinery industry, the steel making industry (coke oven and blast furnace gases), and the chemical industry. Approximately, 16.1 Mton/yr. were flared from these sources in the US in 1982. The largest source of waste gases, 9.6 Mton/yr, were from blast furnace gases. Blast furnace gases are 76 % inert, Joseph, et al. 1982. 2.1 Effect of Exit Velocity According to the work done for the EPA in the 1980's, Pohl, et al. (1984), Pohl and Soelberg (1985), and Pohl and Soelberg (1986), the flare performance, is usually >98%, see Figure 1, if the heating value is greater than 1.2 times heating value required for flame stability. The criteria for flame stability is shown in Figure 2. In this case, flame stability is related to the heating value of the gas and the exit velocity of the gases out of the flare head. Flare heads greater than 3 inches in diameter behave similarly to lager commercial flare heads. The 3 inch rule has been confirmed by Gogolek (2000). This rule seems to the ratio of the perimeter area of the flare head available for diffusion to the flame area for burning the fuel. Such a relationship results in the flare performance being related to the reciprocal of flare head diameter. This relationship means the performance is very dependent of flare head diameter at small diameters, but is insensitive to diameter for large flare heads sizes. Figure 1. Combustion Efficiency of Flare Flames Versus High Heating Value of the Flared Gas. Pohl, et al. (1984), Pohl and Soelberg (1985), and Pohl and Soelberg (1986). Figure 2. Flame Stability, Exit velocity versus Gas Heating Value for 1/16-12 Inch Flare Heads. Pohl, et al. (1984), Pohl and Soelberg (1985), and Pohl and Soelberg (1986). The flame stability was correlated with the exit velocity of the flare head and the heating value of the gas. While this correlation successfully correlated the stability of several individual gases, it could not correlate the stability of gases with different physical properties. Neither could The adiabatic flame temperature, see Figure 3. A another possibility is that flame speed or some other parameter related to gas chemistry is a better correlation parameter. Such properties may be able to correlate the performance of different gases. This work purposely excluded the effect of wind. Trials were performed in a canyon early in the morning behind a wind screen. 2.2 EFFECT OF GAS COMPOSITION The combustion efficiency shown in Figure 1. is for the propane-nitrogen flares. Figure 3. shows the combustion efficiency for other gases investigated. In the 1980's investigation. we studied 1,3 butadiene, Ethylene Oxide, Hydrogen Sulfide in propane-nitrogen mixtures, and ammonia in propane-nitrogen mixtures. The combustion efficiency of these gases is shown in Figure 3. The correlation used for propane-nitrogen flames seems to be similar for other different gas composition. However, no correlation was discovered that could correlate all the gases with a single correlation parameter. Figure 4. Seebold Plot of Reynolds Number (dv/u) versus Richardson Number (gd/v2), Pohl and Soelberg (1985). 3.0 FLARE STRUCTURE Flare flames are turbulent diffusion flames whose structure is controlled by the Reynolds Number and the Richardson Number. These numbers represents the effect inertia forces versus the effect of buoyancy forces, see figure 4. High speed photographs can show the structure of these flames. A photograph of the flame at the Energy and Environmental Research Company facility is shown in Figure 5. This figure shows a picture of EER flare test facility. This figure shows a propane-nitrogen flame burning on a 12 inch commercial flare head. This facility intentionally excluded the effect of wind on flare performance. Three precautions were taken to minimize the effect of wind: * The facility was placed in a box canyon * The facility included a screen to minimize the effect of wind * Experiments were done early in the morning before the wind came up 3.1 FLAME LENGTH There have been a number of correlations of flame length. The correlation proposed by Hottle and Hawthorne (1949), is shown in figure 5. , has proved to be the best. This correlation has been adopted by the American Petroleum Institute (API). It also reasonably correlated the data in the flares studies by Pohl, et al. (1982,1984, 1985, and 1986). However, this correlation required that the coefficients be slightly adjusted. 