White Paper on Multi-Spectral Infrared Camera for Flare Efficiency Measurement

Flares are widely used in chemical process industries (e.g., petroleum refineries, chemical plants, etc.). Due to the intended function and nature of flare design and operations, determination of flare Combustion Efficiency (CE) and Destruction and Removal Efficiency (DRE) is extremely challenging. There has been a protracted debate on how much air pollutants are emitted from flares. The fact is that no one has a good answer to this question and this level of uncertainty regarding flare emissions is problematic for both regulators and industry.

1 White Paper on Multi-Spectral Infrared Camera for Flare Efficiency Measurement Original: March 27, 2012; updated May 3, 2012 Background and Problem Definition Flares are widely used in chemical process industries (e.g., petroleum refineries, chemical plants, etc.). Due to the intended function and nature of flare design and operations, determination of flare Combustion Efficiency (CE) and Destruction and Removal Efficiency (DRE) is extremely challenging. There has been a protracted debate on how much air pollutants are emitted from flares. The fact is that no one has a good answer to this question and this level of uncertainty regarding flare emissions is problematic for both regulators and industry. In 2010, Texas Commission on Environmental Quality (TCEQ) contracted University of Texas at Austin (UT) to conduct a comprehensive study on flare CE and DRE. The field work was conducted at John Zink facility in Tulsa, Oklahoma. The report for this study is available for download at the TCEQ website and a link is provided here for reference: (http://www.tceq.texas.gov/assets/public/implementation/air/rules/Flare/TCEQ2010FlareStudyDraftFinalReport.pdf). The results from this study are very valuable in characterizing flare CE and DRE and had a lasting impact on flare operations and emission management. It should be noted that the study was successful in characterizing flare efficiency under the specific conditions targeted by the experiment design, however it did not cover flare operations under upset or emergency conditions, hydrogen flares, or flares specifically designed for routinely low flow applications. The TCEQ-UT flare study was a major undertaking. The method used could be referred to as "grab and measure" - see the photo on the right for the study setup. It is not practical to use the same approach to measure or monitor flare operations on a regular basis. The TCEQ-UT study did include two supplemental remote sensing based measurement systems with an intention to evaluate their effectiveness for practical flare monitoring. The two systems were: 2  Infrared (IR) Hyper-Spectral Imager by Telops Inc. (Hyper-Cam)  Passive and Active Fourier Transform Infrared (PFTIR and AFTIR, or FTIR for either) Spectroscopy by Industrial Monitor and Control Corporation (IMACC) The study results suggested that the flare CE determined by IMACC PFTIR/AFTIR was generally consistent with the CE determined by analysis of pre- and post-combustion gas samples thru the "grab and measure" method. The mean differences between the two were about 2-2.5% and average standard deviations were 2.8-3.2%. The data availability was 99-100%. The performance of Telops Hyper-Spectral Imager was less desirable. The mean differences were 19.9%, standard deviations were 57.8%, and data availability was 39%. Both the Telops Hyper-Spectral Imager and the IMACC FTIR are powerful instruments for many applications, particularly research projects. However, they have some significant shortcomings if they are to be used as industrial analyzers to determine flare CE. These shortcomings are identified below: Fundamental/Technical Issues:  Telops' Hyper-Cam can be considered a two-dimensional array of FTIR spectrometers that can be combined to form images (i.e., each pixel in the image is equivalent to a single FTIR spectrometer). It has a scan rate of approximately 1 second per scan (depending on spectral resolution and other parameter settings). The flare plume changes rapidly in shape and position, and the resulting path length of the pixels in the Hyper-Cam imager may change dramatically within the same data cube. This variability introduces unknown and uncontrollable factors into the pixel intensity-concentration equation, rendering calculations and results unreliable.  IMACC's FTIR is a path measurement instrument. The results only represent the region where the IR light path intersects the flare plume. Due to the heterogeneous and dynamic nature of a flare, using the measurement from a small path to represent the entire flare is a concern. The IMACC FTIR also has a relatively long scan time (seconds) and suffers the same problem as the Telops Hyper-Cam. Since the IMACC FTIR is a single-path measurement instrument, this variability can be minimized by pointing the instrument to the middle, thick portion of the flare plume where the relative change in path length is small. If the IMACC FITR is aimed at the fringe of the flare plume, or if the flare diameter is small, the effect of this temporal mismatch due to flare plume dynamics is expected to be much more salient and problematic. Selection and alignment of the measurement path could significantly influence results. This makes it impractical for routine monitoring as the system would need some sort of targeting system to ensure it is consistently aimed at the correct position in the flare plume while the plume may be constantly shifting in wind. Practical/Implementation Issues:  Both the Telops Hyper-Cam and the IMACC FTIR are delicate research instruments and require expert-level personnel to operate. They require significant effort to set up and 3 maintain, and significant effort is required for post-processing/analyzing data in order to derive flare CE results. They do not provide real-time or near real-time measurements, and are not suitable instruments to provide continuous real-time feedback to operational personnel.  