3.2 STRUCTURE Figure 6. shows the structure of a propane -nitrogen flame burned in the EER Flare Test Faclity on 12 inch commercial flare head, Pohl et al. (1985). This figure shows that the eddies formed in the wind protected flame are approximately the size of the flare head. High speed movies of this facility shows that the eddies have a frequency of about 5 ms. Figure 6. Photograph of Propane-Nitrogen Flame Burning on a 12 Inch Commercial Head, Pohl, et al. 1985. EFFECT OF WIND The 1980's EPA study intentionally excluded wind. Information on the effect of wind comes from the literature and theoretic reasoning. The number of studies of industrial related flares are limited. One of the first studies of commercial flare was done in Germany by Siegle (1980). This study did include some effects of wind. This work was reviewed by EFFECT OF WIND Elevated flares are subject to major fluxuation in wind velocity. In clam conditions the flame flucatuates around the center-line of the flare head. As cross-wind increases the flame is bent in the direction of the wind; eventually becomes stabilized on the flare head. Wind also increases mixing of air into the flared gas. Such mixing can result in stripped eddies from the flame. Such stripping can result in decrease in combustion efficiency (CE). At higher wind speeds more air is induced into the flare and unburned material is increasingly stripped from the flame. The effect of wind has been studied in a limited cases. Wind has been studied in industrial flares by Siegel (1980). A research on flare heads up to 6 inches in a wind tunnel was performed by Gogolek (20 ). He correlated the combustion efficiency with MFR, momentum flux ratio of wind to exit velocity. He concluded that flames were stabilized behind the stack for MFR >3. Such flames stabilized in this manner have different combustion stability than normally stabilized flames. Characteristic small flame in wind have also been studied by Howes (1981). Siegel (1980) reported that wind reduced the combustion efficiency of industrial flares in a 6.7 mps (4.16 mph) reduced the combustion efficiency (CE) from 99 % to 97 % at the same position in the flame. Wind effects have been correlated with the ratio of wind momentum and gas momentum. Seebold has reviewed the CANMET study and concluded that the correlation did not work well for industrial flares, Seebold (2011). Photographs of the Effect of Wind on turbulent diffusion buoyant flames as show in Fig. 7 and 8. Figure 5. Photograph of an Kuwait Oil Fire in the Near Absence of Wind. Performance of flares is controlled largely by the nature of the fuel and the design of the flare head to increase mixing. Design of the flare head is thought to be critical to maintaining the stability of the flame. Flares are mainly divided into: * Elevated * Ground Flares * Air Assisted Flares * Steam Assisted flares * Pressure Flares The flare heads are designed to stabilize the flame. Ground flares are burned in pits. Air assisted flares are usually fired with a center stream of air. Steam assisted flares can be burned with a central steam, but steam is more commonly injected with a ring of steam injectors surrounding the flare head. The steam jet induce air into the steam which then increases the air fuel mixing. Flares are turbulet diffusion flames. Where the fuel enters from the flare head and the oxygen entrains with air from the outside. Use of air-assisted or steam assisted flare helps with the air mixing into the fuel. In the case of steam, the steam serves as a means to induce air into the flame. TCEQ (Texas Commission on Environment Quality) presentation in 2009, listed some parameters that controlled the structure of flare flames:  Wind  Ambient Temperature  Humidity Pohl, et al.(1980,1983,1964,and 1985) added several more conditions that affect flame structure of flare flames:  Head Exit Velocity  High Heating Value of the Gas Burned  Eddy Size~D  Eddy frequency~5 ms Figure 7. Large Kuwait Oil Flame in the Absence of wind. FLAME LENGTH There have been a number of attempted correlation of flame length with physical properties. The most useful correlation was developed by Hottle and Hawthorne (1947). This correlation was later adopted by API (American Petroleum Institute. This correlation also worked reasonably well with the data of this study Pohl, et al. (1984), with a slight alteration of the coefficients. A comparison of these correlations is shown in Figure 6. Figure 7. Large Kuwait Oil Fire Affected by Wind. 2.0 CORRELATION OF FLARE PERFORMANCE Other correlations have been proposed to predict flare performance. The most used correlation for stability of the flame is shown in Figure 2. The adiabatic flame temperature has also been tried as shown in Figure 3. Neither could correlate the flare performance with physical properties of different gases. Another correlation, shown in figure 9., has been tried that includes a temperature (HHV/Cp2700 F), and the ratio of flammability limits (UFL/LFL). This correlation seemed to work reasonably well for most gases. This correlation also worked reasonably for the data of the Pohl, et al., (1984), Pohl and Soelberg (1985), and Pohl and Soelberg (1986). However, the correlation proved to be highly sensitive to the parameters, and to have significantly scatter. The correlation was also checked against the EER (Energy and Environmental Research Corporation) data. This correlation also worked well, but was still highly sensitive with a lot of scatter for a number of gases. A new correlation is needed to collapse the date for different gases. The correlation parameter has to incorporate the combustion properties of the gases. Flame speed seems to be a logical parameter,( a Pohl, et al. (1984), Pohl and Soelberg (1985), and Pohl and Soelberg (1986). The combustion efficiency of the different gases is shown in Figure 10. Figure 9. Correlation of Flame Stability, Nobel, Keller, and Schwartz (1984). Figure 10. Combustion Efficiency of different Gases Versus Heating Value required for Stability, Pohl et al. (1985). 2.1 Effect of Exit Velocity The exit velocity and high heating value of the gases has been correlated with flame stability for several different gases, Pohl and Soelberg (1985). This correlation worked for different gases on heads larger than 3 inches, however the curves were not the same for different gases. 2.2 Effect of Gas While the effect of all gases could not be correlated; Individual gases were attempted to correlated with adiabatic flame temperature, see Figure 3. In figure 3. the adiabatic flame temperature successfully correlated the limiting velocity for stability for different gases, but not the same correlation for all gases. The gases correlated were C3H8-N2,1,3-butadien, Ammonia-C3H8-N2, Ethylene Oxide, and H2S-C3H8-N2. Figure . The Combustion Efficiency of Different Gases with Heating Value Stability. Figure . The Limiting Velocity versus Adiabatic Flame Temperature for Different Gases, Pohl and Soelberg (1986). 2.3 Effect of Wind The effect of wind like many other properties of flare have been poorly investigated. Siegle (1980) investigated a refinery gas flare burning gas with 45-60% hydrogen on a Flaregas Co. FS-4-6 anti-pollution flare. This was a six inch commercial flare operated at 49-178 MBTU/hr of gas. Other smaller tests of the effect of wind have been done by Palmer (1972), Lee and Whipple (1981), and Howes, et al (). ( Palmer (1972) used a 1/2 inch flare, 4 feet off the ground. The study burned ethylene at 0.4 - MBTU/hr. He also studied the effect of steam injection. Lee and Whipple (1981) studied bench scale propane flares. Lee and Whipple (1981) used a 2 inch flare head with 1/16 center hole. The head also had 2 rings with 16, 1/8 inch holes and two rings of 16, 3/16 inch holes, The velocity of 3 ft/s or 0.3 MBTU/hr. Lee and Whipple (1981) did not study the effect of wind or steam on flare performance. Howes et al. () studied the performance of two small commercial flare heads. The flare heads were manufactured by John Zink and Co. and were an LH an assisted flare head and a LRCO (liner relief gas oxidizer. The LRGO contained 3 burners, 3 feet apart with a combined firing rate of 27 ft/s or 4200 lb/hr of natural gas equivalent to 83.7 MBTU/hr. This Study had exit velocity of 27 ft/s and a firing rate of 44 MBTU/hr. The DMGK, German EPA, stated in their regulations that ground flares are 95 % efficiency and elevated flare are 75 % efficiency. Studies be Pohl, et al. (1982, 1984, 1985, 1986) Seebold (2011) has summarized these studies as well as the recently proposed rules by the EPA. He states that there is no evidence that wind speed less than 22 mph affect combustion efficiency and that MFR (momentum flux Ratio) was a poor correlative parameter for the effect of wind speed on combustion efficiency of flare flames. Wind can seemly have a large effect on combustion efficiency. In the absent of wind, combustion efficiency is reduced when eddies are stripped from the flame and quenched by over induction of air. In normal flares, the eddies are about the size of the flare head and have a frequency of about 5 ms, Pohl, et al. (1985). Wind encourages over induction of wind and quenching of eddies. In addition, wind can push the flame in the direction of the wind and eventually stabilize the flame behind a circular cylinder. Figure 8. shows the aerodynamics of induction of air into a flames. The time to diffuse a distance l is t=l/v Figure 8. Aerodynamics of