The total ownership cost is very high (this is particularly true for Telops' Hyper-Cam). Solution Proposed by Providence Providence is proposing development of a 4-band mid-wave IR staring imager (camera) for flare CE measurement that will overcome the problems described in the previous section and be effective even under upset, emergency, or routinely low flow conditions not covered by the TCEQ-UT study. The use of this new IR camera will be analogous to the GasFindIR cameras manufactured by FLIR Systems, Inc. (FLIR). However, the capability and application of the camera will be drastically different. While the FLIR camera has a single spectral band, the proposed IR camera will utilize a special arrangement of micro lens array (MLA) optics which allow the camera to simultaneously image four spectral bands, each capable of imaging one gas as follows:  Band 1 - Carbon Dioxide (CO2)  Band 2 - Carbon Monoxide (CO)  Band 3 - Hydrocarbon (HC)  Band 4 - Background reference or spare (can be used as general background signal for correcting effect of sky/cloud condition changes, steam plume, etc.; or for a gas of particular interest) The camera will have a frame rate of 30 frames per second (fps) for each band, i.e., 30 data cubes per second. Each data cube has two spatial dimensions and one spectral dimension providing four two-dimensional sub-frames, each at a rate of 30 fps. Actually every "frame" in a conventional sense will consist of four "sub-frames". All sub-frames will have the same field of view and be spatially and temporally synchronized. Dividing the sensor in this manner provides four identical fields of view in different spectral bands. The video stream generated through Band 3 (i.e., series of sub-frame 3) will be similar to the video generated by the FLIR GasFindIR camera, and it will be used to image hydrocarbons. Similar video images of CO2 and CO gases will be generated through Bands 1 and 2, respectively. The operators can choose to view images of one of these gases by pressing a selection button. The images for all four bands are also captured and stored for further analysis. Through sensor design, the camera will have a large dynamic range (linearity range) to prevent signal saturation without sacrificing sensitive for low concentrations. From linearity viewpoint, this range will be sufficient to calculate CE to the accuracy of at least 0.1%. Similar to the IR images generated by the FLIR GasFindIR camera, the proposed camera will image the entire flare plume (or a substantial portion of the flare plume). Each pixel in the image will represent a small portion of the flare plume. Computer vision algorithms will identify the 4 boundaries of the plume and the intensity of each pixel inside the plume will be proportional to (1) the concentration of the target gas and (2) the effective path length through the plume. The flare CE can be determined by the following equation (ref. the TCEQ-UT Tulsa flare study report): ( ) ( Σ ) Eq. (1) All terms on the right side of Equation (1) are for concentrations of indicated compounds in parts per million by volume (ppmv). The three compounds (i.e., CO2, CO, and combined HC) are measured by spectral Bands 1-3, respectively. Each pixel represents a small volume of the flare plume, and is equivalent to the path measurement made by the IMACC PFTIR. The difference between this instrument and the IMACC PFTIR is that IMACC PFTIR scans a wide spectral range through a single path length, whereas this instrument images each gas in each of four fixed spectral ranges for each pixel in the field of view. As a result, the proposed camera can achieve a frame rate of 30 fps with complete spatial and temporal synchronization, therefore overcoming the technical problems of IMACC FTIR and Telops Hyper-Spectral Imager described in the previous section. Because the measurement is done for the entire flare plume, the issue of PFTIR or AFTIR regarding aiming at the right point of the flare is no longer a concern. The intensity of each plume pixel in the image captured by the new device can be expressed as: Eq. (2) Where I = Intensity of signal at each pixel in the sub-frame corresponding to each spectral band. ε = Absorption/emission coefficient specific to each compound in the narrow spectral window represented by the corresponding narrow bandpass filter in the camera. C = Concentration of the compound in the small volume of flare plume represented by the pixel. L = Path length of the small volume of flare plume represented by the pixel. Equation (2) can be rearranged as follows: Eq. (3) Substituting Eq. (3) into the concentration terms in Eq. (1), the resulting equation is: ( ) ( Σ ) Eq. (4) 5 For each small volume of the flare plume represented by a pixel in the image, the path length, L, is identical for all three gases. Therefore, L is canceled and Eq. (4) can be simplified as Eq. (5): ( ) ( Σ ) Eq. (5) The three coefficient (ε) terms for the three compounds are wavelength-dependent constant and pre-determined based on the spectra of these compounds. The term εHC represents a weighted average of constants of common hydrocarbon compounds typically found in the vent gas could be adjusted by the user based on major constituents expected in the vent gas. Although the coefficients ε will be affected by temperature, impact to the three compounds in the same region of the flare will be similar and therefore cancelled out. CO2 in the atmosphere is expected to be approximately 390 ppm. Coupled with the long optical length between the flare and IR camera, atmospheric CO2 would significantly attenuate the intensity of CO2 signal from the flare, therefore interfering with the proposed method. This interference can be avoided by using the so called "red spike" and "blue spike" of CO2. Spectral bands of CO2 widen when CO2 gas is heated. In contrast, the spectral bands of atmospheric CO2 remain unchanged. When IR light passes through atmosphere, the center portion of the CO2 bands is