Eddies, Seebold (2011). There have been few valid reports on the influence of wind on commercial flares. The one example is Siegel (1980) who studied the emission from 6 inch operating refinery flare including the effect of wind. CANMET, Gogolek (20 ) has studied up to six inch flares in a wind tunnel, The wind velocities studied varied up to kmp. Seebold (2012) has summed up the evidence available on the effect of wind on flare flame combustion efficiency. He suggested that there is insufficient evidence to support the EPA position on the effect of wind on flare flame combustion efficiency that wind reduces combustion efficiency if the wind velocity is below 22 mph. Seebold (2012) claims that no evidence exists of wind induced loss of combustion efficiency. Seebold states that he is unaware of any data on industrial flares that shows that wind decrease combustion efficiency. Seebold (2012) states that MFR (ratio of wind momentum to flare momentum) >3 yields a flame stabilized behind the stack in a wind tunnel flares. However, MFR is not useful in correlating combustion efficiency of industrial or other flares. . . Figure 3. Adiabatic Flame Temperature Versus Maximum Velocity for Flame Stability. Pohl, et al. (1984), Pohl and Soelberg (1985), and Pohl and Soelberg (1986). 2.1 FLAME STABILITY Figure 1., Pohl, et al. (1984) shows the stability curve as the relationship between high heating value and the exit velocity of the flared gases. The cross hated area includes 12 inch diameter commercial flare heads. The correlation worked for flare heads larger than 3 inches in diameter. The cross hatched area is the stability of 3, 6, and 12 inches diameter flare heads including commercial 12 inch diameter heads. The performance of flare heads larger than 3 inches in diameter can be scaled. This correlation was proposed by Pohl et al. (1984) and has been confirmed by Gogolek, et al. (2000). Flare performance has been correlated using flame stability; adiabatic flame temperature has also been tried. Neither could correlate the performance of different gases. 2.2 FLARE HEAD DESIGN The stability of some commercial heads are compared with the stability of pipe heads tested in this study. Several commercial flare heads were tested, John Zink, McGill, and Flare Gas (UK). Pohl and Soelberg (1985). The normal flame stability of flares tested in this study is shown as the cross hatched area. The commercial flare heads tested are slightly more stable than the pipe heads tested in this study. For instance, at an exit velocity of 10 ft/sec, the stability of heads tested required a gas heating value 400-500 BTU/ft3, while Flare Gas's Conada head required about 300 BTU/ft3, the pressure flare tested required between 300 BTU/ft3, and 100-200 BTU/ft3. REFERENCES Joseph, D., J. Lee, C. McKinnon, R. Payne, and J.H. Pohl, "Evaluation of the Efficiency of Industrial Flares Used to Destroy Waste Gases: Phase I Interim Report-Experimental Design, EPA Contract No. 68-02-3661, January, 1982. Hottle, H. C. and W. R. Hawthorne, Third Symposium on Combustion, The Williams and Wilkins Company, P. 254, 1949. Howes, J.E. Jr, T.E. Hill, R.N. Smith, G.R. Ward, and W.F. Herget, "Field Test Plan for Development of Flare Emission Measurement Methodology," EPA Contract No. 68-02-2686 (1979). Gogolek, Lee, K.C. and G.M. Whipple, Waste Gas Hydrocarbon Combustion in a Flare, Union Carbide Corporation, Charleston, WV, 1981. Nobel, R.K., M.R. Keller, and R.E. Schwartz, "An Experimental Analysis of Flame Stability of Open Air Diffusion Flames", AFRC International Symposium on Alternate Fuels and Hazardous Waste, Tulsa, OK, 1984. Pohl, J.H., R. Payne, J. Lee, "Evaluation of the Efficiency of Industrial Flares: Test Results", EPA -600/2-84-095, May 1984. Pohl, J.H. and N.R. Soelberg, "Evaluation of the Efficiency of Industrial Flares: Flare Head Design and Gas Composition", EPA -600/2-85-106, September 1985 Pohl, J.H. and N.R. Soelberg, "Evaluation of the Efficiency of Industrial Flares: H2S Gas Mixtures and Pilot Assisted Flares", EPA -600/2-86-080, September 1986. Seebold, J.G., "Parameters for Properly Designed and Operating Flares", Peer Review of a Report by: U.S. EPA's ffice of Air Quality planning and Standards (QAQPS), May 2012. Seebold, J.G., B.C. Davis, P.E.G. Gogolek, L.W. Kostiuk, J.H. Pohl, R.E. Schwartz, N.R. Soelberg, M. Strosher, and P.M. Walsh, "Reaction Efficiency of Industrial Flares-Perspective of the Past", Combustion Canada 03, Vancouver, BC, September 2003. Siegel, K.D., "Degree of Conversion of Flare Gas in Refinery High Flares", Ph. D. Dissertation, University of Karlsruhe (Germany), February 1980. .