attenuated due to absorption of atmospheric CO2, forming red and blue spikes on each side of the widened CO2 bands. Either the red or the blue spike can be selected to avoid attenuation caused by atmospheric CO2. Similar methods are currently used to identify exhaust plumes from jets and missiles in military tracking applications. Alternatively, the effect of atmospheric CO2 could be corrected based on a known concentration of CO2 (390 ppm) and a known distance between the camera and flare. Once the three intensity (I) terms in Eq. (5) are measured by this new device, the CE for the small volume of the flare plume represented by a given pixel can be calculated. The same calculation can be done for all pixels in an image using computer vision techniques applied within the firmware of the camera. The results will be a CE contour map covering the flare plume in the image. The figure on the right is a simplified illustration of this concept. Various averages across the flare plume outer envelope and/or time will then be calculated to characterize the overall flare efficiency. These averages will be available on a real-time basis and can be used as metrics to monitor the flare performance. With the addition of an industrial interface, the camera itself can be incorporated into a facility's process instrument network to 999998989998979790929095959598987098989898989895959598989797959594949494947060 6 provide continuous real-time flare CE information to operators in the control room. In summary, the proposed flare efficiency analyzer will be:  A 4-band IR camera monitoring [concentration x path length] of CO2, CO, hydrocarbon, and background in flare plume at a frame rate of 30 fps, and the path length is cancelled out in the CE calculations;  Easy to deploy either as a portable device or fixed installation;  Easy to operate by trained technicians (does not require expert training as required for any FTIR-based instrument);  Suitable for various types of flares operating under steady state conditions or conditions not covered by the TCEQ-UT flare study;  For fixed installation, it will be installed in a similar fashion as other common process instruments at chemical processing facilities (e.g., pressure transducers, flame detectors, etc.) and can be integrated into existing instrument system through a common data protocol;  One device can monitor multiple flares at a facility using "step and stare" methodology;  Anticipated output: o Four channels for video streaming of raw video for each of the 4 spectral bands (i.e., video images of CO2, CO, HC, and background), o One channel for continuous output of processed data in the form of a CE color contour map, o Digital or analog readout of selected key flare performance metrics, which can be integrated into facility DCS or other data systems (MODBUS TCP); and  Cost competitive compared to a typical PFTIR or AFTIR instrument and much more cost competitive compared to a hyper-spectral imager. Providence Credentials Providence is an environmental engineering and consulting firm with 170 employees headquartered in Baton Rouge, Louisiana. Providence has branch offices in Houma (LA), Shreveport (LA), Dallas (TX), Houston (TX), Gulfport (MS), and Fresno (CA) along with a few satellite offices in the gulf coast states. Providence has pioneered multiple new technologies and innovative products (both hardware and software) to solve problems in the areas of air quality monitoring and management. Providence has been engaged in research and development of IR imaging technology since 2006. Providence possesses in-depth expertise that spans vertically through the development and prototyping cycle of IR gas imaging devices, including IR sensors, optics, filters, spectral applications, electronics, data communication, advanced plume recognition algorithms, enclosures, and other deployment related hardware. Providence's credentials in the IR imaging R&D area as well as air monitoring in general are highlighted below. 7  In 2006, Providence developed the concept of the Third Generation Leak Detection and Repair or LDAR3. It can be viewed as Continuous Emission Monitoring Systems (CEMS) for fugitive VOC emissions. Since then Providence has successfully developed autonomous remote gas leak detection systems. Recently (March 2012), Version 2 of such a system was successfully field tested for a confidential client. Providence is currently working on Versions 3 and 4 which significantly extend the capabilities of the system and represent the current state of the art in this field.  Providence has conducted more than $3 million worth of research and development in the field of IR imaging over the past four years. Ongoing R&D efforts currently exceed $1 million annually.  Providence's inventions in this field have earned one patent with two more patents pending.  Providence markets 16-band cameras, very similar to the one proposed here for flare monitoring but in different spectral regions. With additional product development effort, the technology can be adapted to this new instrument.  Providence pioneered the air monitoring method of operating Auto-GC in trigger mode aimed at study of un-conventional air releases.  Providence designs and installs CEMS to monitor emissions from a variety of sources in the chemical process industry. Our expertise includes design specification for sampling systems and probes, installation, maintenance, operation and validations of CEMS systems.  Providence has designed and constructed multiple state-of-the-art mobile air monitoring systems and accompanying web-based data management and visualization systems. These products have been in operations for many years. Providence maintains current knowledge regarding technologies for IR gas imaging and its applications on a global scale. This white paper does not include all technical details. For further information, contact Yousheng Zeng, PhD, PE Managing Partner Providence 1201 Main Street Baton Rouge, LA 70802 225-766-7400 [e-mail]